← back to all models ← RevenueBench home
unbiased

unbiased/pareto-26.9

Mean 0.982 · 37/40 perfect tests · $8.56 total · modalities in: text, image · out: text · each card: the prompt → the correct answer (gold) → this model's actual answer

What do the modalities mean?

Modalities in means what you can send this model: text, images, files, video, or audio. Modalities out means what it can send back. Bench tasks feed text extracts, so modalities do not change the score. They are here so you know what this model can handle in real work.

data-crm 5 tests
stage-hygiene-auditPASS 1.00
Plain-English question: How much of the open DS2+ pipeline is missing a substantive why-buys, and is anything missing DS1 timestamps or out of order?
Prompt — what the model saw
Audit stage hygiene on the open pipeline below. Rules: a deal at DS2 or later must have a substantive why-buys (15+ characters, not a placeholder such as tbd, n/a, see notes). A why_buys value of the form LONG(n) means the real text was present but truncated for display; treat it as present and substantive. Every deal should have a DS1 entry timestamp. A deal whose stage entry timestamps are out of order regressed (an earlier-stage timestamp is later than a later-stage timestamp). Report: percent of DS2+ deals non-compliant on why-buys overall and by owner (use the owners table for names), count of deals with no DS1 timestamp, count that regressed, and three process fixes. State definitions before numbers.
cost $0.3166 · 157s · in 14,703 / out 36,735 tok
✓ noncomp_pct✓ no_ds1_0✓ regressed_0✓ per_owner_breakdown✓ definitions_first✓ three_fixes

✓ Correct Answer (ground truth)

Plain-English answer: 81.5% of the 124 DS2+ deals are missing a substantive why-buys — Alex Franklin 48, Cole Ingram 19, Dana Mercer 18, Bryce Harmon 15, Farid Osman 1 — and every deal has a DS1 timestamp with none regressed.
The answer in one line: n_deals = 156; ds2plus = 124; noncomp_pct = 81.5; bryce_noncomp = 15; no_ds1 = 0; regressed = 0 (+1 more below)
n_deals156
bryce_noncomp15
ds2plus124
no_ds10
noncomp_pct81.5
regressed0
per_owner_noncompAlex Franklin=48 of 61 (78.7%); Cole Ingram=19 of 22 (86.4%); Dana Mercer=18 of 22 (81.8%); Bryce Harmon=15 of 18 (83.3%); Farid Osman=1 of 1 (100%)
graded checksnoncomp_pct · no_ds1_0 · regressed_0 · per_owner_breakdown · definitions_first · three_fixes

Model's actual answer

show the model's full answer
DEFINITIONS

- DS2+ deal: current stage DS2, DS3, DS4, or DS5. Scope: 124 of the 156 open deals (32 are DS1).
- Why-buys compliant (DS2+ only): why_buys populated, ≥15 chars, and not a placeholder (tbd / n/a / see notes). LONG(n) is treated as present and substantive. In this file every why_buys is either blank or LONG(n) — no short strings or placeholders occur — so non-compliant = blank/missing.
- No DS1 timestamp: t_ds1 is blank.
- Regressed: any earlier-stage entry timestamp strictly later than a later-stage one (equal same-day timestamps are not regressions).

NUMBERS

Why-buys non-compliance, DS2+ deals: 101 non-compliant / 124 = 81.5%. (Compliant = the 23 deals carrying LONG(n).)

By owner (name per owners.csv):
- Alex Franklin (84342457): 48/61 = 78.7%
- Dana Mercer (83155923): 18/22 = 81.8%
- Cole Ingram (83155924): 19/22 = 86.4%
- Bryce Harmon (119337721): 15/18 = 83.3%
- Farid Osman (716654662): 1/1 = 100.0%
- Elena Sinclair (701163055): no DS2+ deals (her only deal, Deal-57FF13, is DS1) — not applicable

Deals with no DS1 timestamp: 0 of 156 — every row has a t_ds1 value.

Regressed deals: 0 of 156 — every stage-entry series present is non-decreasing (same-day entries are equal, not inverted).

Missing-data caveats (explicitly excluded from the counts above):
- t_ds5 is blank on all 10 DS5 deals; t_ds4 is also blank on 3 of them (Deal-547B2B, Deal-403845, Deal-584EE5; 403845 also lacks t_ds3). That is missing stage history, not timestamp inversion, so it does not count as regressed under the stated definition.
- Two rows have trailing-field irregularities (Deal-3EED2C, Deal-57FF13); why_buys-blank and t_ds1-present are unambiguous for both, so the metrics are unaffected.

THREE PROCESS FIXES

1. Stage-gate why-buys at promotion: block any save into DS2+ unless why_buys is ≥15 chars with placeholder rejection, and run a one-time backfill on the 101 open gaps — starting with the 10 of 11 COMMIT deals missing it (only Deal-584EE5 has text).
2. System-write stage timestamps: auto-stamp t_ds1 at creation (preserves today's 0-blank record) and write t_ds2–t_ds5 automatically on each stage change so late-stage deals stop carrying blank entry dates (the all-DS5 t_ds5 gap above).
3. Weekly owner scorecard: publish why-buys compliance by owner (Cole Ingram 86.4% and Farid Osman 100% non-compliant vs Alex Franklin 78.7%) and route failing deals back to the owner before the weekly COMMIT/BEST_CASE forecast call.
owner-resolutionPASS 1.00
Plain-English question: Every open deal has a blank owner name — resolve each one to a real person using the owners table.
Prompt — what the model saw
The deal extract below has an owner name field that is blank on nearly every row and an owner id field that is populated. Using the owners table, resolve each open deal to an owner name. Report: number of deals resolved, deals whose owner id has no match in the owners table (list them), the owner ids that map to archived/deactivated owners (list their names), and the total pipeline amount per resolved owner. Do not guess a name for any unmatched id.
cost $0.2823 · 77s · in 30,339 / out 25,926 tok
✓ resolved_count✓ unmatched_none✓ deactivated_flagged✓ bryce_amount✓ alex_amount✓ no_invented

✓ Correct Answer (ground truth)

Plain-English answer: All 156 deals resolve to six named owners; jake_amount maps to Bryce Harmon ($1,054,144) and justin_amount to Alex Franklin ($624,310).
The answer in one line: jake_amount = 1,054,144; justin_amount = 624,310; resolved = 156
jake_amount1,054,144
justin_amount624,310
resolved156
deactivated
per_ownerBryce Harmon=1,054,144; Alex Franklin=624,310; Dana Mercer=341,195; Cole Ingram=288,161.43; Farid Osman=4,134; Elena Sinclair=2,100
graded checksresolved_count · unmatched_none · deactivated_flagged · bryce_amount · alex_amount · no_invented

Model's actual answer

show the model's full answer
Owner resolution for all 156 open deals in deals_open.csv (every row has a populated owner_id):

**Deals resolved: 156 of 156** — every owner_id matched a row in owners.csv.

**Deals with unmatched owner_id: none.** No owner_id in the deal file is absent from the owners table.

**Deals whose owner is archived/deactivated: none.** The two archived owners in owners.csv (Gavin Porter, id 1520255671; Hugo Lindqvist, id 77260721) have no open deals assigned to them. All open deals map to the six active owners.

**Total pipeline amount per owner** (sum of `amount` across that owner's open deals; count in parentheses):

| Owner (owner_id) | Deals | Pipeline total |
|---|---|---|
| Bryce Harmon (119337721) | 35 | 1,054,144 |
| Alex Franklin (84342457) | 67 | 624,310 |
| Dana Mercer (83155923) | 24 | 341,195 |
| Cole Ingram (83155924) | 22 | 288,161.43 |
| Farid Osman (716654662) | 7 | 4,134 |
| Elena Sinclair (701163055) | 1 | 2,100 |
| **Grand total** | **156** | **2,314,044.43** |

Arithmetic check: 1,054,144 + 624,310 + 341,195 + 288,161.43 + 4,134 + 2,100 = 2,314,044.43, which equals the sum of all 156 amount values.

Notes on the data as given: Cole Ingram's total includes four fractional amounts (7,225.40 + 7,781.20 + 6,947.50 + 2,480.40 + 4,779.88 + 3,334.80 + 58,529.25 among his deals), hence the .43. Bryce Harmon's total includes three deals with amount = 1 (Deal-012CB1, Deal-483B2D, Deal-3795AD), which are counted as-is. No owner name was inferred for any row; all names come directly from owners.csv.
gong-hubspot-join-integrityPASS 1.00
Plain-English question: How many call records point at deal ids that do not exist in the open pipeline?
Prompt — what the model saw
Check the integrity of the call-to-deal mapping below. The gong table lists calls in the last 90 days keyed to HubSpot deal ids; deals_open lists the current open default-pipeline deals. Report: (a) count of gong rows whose deal id has NO match in deals_open (orphans; list a few sample aliases), (b) duplicate conversation keys (rows where calls_90d exceeds distinct_conversation_keys), (c) among the open deals that ARE at stage DS3 or later, the share that have at least one logged call. Give sample aliases for the orphan finding, all taken from the data. Note: gong_owner_id is a Gong user id, not a HubSpot owner id.
cost $0.6229 · 153s · in 58,116 / out 57,955 tok
✓ orphan_count✓ dup_keys_0✓ ds3_coverage✓ sample_ids_present

✓ Correct Answer (ground truth)

Plain-English answer: 30 Gong call rows are orphans keyed to dead deal ids, none are duplicates, and DS3 call coverage is 29.4%.
The answer in one line: orphans = 30; ds3_cov_pct = 29.4; dup = 0
orphans30
ds3_cov_pct29.4
dup0
orphan_aliasesDeal-8FA85D, Deal-8FC3F9, Deal-3B7945, Deal-42B265, Deal-9CCC42, Deal-36EA09
graded checksorphan_count · dup_keys_0 · ds3_coverage · sample_ids_present

Model's actual answer

show the model's full answer
**Data integrity check — Gong calls (last 90d) vs. open default-pipeline deals**

Base counts: 67 gong rows (67 unique hs_deal_id), 156 deals_open rows (156 unique deal_id). Aliases are consistent across both files for every matched id.

**(a) Orphans — gong rows whose deal id has no match in deals_open: 30 of 67 rows (44.8%).**
These 30 rows carry 304 of the 940 total logged calls. Sample aliases (all from the data): Deal-8FA85D, Deal-8FC3F9, Deal-42B265, Deal-9CCC42, Deal-3F86A0, Deal-B038F0. Full list available if needed. The heaviest offenders (Deal-8FA85D, 46 calls; Deal-42B265 and Deal-9CCC42, 17–21 calls each) have last_call dates extending to 2026-11-25/27 — dates in the future relative to a 90-day window ending mid-September — so these look like a separate import/source with inconsistent period and matching, not just closed/lost deals: this is a real mapping break, not expected churn. Caveat: deals_open contains only open deals, so some of the 30 may be legitimately closed/lost; the data provided doesn't say which.

**(b) Duplicate conversation keys: 0 rows.** In every one of the 67 rows calls_90d equals distinct_conversation_keys (no row has calls_90d > distinct), so there are no duplicated conversation keys — call-level dedupe looks clean.

**(c) Logged-call coverage on open deals at DS3+: 29.4%.**
- Open deals at DS3 or later: 85 (DS3 = 52, DS4 = 21, DS5 = 12)
- With at least one logged call: 25
- Share = 25 / 85 = 0.294 → **29.4%**; the other 60 DS3+ deals have zero logged calls.

Arithmetic: 25/85 = 0.29412 → 29.4%. Caveat: this compares "calls logged in the last 90 days" to deals that may have been worked earlier; gong_owner_id was correctly ignored (Gong ids, not HubSpot owner ids — not joinable to deals_open.owner_id).

⚠️ File-mutation verifier: 2 file edit(s) FAILED this turn despite any wording above that may suggest otherwise. Run `git status` or `read_file` to confirm what actually landed.
  • `/Users/amaniphipps/.hermes/cache/scratch/gong_calls_by_deal_90d.csv` — [write_file] Refusing to overwrite `/Users/amaniphipps/.hermes/cache/scratch/gong_calls_by_deal_90d.csv`: `/Users/amaniphipps/.hermes/cache/scratch/gong_calls_by_deal_90d.csv` exists but this task…
  • `/Users/amaniphipps/.hermes/cache/scratch/deals_open.csv` — [write_file] Refusing to overwrite `/Users/amaniphipps/.hermes/cache/scratch/deals_open.csv`: `/Users/amaniphipps/.hermes/cache/scratch/deals_open.csv` exists but this task has not seen its full c…
snowflake-sql-trapped-catalogPASS 1.00
Plain-English question: Write one Snowflake query for first-month usage per customer using only the provided, trap-laden catalog.
Prompt — what the model saw
Using ONLY the catalog provided, write one Snowflake SQL query returning, per customer company, for its first calendar month as a customer: unique givers, recognition count, successful redemption count. Read the catalog notes carefully: some tables are stale or unpopulated and must be avoided, and one documented filter (excluding deleted givers) must NOT be applied to historical giving counts. Put a comment above each FROM or JOIN explaining the table choice. Return only fenced SQL.
cost $0.0694 · 40s · in 1,718 / out 8,085 tok
✓ avoids_stale_deals✓ avoids_unpopulated_live✓ uses_cohort_summary✓ succeeded_filter✓ no_giver_deletion_filter✓ returns_sql

✓ Correct Answer (ground truth)

Plain-English answer: A single correct SQL query that uses the right catalog tables and avoids the trap columns — graded on the exact tables and columns.
The answer in one line: use_table = PRODUCTION.PLG.COMPANY_COHORT_SUMMARY
avoid_stalePRODUCTION.HUBSPOT.DEALS (stale)
avoid_unpopulatedV2_LIVE.OBJECTS_DEALS (unpopulated)
forbidden_filterdeleted-giver exclusion must NOT be applied
outputone fenced SELECT with per-FROM/JOIN table-choice comment
required_filtersucceeded redemptions only
use_tablePRODUCTION.PLG.COMPANY_COHORT_SUMMARY
graded checksavoids_stale_deals · avoids_unpopulated_live · uses_cohort_summary · succeeded_filter · no_giver_deletion_filter · returns_sql

Model's actual answer

show the model's full answer
```sql
-- ============================================================================
-- NOT PRODUCIBLE AS SPECIFIED. Missing data, stated explicitly:
--   1. No event-level recognition (giving) table exists anywhere in this
--      catalog. Unique givers and recognition count require giver IDs and
--      recognition timestamps; no table here has either. The only giving
--      figures in the catalog are pre-aggregated M1_GIVES / M2_GIVES /
--      M3_GIVES in COMPANY_COHORT_SUMMARY, and the catalog does not define
--      the M1 anchor date or calendar alignment, so they cannot be mapped
--      to "first calendar month as a customer".
--   2. PRODUCTION.DEPRECATED_RECOGNITION.REDEMPTION_RECORDS_V2 documents
--      only the STATE = 'succeeded' rule. No company key, event timestamp,
--      or giver column is documented, so successful redemptions cannot be
--      attributed to a company or a calendar month. To discover its actual
--      columns before any use:
--        -- DESCRIBE TABLE PRODUCTION.DEPRECATED_RECOGNITION.REDEMPTION_RECORDS_V2;
--   3. "Per customer company": COMPANY_COHORT_SUMMARY documents no company
--      identifier column, and no documented join key links it to HubSpot
--      companies or Chargebee CUSTOMER_ID. No joins are used because no
--      documented join path exists.
--   4. "Customer" start is undefined: SIGNUP_DATE vs ACTIVATED_DATE vs
--      FIRST_SUB_PAYMENT_DATE. CURRENT_STATUS values are not documented,
--      so no customer-only filter is applied.
--   Catalog rules honored: the deleted-giver exclusion is NOT applied
--   (it understates history; moot here only because no giving table
--   exists). Stale/unpopulated tables (HUBSPOT_HUB_1973303.V2_LIVE.OBJECTS_DEALS,
--   PRODUCTION.HUBSPOT.DEALS, PRODUCTION.HUBSPOT.GONG_HUBSPOT_MAP_FAST)
--   are avoided per catalog; none hold recognition data in any case.
-- ============================================================================
-- NEAREST AVAILABLE PROXY (self-serve cohort month 1; NOT the requested
-- calendar-month window). Arithmetic is a 1:1 passthrough of documented
-- aggregates: unique givers <- M1_USERS (documented as users, NOT givers);
-- recognition count <- M1_GIVES; successful redemptions <- M1_REDEMPTIONS
-- (the STATE = 'succeeded' rule cannot be verified against this aggregate).
SELECT
    DATE_TRUNC('MONTH', FIRST_SUB_PAYMENT_DATE)
        AS first_calendar_month__IF_customer_start_is_first_payment,
    SIGNUP_DATE,
    ACTIVATED_DATE,
    FIRST_SUB_PAYMENT_DATE,
    CURRENT_STATUS,
    CURRENT_ARR_USD,
    M1_USERS        AS unique_givers__PROXY_M1_USERS_NOT_GIVERS,
    M1_GIVES        AS recognition_count__PROXY_COHORT_M1_ONLY,
    M1_REDEMPTIONS  AS successful_redemptions__PROXY_COHORT_M1_ONLY
-- PRODUCTION.PLG.COMPANY_COHORT_SUMMARY: the only table in the catalog
-- carrying any giving or redemption figures (pre-aggregated, one row per
-- self-serve company). Every other catalog table is HubSpot CRM, Gong, or
-- billing metadata: no recognition events, no giver identities, and no
-- documented key to join here. Redemptions cannot be rebuilt from
-- REDEMPTION_RECORDS_V2 because none of its columns are documented.
FROM PRODUCTION.PLG.COMPANY_COHORT_SUMMARY;
```
crm-completeness-enrichmentPASS 1.00
Plain-English question: Audit the CRM for missing and invalid fields, and propose the fixes.
Prompt — what the model saw
Audit the CRM extract for completeness and propose fixes. Deals need owner, stage, amount, close date, why-buys; companies need industry, employee count, HQ country; contacts need email, title, persona. Report completeness percent per field. Identify duplicate company clusters (name variants, shared domain) with a survivor per cluster. List invalid emails and domain mismatches. Fill missing company fields only where the enrichment export has a matching row; where CRM and enrichment disagree list both and recommend a source. Never invent a value. End with the 10 fixes with the most pipeline amount at stake.
cost $0.3059 · 109s · in 30,969 / out 28,929 tok
✓ dup_clusters✓ invalid_emails✓ survivor_language✓ completeness_pct✓ no_invent

✓ Correct Answer (ground truth)

Plain-English answer: 9 companies missing employee count, 6 missing HQ country, 4 invalid emails, 1 domain mismatch, 2 duplicate clusters, and 25 enrichment matches.
The answer in one line: n_contacts = 52; n_companies = 34; enrichment_matches = 25; invalid_email_count = 4; n_dup_clusters = 2; domain_mismatch_count = 1 (+4 more below)
n_contacts52
domain_mismatch_count1
enrichment_matches25
invalid_email_count4
n_companies34
n_dup_clusters2
company_missingindustry=0; employee_count=9; hq_country=6
domain_mismatchesCT-0011
duplicate_clusters['C-0A092931', 'C-0A092932'], ['C-0A092933', 'C-0A092934']
invalid_emailsCT-0010, CT-0080, CT-0081, CT-0192
graded checksdup_clusters · invalid_emails · survivor_language · completeness_pct · no_invent

Model's actual answer

show the model's full answer
# CRM Extract Audit

## 0. Scope gap — read first

The extract contains only `companies.csv`, `contacts.csv`, `zoominfo_enrichment.csv`. **No deals file was provided**, so owner / stage / amount / close date / why-buys completeness cannot be computed (0 of 5 deal fields auditable), and no dollar pipeline figure exists to rank fixes by. The Top-10 list at the end is therefore ranked by records affected, not dollars — every "amount at stake" figure is unknown until the deals file is supplied.

## 1. Completeness by field

Companies (n=34):

| Field | Complete | % | Arithmetic | Missing (aliases) |
|---|---|---|---|---|
| company_alias | 34/34 | 100.0% | — | none |
| domain | 34/34 | 100.0% | — | none |
| industry | 34/34 | 100.0% | — | none (but see §4: only 24/34 = 70.6% use canonical values) |
| employee_count | 25/34 | 73.5% | 25÷34 | C-EC3025, C-96039F, C-44EA29, C-D04904, C-B23205, C-60C75F, C-7BBDFA, C-50D386, C-93C8BF |
| hq_country | 28/34 | 82.4% | 28÷34 | C-2D1F1B, C-D73B89, C-44EA29, C-D04904, C-2C60E5, C-EE9FFB |

Contacts (n=52):

| Field | Complete | % | Arithmetic | Notes |
|---|---|---|---|---|
| email (non-blank) | 52/52 | 100.0% | — | but only 48/52 = 92.3% syntactically valid; 47/52 = 90.4% valid AND on company domain |
| title | 39/52 | 75.0% | 39÷52 | 13 blanks |
| persona | 37/52 | 71.2% | 37÷52 | 15 blanks |
| domain | 52/52 | 100.0% | — | all match the parent company's domain |

Cross-file: all 52 contact `company_alias` values resolve to a company row; 14 of 34 companies (41.2%) have zero contacts (incl. all 4 duplicate-cluster members, C-2C60E5, C-2D7423, C-B97B4E, C-BA969B, C-332637, C-93C8BF, C-EE9FFB, C-C9BB20, C-7BBDFA, C-50D386).

## 2. Invalid emails and domain mismatches

Invalid (missing domain; local part present), 4 rows = 7.7% of contacts:

| contact_key | company_alias | stored value | deterministic repair (append CRM domain; verify before send) |
|---|---|---|---|
| CT-0010 | C-66D1FC | user0@ | user0@66d1fc.com |
| CT-0080 | C-92D97D | user0@ | user0@92d97d.com |
| CT-0081 | C-92D97D | user1@ | user1@92d97d.com |
| CT-0192 | C-425E2A | user2@ | user2@425e2a.com |

Domain mismatch, 1 row: **CT-0011 (C-66D1FC)** — `user1@other-domain.com` vs expected `66d1fc.com`. Could be legitimate (agency/consultant) or stale; verify with the account owner — do not auto-rewrite. Note C-92D97D has 2 of its 3 contacts broken.

## 3. Duplicate company clusters

Both found via shared domain; aliases differ (name variants not observable — no company-name column exists in the extract).

| Cluster (domain) | Members | Evidence | Survivor | Merge |
|---|---|---|---|---|
| acme-corp.com | C-0A092931 (Technology, 500, US) / C-0A092932 (tech, 510, USA) | shared domain; Δemployees = 510−500 = 10 (2.0%, same-company range); US/USA same country | **C-0A092931** (canonical industry; the 500 headcount) | fold C-0A092932 in |
| globex.io | C-0A092933 (SaaS, 200, US) / C-0A092934 (Technology, 200, US) | shared domain; identical headcount and country | **C-0A092933** (retains the more specific "SaaS" as sub-segment; normalize industry to Technology) | fold C-0A092934 in |

Neither cluster has any enrichment row or contacts, so the merge is low-risk but still splits account-level reporting today.

## 4. Enrichment join (fill only where a matching row exists)

Coverage: 25 of 34 CRM domains have an enrichment row; 9 have none (C-BA969B, C-332637, C-93C8BF, C-EE9FFB, C-C9BB20, both acme rows, both globex rows).

**Fills possible (CRM blank, enrichment has value) — employee_count, 8 companies:** C-EC3025←400, C-96039F←400, C-44EA29←400, C-D04904←400, C-B23205←400, C-60C75F←400, C-7BBDFA←400, C-50D386←400. Lifts employee_count completeness 25/34 → 33/34 (73.5% → 97.1%). Not fillable: C-93C8BF (no enrichment row).

**Not fillable from enrichment — hq_country, all 6 blanks:** for C-2D1F1B, C-D73B89, C-44EA29, C-D04904, C-2C60E5 the enrichment row exists but `zi_hq_country` is also blank (the gap is correlated across both systems); C-EE9FFB has no enrichment row. Requires re-enrichment or manual research.

**Disagreements (both sides populated):**

| Field | Count | Nature | Recommendation |
|---|---|---|---|
| employee_count | 0 conflicts; 17 exact agreements (17÷17) | — | Enrichment agrees everywhere comparable; use it as fill source |
| industry | 10 conflicts | All taxonomy, not fact: CRM Technology/tech/"Tech " vs enrichment "Computer Software" (e.g., C-77A95A, C-92D97D, C-B25F40, C-425E2A) | Keep CRM taxonomy; normalize CRM to canonical set {Manufacturing, Retail, Finance, Healthcare, Technology} — 10 rows need cleanup (tech ×3, "Tech " ×4, "health care" ×2, SaaS ×1); treat Computer Software ≡ Technology |
| hq_country | 10 conflicts | All formatting: US/USA vs "United States" (e.g., C-66D1FC US, C-E51FB7 USA); zero substantive disagreements | Standardize to one vocabulary (enrichment's "United States" or ISO codes); 15 of 28 populated CRM rows use US/USA shorthand, 2 use "United States", enrichment uses "United States" |

## 5. Top 10 fixes, ranked by exposure

No deal amounts exist in the extract (see §0), so rank = records affected, descending; the dollar ranking must wait for the deals file.

1. **Supply the deals file.** Owner/stage/amount/close date/why-buys are 0% auditable; every $-weighted priority and forecast number is unverifiable until it arrives.
2. **Merge acme-corp.com cluster → C-0A092931** (§3). Two accounts, one company; splits any pipeline and firmographic reporting.
3. **Merge globex.io cluster → C-0A092933** (§3). Same defect class.
4. **Repair the 4 malformed emails** (CT-0010, CT-0080, CT-0081, CT-0192) by appending the CRM domain, then verify deliverability. C-92D97D loses 2 of 3 contacts to this one defect.
5. **Adjudicate CT-0011** (`user1@other-domain.com` on C-66D1FC): confirm off-domain or fix to the CRM domain.
6. **Backfill 8 employee_count values from enrichment** (§4). Highest-yield mechanical fill: 73.5% → 97.1%.
7. **Re-enrich or manually research hq_country for the 6 blank companies.** Enrichment cannot fill any of them — correlated blank in both systems; C-EE9FFB has no enrichment row at all.
8. **Re-pull enrichment for the 4 non-duplicate companies missing rows** (C-BA969B, C-332637, C-93C8BF, C-C9BB20); this is the only path to fixing C-93C8BF's headcount.
9. **Standardize industry taxonomy** (10 non-canonical CRM rows) and lock the Technology ≡ Computer Software mapping so enrichment joins stop flagging false conflicts.
10. **Standardize country vocabulary** (15 US/USA rows + 2 "United States") and **backfill contact title (13 blanks) / persona (15 blanks)** — routing and persona scoring run on guesses today. Also flag the 14 zero-contact companies (41.2% of accounts) for coverage review.
deal-intelligence 5 tests
closed-lost-classificationPASS 1.00
Plain-English question: Why did we lose the last six months of deals — one primary category per deal, and whose side is the reason on?
Prompt — what the model saw
Classify every deal lost in the last 6 months below into one primary category: pricing, competitor, no decision, timing, product gap, champion left, or other. Also assign a side (Bonusly, buyer, unknown). Then summarize: category counts, the side split, the number of deals where the structured closed_lost_tag clearly disagrees with the free-text reason, and the two patterns most worth acting on. Base everything only on the tag and free-text provided.
cost $0.9278 · 502s · in 25,588 / out 108,213 tok
✓ total_90✓ competitor_top✓ category_counts_present✓ side_split✓ two_patterns

✓ Correct Answer (ground truth)

Plain-English answer: Every closed-lost deal gets exactly one category (pricing, competitor, no decision, timing, product gap, champion left, other) plus a side, per the computed classification.
The answer in one line: total_lost_6mo = 90; top_reason = competitor (26)
total_lost_6mo90
categoriespricing, competitor, no decision, timing, product gap, champion left, other
patternstwo most actionable, tag-vs-text disagreements counted
sideBonusly / buyer / unknown per deal
top_reasoncompetitor (26)
graded checkstotal_90 · competitor_top · category_counts_present · side_split · two_patterns

Model's actual answer

show the model's full answer
All 90 deals classified. Method: primary category from the free-text reason where it's specific; the structured tag governs only where the text is silent or consistent with it. "Side" = Bonusly (something on our side: product, price/value, process) vs buyer (their internal budget/priorities/process) vs unknown (not determinable from text). Arithmetic checked in code: category counts sum to 90, side counts sum to 90, amounts sum to $1,267,945.16.

## Per-deal classification (grouped by category; side in brackets)

**No decision — 35 deals**
- [buyer] Tag "Doing nothing/Not a priority": Deal-13E9CF, Deal-E74A73, Deal-50E5D8, Deal-FEDBCB, Deal-7FBAC6, Deal-413C56, Deal-ABD14C, Deal-2FEDDB. Tag "Lost DM": Deal-F325A5 (layoffs/leadership change deprioritized), Deal-ED9AE7 (timing+budget+authority never converged), Deal-FAC17C (no final exec approval). Tag "Feature Request": Deal-8E27DA (didn't want R&R at all; went with a swag provider).
- [bonusly] Deal-E0441F (tag MIA; went stale after inheriting from a departed rep — our hygiene).
- [unknown] Tag MIA: Deal-AC944F, Deal-214060, Deal-21B045, Deal-988493, Deal-F308CA, Deal-4664E1, Deal-D48E0B, Deal-583ADB, Deal-7CB44D, Deal-AFA56C, Deal-D1AABF, Deal-2BBA21, Deal-386F6E, Deal-3F86A0, Deal-096750, Deal-79E61A, Deal-AE7C4E, Deal-DAB4F1, Deal-B4B50F, Deal-5885B9. Plus Deal-55867E (tag Timing; text is a soft evaluation close, no timeline given), Deal-70F704 (narrow use case, went dark).

**Competitor — 26 deals**
- [buyer] 9: Deal-A2C349, Deal-47F1A1, Deal-BF2A98, Deal-369281, Deal-64B19A, Deal-D0C698 (incumbents: Awardco, WorkTango, HiThrive, Paylocity, Motivosity, Kudos), Deal-8A0992, Deal-9FCD0D (Canadian-provider preference), Deal-1BCA50 (stakeholder already committed elsewhere).
- [bonusly] 6 (lost on a named capability gap): Deal-F97C37, Deal-422BA6 (ADP TotalSource preferred-partner integrations), Deal-DDAB52 (currency exchange handling), Deal-DC77FE (points-as-dollars customization), Deal-242273 (digitized internal points currency spendable on-site), Deal-5AD03E (budget access).
- [unknown] 11: Deal-F7F635, Deal-381C8C, Deal-F1E8A6, Deal-ACE061, Deal-2D2F8D, Deal-0F96AA, Deal-7CC678, Deal-C7156E, Deal-EECC02, Deal-1E7DA9, Deal-286F9C.

**Timing — 19 deals, all [buyer]**: Deal-DB0AAC, Deal-91A056, Deal-29326C, Deal-831B7B, Deal-39E25C, Deal-B3ABED, Deal-B6AC09, Deal-E6E80A, Deal-B038F0, Deal-BB78F3, Deal-15DA99, Deal-F4AF5D, Deal-79B7A1, Deal-9F176A, Deal-69CF3D, Deal-ECBF89, Deal-D1A623, Deal-175756, Deal-5E64CE (tag "Doing nothing/Cost"; text is explicit timing: locked into Nectar contract through Oct 2027).

**Pricing — 5 deals, all [buyer]**: Deal-7ED004, Deal-C33D91, Deal-8A119B, Deal-DAFB82, Deal-7B2236 (tag "Doing nothing/Cost"; text cites budget + wants simpler/cheaper).

**Product gap — 3 deals, all [bonusly]**: Deal-9048EB (tag MIA; text cites bad fit + multiple feature gaps), Deal-981AD4 (UI fit, not UK-focused), Deal-3618CC (tag "Lost DM"; text: wanted Surveys).

**Other — 2 deals**: Deal-5DB9B0 [unknown] (tag "not ICP"; text: "Spam."), Deal-2A292B [buyer] (building internally).

**Champion left — 0 deals.** No free-text shows a buyer champion departing. Closest calls, both rejected: Deal-F325A5 (leadership changed → priority shift, not sponsor loss) and Deal-E0441F (a Bonusly rep left, not theirs).

## Summary

Category counts (deal count, closed-lost amount):
- No decision: 35, $322,204
- Competitor: 26, $406,235
- Timing: 19, $256,011
- Pricing: 5, $172,450
- Product gap: 3, $94,245
- Other: 2, $16,800
- Champion left: 0
- Check: 35+26+19+5+3+2+0 = 90; $322,204+$406,235+$256,011+$172,450+$94,245+$16,800 = $1,267,945 (file total $1,267,945.16)

Side split: buyer 46 ($608,394), unknown 34 ($459,581), Bonusly 10 ($199,970). Bonusly-attributable losses are concentrated in competitor-capability ($103,320) and product-gap ($94,245) deals; the majority of dollars died of buyer-side inaction or are unattributed.

Tag vs free-text disagreements: 5 deals.
- Deal-8E27DA — tag "Feature Request"; text says they never wanted R&R (went with a swag provider)
- Deal-55867E — tag "Timing"; text contains no timing, just a soft close
- Deal-5E64CE — tag "Doing nothing/Cost"; text is a hard contractual timing lock (Nectar to Oct 2027)
- Deal-9048EB — tag "MIA"; text cites feature gaps as the real cause
- Deal-3618CC — tag "Lost DM"; text cites a missing Surveys capability

Two patterns most worth acting on:
1. Engagement collapse (no decision/MIA): 35 deals, $322,204 — the largest count bucket. 22 went formally MIA; 11 of the 35 are ≥$10K and hold $228,317 of the bucket (33,750+31,860+30,321+23,400+21,060+21,000+14,931+14,400+13,895+12,000+11,700). MIA deals average 3.4 contacts vs 3.3 file-wide, so more touching didn't prevent the fade — these died post-intro/demo without a forcing event. Action: revive sequence for the dark deals plus earlier qualification/exit criteria so they exit as no-decision before consuming cycle time.
2. Dated timing deferrals: 19 deals, $256,011 — 12 of 19 name an explicit re-entry window (2027, 2028, Q2, "new year", "end of the year"), e.g. Deal-B3ABED ($40,001, Q2 for 2028 budget), Deal-9F176A ($54,600, year-end), Deal-5E64CE (Nectar renewal Oct 2027). This is near-zero-cost recoverable pipeline: dated, calendar-triggered follow-ups rather than generic nurture.

Runner-up worth logging: of 26 competitor losses, only 6 name a concrete cause ($103,320), but 3 of those share one theme — points currency flexibility (exchange rates Deal-DDAB52, points-as-dollars Deal-DC77FE, on-site spending of digitized currency Deal-242273).

Data limitations: every row has both tag and free text, so nothing was unclassifiable, but 11 competitor deals and all 20 MIA deals give no reason at all; "MIA" and "Lost DM" tags carry no structured meaning, so those sides lean on inference and are marked unknown where the text is silent.
pipeline-tieringPASS 1.00
Plain-English question: Score all 156 open deals into exactly one of LOCK / ACTION / BUILD / REVIVE / WATCH / RISKY.
Prompt — what the model saw
Score the open pipeline below on these signals: forecast category, stage, meetings_30d, contact count, and engagement recency. Assign exactly one tier per deal: LOCK, ACTION, BUILD, REVIVE, WATCH, or RISKY (RISKY = the forecast category disagrees with the engagement evidence). Never tier a deal LOCK if it has zero meetings_30d. Do NOT list every deal. Return ONLY a compact JSON object with exactly these keys: - "tier_counts": object mapping each tier to the number of deals in it (must sum to the total deal count), - "tier_examples": object mapping each tier to up to 3 example deal aliases from the data, - "risky_deals": array of the deal aliases you tiered RISKY, - "lock_violations": number of deals tiered LOCK that have zero meetings_30d (should be 0), - "pipeline_shape": one short paragraph on the overall pipeline shape. Note: inbound_emails_30d is 0 for every row due to a data defect, so treat meetings_30d as the inbound signal.
cost $0.5333 · 204s · in 19,890 / out 63,901 tok
✓ json_returned✓ tier_vocabulary✓ tier_counts_sum✓ risky_deals_listed✓ no_lock_violation_claimed

✓ Correct Answer (ground truth)

Plain-English answer: All 156 open deals tiered exactly once across the six tiers using forecast category, stage, meetings, contacts, and recency.
The answer in one line: total = 156
total156
graded checksjson_returned · tier_vocabulary · tier_counts_sum · risky_deals_listed · no_lock_violation_claimed

Model's actual answer

show the model's full answer
{
  "tier_counts": {
    "LOCK": 3,
    "ACTION": 19,
    "BUILD": 44,
    "REVIVE": 1,
    "WATCH": 84,
    "RISKY": 5
  },
  "tier_examples": {
    "LOCK": ["Deal-D348E1", "Deal-C26D20", "Deal-403845"],
    "ACTION": ["Deal-25F752", "Deal-584EE5", "Deal-499BF6"],
    "BUILD": ["Deal-93C8BF", "Deal-C6FE92", "Deal-03BA69"],
    "REVIVE": ["Deal-2D1F1B"],
    "WATCH": ["Deal-66D1FC", "Deal-EC3025", "Deal-92D97D"],
    "RISKY": ["Deal-547B2B", "Deal-A2B47C", "Deal-2465CE"]
  },
  "risky_deals": ["Deal-547B2B", "Deal-B7EBD1", "Deal-A2B47C", "Deal-2465CE", "Deal-7BBDFA"],
  "lock_violations": 0,
  "pipeline_shape": "As of the ~Sep 4 snapshot implied by the data, this is a bottom-heavy 156-deal pipeline: 132 deals (85%) sit in DS1-DS3 and only 24 in DS4-DS5, and just 53 deals (34%) show any meeting in the last 30 days, so most mid-funnel motion is email-only. Late stage is thin and mixed in quality: all 3 LOCKs are DS5 COMMITs with meetings_30d >= 1 and contact within days of the snapshot, while 6 of the 11 COMMIT deals have zero meetings_30d, 4 of them closing within about a week (tiered RISKY: Deal-547B2B, Deal-B7EBD1, Deal-A2B47C, Deal-2465CE), i.e., forecast running ahead of buyer proof. The 84-deal WATCH block is freshly touched (only Deal-2D1F1B and Deal-7BBDFA are silent 30+ days) but stalled on meetings, and two brand-new deals (Deal-3EED2C, Deal-57FF13) have no engagement rows at all. Hygiene flags: Deal-333EBB's close date (2026-08-28) is already past due, and inbound_emails_30d is 0 for every row (stated defect), so meetings_30d carried the inbound signal."
}
call-transcript-extractionPASS 1.00
Plain-English question: Extract the CRM write-back fields from each call transcript as JSON.
Prompt — what the model saw
For each transcript, extract CRM write-back fields as JSON: why-buys (prospect statements only), pain points, stakeholders from the speaker list, budget signal (prospect-stated or null), timeline signal, competitor mentioned (only if the prospect raised it), next step (only if explicitly agreed), objections, confidence. Never fill a field from what the rep said or hoped.
cost $0.0926 · 54s · in 2,634 / out 10,876 tok
✓ json_emitted✓ budget_signal✓ null_discipline✓ stakeholders✓ competitor_prospect_only

✓ Correct Answer (ground truth)

Plain-English answer: Per transcript: prospect-only why-buys, pain points, stakeholders, budget and timeline signals, and competitors — e.g. TX-001: '$40k earmarked', live before open enrollment.
The answer in one line: rep = Alex Franklin
allowed_newTX-001, TX-002, TX-003, TX-004, TX-005, TX-006, Achievers, Bucketlist, Nectar
per_transcript{"TX-001": {"why_buys": ["automating anniversary and birthday awards"], "pain_points": ["HR team of three cannot keep up with it manually", "track everything in a spreadsheet, and people slip through
repAlex Franklin
rep_sourced_nulls{"TX-003": ["budget_signal \u2014 only the rep mentioned pricing/budget"], "TX-004": ["next_step \u2014 rep proposed, prospect said 'no promises'"], "TX-006": ["budget_signal \u2014 rep decoy, prospec
graded checksjson_emitted · budget_signal · null_discipline · stakeholders · competitor_prospect_only

Model's actual answer

show the model's full answer
Rubric used for confidence (prospect-stated signals only; rep lines never score): budget explicit +2 / conditional or non-figure +1 / none 0; timeline date- or event-bound +2 / quarter-level +1 / none 0; next step agreed+dated +2 / agreed undated +1 / none 0; −1 per distinct unresolved objection; competitor raised by prospect: active eval −1, past & rejected +1, past affinity −1; economic buyer on call +1. Bands: ≥6 High, 3–5 Medium, ≤2 Low.

```json
{
  "TX-001": {
    "deal_alias": "Deal-CFE7F4",
    "stakeholders": ["Prospect (VP People)", "Prospect (HR Admin)"],
    "mentioned_not_present": ["IT (sign-off requirement, per HR Admin)"],
    "why_buys": [
      "\"The big win for us would be automating anniversary and birthday awards\" (VP People)"
    ],
    "pain_points": [
      "HR team of three \"cannot keep up\" with manual anniversary/birthday awards (VP People)",
      "\"Right now we track everything in a spreadsheet, and people slip through the cracks\" (HR Admin)"
    ],
    "budget_signal": "\"We have about $40k earmarked for engagement tools this fiscal year\" (VP People)",
    "timeline_signal": "\"Ideally we would have this live before open enrollment in November\" (VP People)",
    "competitor_mentioned": "Achievers — prospect-raised, past evaluation, rejected: \"too heavy for a team our size\"",
    "next_step": "Security review with IT lead on September 12 (prospect: \"Yes — let's do the security review on September 12\")",
    "objections": [
      "\"we need SSO and audit logs for IT to sign off\" (HR Admin)"
    ],
    "confidence": {"level": "High", "score": 7, "math": "budget 2 + timeline 2 + next step 2 + competitor rejected +1 + budget holder on call +1 − objection 1 = 7"}
  },
  "TX-002": {
    "deal_alias": "Deal-70BB30",
    "stakeholders": ["Prospect (Head of Total Rewards)", "Prospect (CFO)"],
    "mentioned_not_present": ["Legal (agreement routing, per CFO)"],
    "why_buys": [
      "\"We want to tie recognition to retention for our hourly workforce\" (Head of Total Rewards)"
    ],
    "pain_points": [
      "\"regretted turnover there is over 30%\" in the hourly workforce (Head of Total Rewards)"
    ],
    "budget_signal": "\"Finance has approved a $25k pilot budget for this quarter\" (CFO)",
    "timeline_signal": "\"We want a decision by end of September\" (CFO); pilot budget scoped to this quarter",
    "competitor_mentioned": null,
    "next_step": "Rep sends pilot agreement; prospect routes it to legal this week (CFO: \"Yes — send the pilot agreement and we'll route it to legal this week\")",
    "objections": [
      "\"Integration with Workday has to be rock solid — that's my one condition\" (CFO)"
    ],
    "confidence": {"level": "High", "score": 6, "math": "budget 2 (approved) + timeline 2 + next step 2 + economic buyer (CFO) 1 − objection 1 = 6"}
  },
  "TX-003": {
    "deal_alias": "Deal-530B50",
    "stakeholders": ["Prospect (People Ops Manager)"],
    "mentioned_not_present": ["CEO (prospect-stated decision-maker: \"she decides anything people-related\")"],
    "why_buys": [
      "\"We need to make recognition visible across our 12 retail locations\" (People Ops Manager)"
    ],
    "pain_points": [
      "\"Store managers have zero budget autonomy for on-the-spot recognition today\" (People Ops Manager)",
      "Recognition not visible across 12 retail locations (People Ops Manager)"
    ],
    "budget_signal": null,
    "timeline_signal": "\"Honestly there's no rush on our side until Q1\" (People Ops Manager)",
    "competitor_mentioned": "Bucketlist — prospect-raised; CEO \"used Bucketlist at her last company and liked it\"",
    "next_step": "Call with the CEO to be scheduled; prospect will send two times (People Ops Manager: \"Yes, let's schedule a call with our CEO — I'll send two times\")",
    "objections": [
      "\"The CEO has to be sold first — she decides anything people-related\" (People Ops Manager)",
      "No urgency until Q1 (People Ops Manager)"
    ],
    "confidence": {"level": "Low", "score": 0, "math": "budget 0 + timeline 1 + next step 1 (agreed, undated) − CEO-gate objection 1 − Bucketlist affinity 1 + economic buyer 0 = 0"}
  },
  "TX-004": {
    "deal_alias": "Deal-180D02",
    "stakeholders": ["Prospect (VP People)", "Prospect (IT Security Lead)"],
    "mentioned_not_present": ["CFO (follow-up proposed, unconfirmed)"],
    "why_buys": [
      "\"We want to consolidate three separate recognition tools into one\" (VP People)"
    ],
    "pain_points": [
      "\"We're paying for three tools and none of them talk to our HRIS\" (VP People)"
    ],
    "budget_signal": "\"If it's under $15k annually, I can approve it without going to the board\" (VP People) — conditional threshold, no committed figure",
    "timeline_signal": "\"Our procurement cycle runs six to eight weeks minimum\" (IT Security Lead)",
    "competitor_mentioned": null,
    "next_step": null,
    "objections": [
      "\"The security review took three months for our last vendor — that's my hesitation\" (IT Security Lead)",
      "CFO follow-up not committed: \"Maybe — I need to check her calendar, no promises\" (VP People)"
    ],
    "confidence": {"level": "Low", "score": 2, "math": "budget 1 (conditional threshold) + timeline 1 (constraint, no target date) + next step 0 (not agreed) − objection 1 + economic buyer 1 = 2"}
  },
  "TX-005": {
    "deal_alias": "Deal-F8767A",
    "stakeholders": ["Prospect (HR Director)", "Prospect (People Ops Coordinator)"],
    "mentioned_not_present": ["Exec team (prospect-stated presentation audience)"],
    "why_buys": [
      "\"automate service milestones, and give us analytics on recognition equity across departments\" (HR Director)"
    ],
    "pain_points": [
      "\"Our night-shift teams feel invisible — their engagement scores run 20 points lower\" (People Ops Coordinator)",
      "\"Our exec team is skeptical after a failed rollout two years ago\" (HR Director)"
    ],
    "budget_signal": "\"We have $12k approved under our engagement line\" (HR Director)",
    "timeline_signal": "\"We need this running before our January all-hands\" (HR Director)",
    "competitor_mentioned": "Nectar — prospect-raised, active: \"We're mid-pilot with Nectar right now\"",
    "next_step": "Rep presents to the exec team on October 2 (HR Director: \"Yes — come present to our exec team on October 2\")",
    "objections": [
      "Exec skepticism after failed rollout two years ago (HR Director)",
      "\"you'd need to beat that experience\" — incumbent Nectar pilot (HR Director)"
    ],
    "confidence": {"level": "Medium", "score": 4, "math": "budget 2 (approved) + timeline 2 + next step 2 − objections 2 − active incumbent 1 + budget holder on call 1 = 4"}
  },
  "TX-006": {
    "deal_alias": "Deal-EE195F",
    "stakeholders": ["Prospect (HR Manager)"],
    "mentioned_not_present": ["COO (prospect-stated: forward target and presumed build-vs-buy decision influence)"],
    "why_buys": [
      "\"I'd love to cut the admin time on service awards\" (HR Manager)",
      "\"Budget isn't the issue — time is\" (HR Manager) — value driver is time savings, not price"
    ],
    "pain_points": [
      "\"I personally spend five hours a month ordering and shipping plaques\" (HR Manager)"
    ],
    "budget_signal": "\"Budget isn't the issue — time is\" (HR Manager) — qualitative only, no figure stated",
    "timeline_signal": "\"Our fiscal year resets in February, so a Q1 start is realistic\" (HR Manager)",
    "competitor_mentioned": null,
    "next_step": "Rep sends one-page overview; HR Manager forwards to COO this week (HR Manager: \"Yes — send the one-page overview and I'll forward it to our COO this week\")",
    "objections": [
      "\"our COO usually prefers building things in-house\" (HR Manager)",
      "Buy is weighed against an internal build: \"we're comparing this against just doing it internally\" (HR Manager)"
    ],
    "confidence": {"level": "Medium", "score": 3, "math": "budget 1 (non-figure statement) + timeline 1 + next step 2 − objection 1 + competitor 0 + economic buyer 0 = 3"}
  }
}
```

Notes on gaps: TX-003 budget is null because the prospect never stated one (the $8/employee figure was rep-supplied and excluded per instructions). TX-004 has no agreed next step — the CFO meeting was only "maybe." TX-006's budget signal is kept as qualitative prospect statement, not a number.
next-to-close0.75
Plain-English question: Which 3 deals are closest to signature right now?
Prompt — what the model saw
Which 3 deals are closest to signature right now? Use the CRM data and the Slack thread; Slack often has the real status before the CRM does. For each: deal, amount, why it is close, what is left. Under 250 words.
cost $0.0421 · 24s · in 1,969 / out 4,366 tok
✗ exactly_three✓ top3_named✓ slack_first✓ blocked_excluded

✓ Correct Answer (ground truth)

Plain-English answer: Deal-547B2B ($11,200), Deal-B7EBD1 ($9,000), and Deal-403845 ($9,000) — all DS5 COMMIT deals closing September 10-11, with what is left on each.
blocked_excludeddeal_alias=Deal-2465CE; reason=champion left + procurement freeze (Slack #sales-sparring 2026-09-04)
exactly_three1
ranking_ruleSlack signature-imminent deal ranks first; remaining ranked by stage (DS5>DS4), forecast (COMMIT>BEST_CASE>PIPELINE), close_date asc, amount desc
top3rank=1; deal_alias=Deal-547B2B; amount=11200; stage=DS5; forecast_category=COMMIT; close_date=2026-09-11 | rank=2; deal_alias=Deal-B7EBD1; amount=9000; stage=DS5; forecast_category=COMMIT; close_date=2026-09-10 | rank=3; deal_alias=Deal-403845; amount=9000; stage=DS5; forecast_category=COMMIT; close_date=2026-09-11
graded checksexactly_three · top3_named · slack_first · blocked_excluded

Model's actual answer

show the model's full answer
Closest to signature (thread dated 2026-09-04):

1. Deal-547B2B — $11,200 — DS5/COMMIT, close date 2026-09-11
Why: DS5/COMMIT; Slack says redlines returned clean, the signing page is out, and their VP People said they are signing 2026-09-05 ("signature-imminent").
What is left: customer signature only.

2. Deal-403845 — $9,000 — DS5/COMMIT, close date 2026-09-11
Why: DS5/COMMIT; Slack says the order form is with their finance team and the deal is "moving fine."
What is left: their finance team's approval of the order form, then signature.

3. Deal-A2B47C — $6,360 — DS5/COMMIT, close date 2026-09-11
Why: DS5/COMMIT; Slack says it is "still warm, just normal legal-review pace" as of 2026-09-04.
What is left: legal review, then signature.

Combined value: 11,200 + 9,000 + 6,360 = $26,560.

CRM/Slack divergences that shaped the ranking:
- Deal-2465CE ($5,400, DS5/COMMIT, close 2026-09-10) is excluded: Slack says the champion left, procurement froze new vendors, and it is being pulled from commit (realistically Q4). CRM is stale here.
- Deal-B7EBD1 ($9,000, DS5/COMMIT, close 2026-09-10) would rank by CRM alone but has no Slack corroboration, so its status is unverified; it is the next candidate.
- Deal-D348E1 ($13,770, DS5/COMMIT) is in the same "warm, legal review" state as Deal-A2B47C but closes 2026-10-15, a month further out.

Missing data: no post-2026-09-04 Slack updates were provided, so "right now" is anchored to that date.
product-gap-detectionPASS 1.00
Plain-English question: Which prospect complaints are true product gaps versus rollout or enablement issues?
Prompt — what the model saw
Surface true product gaps raised by prospects in active deals. For each candidate: quote the prospect line, classify as real gap, plan gate, or rollout/enablement issue using the product docs excerpt, and name the deal and amount. Only real gaps go in the summary. Rep language does not count as prospect voice.
cost $0.0216 · 12s · in 1,517 / out 1,777 tok
✓ real_gaps✓ plan_gate_classified✓ enablement_classified✓ gap_summary

✓ Correct Answer (ground truth)

Plain-English answer: Two real gaps (SCIM provisioning, ADP Workforce Now integration) separated from plan gates and enablement issues, each with the prospect quote and deal.
allowed_newTG-001, TG-002, TG-003, TG-004, TG-005, SCIM user provisioning, ADP Workforce Now integration, Custom report builder
enablement_issuestranscript_id=TG-004; deal_alias=Deal-84DBA6; amount=16000; feature=Slack integration recognition adoption; quote=We turned on the Slack integration in July, but half our managers still haven't given a single recognition — nobody ever trained them on it.; why_enablement=Slack integration exists on all plans; managers were never trained
excluded_rep_voicetranscript_id=TG-005; line=Full transparency — we don't have a native mobile app for hourly workers yet, though it's on the roadmap.; reason=rep voice, not prospect-raised
plan_gatestranscript_id=TG-003; deal_alias=Deal-CFE7F4; amount=32000; feature=Custom report builder; tier_required=Enterprise; quote=I tried to build a custom report combining recognition frequency with turnover and couldn't — is that possible on our current plan?
real_gapstranscript_id=TG-001; deal_alias=Deal-EC3025; amount=62000; feature=SCIM user provisioning; quote=We need SCIM user provisioning — our security team will not approve any tool that requires manual account management.; why_real=not listed in product docs at any tier | transcript_id=TG-002; deal_alias=Deal-D0D6B5; amount=23800; feature=ADP Workforce Now integration; quote=Our payroll runs on ADP Workforce Now and I don't see ADP anywhere in your integration list — that's a dealbreaker for us.; why_real=supported HRIS list is Workday/BambooHR/Gusto/Rippling only
graded checksreal_gaps · plan_gate_classified · enablement_classified · gap_summary

Model's actual answer

show the model's full answer
Deal amounts: not provided. Neither transcripts_gaps.csv nor product_docs.md contains any deal amount field, so amounts are reported as "not in source data" for every candidate.

Arithmetic: 5 transcripts reviewed → 2 real gaps, 1 plan gate, 1 rollout/enablement issue, 1 rep-voice line (excluded per "rep language does not count as prospect voice").

Candidate-by-candidate:

1. TG-001 / Deal-EC3025 — REAL GAP
   Prospect line (IT Security Lead): "We need SCIM user provisioning — our security team will not approve any tool that requires manual account management."
   Docs check: "SCIM user provisioning and ADP Workforce Now integrations are NOT currently listed as supported capabilities." The capability does not exist on any tier.
   Amount: not in source data.

2. TG-002 / Deal-D0D6B5 — REAL GAP
   Prospect line (HRIS Manager): "Our payroll runs on ADP Workforce Now and I don't see ADP anywhere in your integration list — that's a dealbreaker for us."
   Docs check: supported HRIS list is "Workday, BambooHR, Gusto, Rippling"; ADP Workforce Now is explicitly not listed. Missing integration, not a tier restriction.
   Amount: not in source data.

3. TG-003 / Deal-CFE7F4 — PLAN GATE (not a gap)
   Prospect line (VP People): "I tried to build a custom report combining recognition frequency with turnover and couldn't — is that possible on our current plan?"
   Docs check: "Custom report builder" is Enterprise-only (— on Core and Pro, yes on Enterprise). The capability exists; it is gated by tier. This is an upgrade-path conversation, not a product gap.
   Amount: not in source data.

4. TG-004 / Deal-84DBA6 — ROLLOUT/ENABLEMENT ISSUE (not a gap)
   Prospect line (People Ops Manager): "We turned on the Slack integration in July, but half our managers still haven't given a single recognition — nobody ever trained them on it."
   Docs check: Slack integration is supported on all plans. The product capability works; adoption failed due to lack of training. Rep response (enablement session) is consistent with this classification.
   Amount: not in source data.

5. TG-005 / Deal-36C33F — EXCLUDED (rep voice, not prospect voice)
   The mobile-app statement ("Full transparency — we don't have a native mobile app for hourly workers yet...") is spoken by Alex Franklin, the rep. The prospect's only line — "Good to know. The web version should be fine for our office staff for now." — raises no gap. Per instructions, rep language does not count as prospect voice, so this is not a prospect-raised gap.
   Amount: not in source data.

SUMMARY — true product gaps raised by prospects (only real gaps):

- SCIM user provisioning absent from product — Deal-EC3025 (amount not in source data). Quote: "We need SCIM user provisioning — our security team will not approve any tool that requires manual account management."
- ADP Workforce Now integration missing — Deal-D0D6B5 (amount not in source data). Quote: "Our payroll runs on ADP Workforce Now and I don't see ADP anywhere in your integration list — that's a dealbreaker for us."

Caveat: "active deals" is taken as given from the task framing; the files contain no deal-stage field to verify it.
rep-performance 5 tests
stale-pipeline-by-rep0.67
Plain-English question: Which open deals have had no email, call, or meeting in the last 7 days?
Prompt — what the model saw
The snapshot date is 2026-09-05. Using the open deals and the engagements-by-deal table, list every open deal with no logged email, call, or meeting in the last 7 days. Compute recency from the engagements table's last_email/last_call/last_meeting fields, NOT from any date field on the deal (the deal's last_contacted_field is unreliable). For each stale deal give: deal alias, owner name, stage, amount, days since last contact. Group by owner, order by amount descending within each owner. Then state, per owner, how many stale deals they carry and the total stale amount. (Do not write outreach text.)
cost $0.3714 · 135s · in 19,844 / out 42,336 tok
✓ stale_recall_ge_80✓ grouped_by_owner✓ bryce_top_amount✓ cole_stale_amt✗ dana_stale_amt✗ stale_count

✓ Correct Answer (ground truth)

Plain-English answer: 65 open deals are stale; the largest is Deal-2D1F1B at $240,000, with Bryce Harmon carrying the most stale dollars at $626,243.
The answer in one line: jake_amt = 626,243; jenn_amt = 261,645; kevin_amt = 252,905.03; top_amount = 240,000; stale_count = 65; top_alias = Deal-2D1F1B
jake_amt626,243
jenn_amt261,645
kevin_amt252,905.03
stale_count65
top_amount240,000
top_aliasDeal-2D1F1B
graded checksstale_recall_ge_80 · grouped_by_owner · bryce_top_amount · cole_stale_amt · dana_stale_amt · stale_count

Model's actual answer

show the model's full answer
**Method**
- Snapshot: 2026-09-05. Last contact = max(last_email, last_call, last_meeting) from engagements_by_deal_90d.csv. The deals table's last_contacted_field was ignored, as instructed.
- "Last 7 days" = dated 2026-08-29 through 2026-09-05 (≤7 days). Stale = latest engagement dated 2026-08-28 or earlier, or no logged email/call/meeting at all. (No deal's latest engagement falls on 08-29/08-30, so the boundary choice changes nothing.)
- Days since last contact = 2026-09-05 − last-contact date. Dates below are 2026, shown MM-DD.
- Future-dated last_meeting values (> 09-05, i.e., scheduled meetings) were NOT counted as contact in the window; recency uses the most recent engagement dated on/before the snapshot. Marked "(mtg MM-DD fut)".
- Deals with no row in the engagements table at all: Deal-3EED2C, Deal-57FF13 → no logged email/call/meeting; days = N/A.

---

**BRYCE HARMON (119337721) — 18 stale deals — $692,964.00**

```
Deal-2D1F1B  DS1  240,000.00  meeting 06-16 (last email 06-11)        09-05 − 06-16 = 81d
Deal-66D1FC  DS1   99,000.00  email 08-20                             09-05 − 08-20 = 16d
Deal-950043  DS1   70,000.00  email 08-17                             09-05 − 08-17 = 19d
Deal-B23205  DS1   45,000.00  email 08-20 (meeting 08-20)             09-05 − 08-20 = 16d
Deal-7BBDFA  DS3   37,440.00  email 07-21 (meeting 06-18)             09-05 − 07-21 = 46d
Deal-332637  DS2   36,000.00  email 08-27 (meeting 07-22)             09-05 − 08-27 = 9d
Deal-1BEEBF  DS1   31,500.00  email 08-17 (call 07-30)                09-05 − 08-17 = 19d
Deal-A414F6  DS1   25,200.00  email 08-17 (mtg 09-10 fut)             09-05 − 08-17 = 19d
Deal-C5658B  DS1   23,400.00  email 08-20 (meeting 07-31)             09-05 − 08-20 = 16d
Deal-40522D  DS3   21,000.00  email 08-17 (meeting 08-04)             09-05 − 08-17 = 19d
Deal-C1FA6D  DS1   18,000.00  email 08-20 (mtg 09-15 fut)             09-05 − 08-20 = 16d
Deal-01E193  DS1   12,600.00  email 08-28 (mtg 09-09 fut)             09-05 − 08-28 = 8d
Deal-F0EBBB  DS3   11,400.00  email 08-12 (meeting 07-02)             09-05 − 08-12 = 24d
Deal-927338  DS1   10,920.00  email 08-18 (mtg 09-17 fut)             09-05 − 08-18 = 18d
Deal-E25A09  DS1    6,000.00  email 08-27 (meeting 07-15)             09-05 − 08-27 = 9d
Deal-C9C286  DS2    5,502.00  email 08-27 (meeting 08-05)             09-05 − 08-27 = 9d
Deal-012CB1  DS1        1.00  email 08-13 (meeting 08-12)             09-05 − 08-13 = 23d
Deal-3795AD  DS2        1.00  email 08-28 (mtg 10-02 fut)             09-05 − 08-28 = 8d
```
Sum: 240,000+99,000+70,000+45,000+37,440+36,000+31,500+25,200+23,400+21,000+18,000+12,600+11,400+10,920+6,000+5,502+1+1 = 692,964

---

**DANA MERCER (83155923) — 16 stale deals — $279,295.00**

```
Deal-44EA29  DS2   60,000.00  email 08-26                             09-05 − 08-26 = 10d
Deal-E51FB7  DS2   43,875.00  call 08-24 (last email 08-18)           09-05 − 08-24 = 12d
Deal-B42F46  DS1   27,000.00  email 08-17                             09-05 − 08-17 = 19d
Deal-BA3DDC  DS3   23,400.00  call 08-21 (last email 08-20)           09-05 − 08-21 = 15d
Deal-9DDE86  DS2   20,000.00  email 08-21 (meeting 07-27)             09-05 − 08-21 = 15d
Deal-215CCA  DS3   18,900.00  meeting 08-19 (last email 07-02)        09-05 − 08-19 = 17d
Deal-5EED42  DS3   16,250.00  email 08-25 (call 08-25, meeting 07-24) 09-05 − 08-25 = 11d
Deal-57887A  DS2   15,000.00  email 08-28 (meeting 08-21)             09-05 − 08-28 = 8d
Deal-944310  DS4   10,500.00  no email/call logged; last_meeting 09-15 fut (see *)   N/A*
Deal-B7EBD1  DS5    9,000.00  email 08-20 (call 08-10, meeting 07-30) 09-05 − 08-20 = 16d
Deal-3974EB  DS4    9,000.00  email 08-28 (meeting 08-28)             09-05 − 08-28 = 8d
Deal-F40F04  DS2    8,100.00  email 08-21 (meeting 08-21)             09-05 − 08-21 = 15d
Deal-7599B8  DS3    7,350.00  email 08-18 (mtg 09-10 fut)             09-05 − 08-18 = 18d
Deal-87DDD1  DS1    5,000.00  email 08-17 (meeting 07-23)             09-05 − 08-17 = 19d
Deal-F336B6  DS3    4,200.00  email 08-21 (meeting 08-18)             09-05 − 08-21 = 15d
Deal-0660B4  DS4    1,920.00  meeting 08-20 (last email 08-10)        09-05 − 08-20 = 16d
```
Sum: 60,000+43,875+27,000+23,400+20,000+18,900+16,250+15,000+10,500+9,000+9,000+8,100+7,350+5,000+4,200+1,920 = 279,295

---

**COLE INGRAM (83155924) — 18 stale deals — $252,904.03**

```
Deal-D04904  DS2   58,529.25  email 08-25 (meeting 06-12)             09-05 − 08-25 = 11d
Deal-B25F40  DS3   40,000.00  email 08-28 (meeting 07-24)             09-05 − 08-28 = 8d
Deal-813836  DS2   32,175.00  email 08-25 (meeting 07-07)             09-05 − 08-25 = 11d
Deal-1BA595  DS2   31,750.00  email 08-25                             09-05 − 08-25 = 11d
Deal-CFE1E8  DS3   18,000.00  email 08-25 (meeting 07-24)             09-05 − 08-25 = 11d
Deal-CD47A6  DS2   12,168.00  email 08-25 (call 08-24)                09-05 − 08-25 = 11d
Deal-627646  DS3   11,193.00  email 08-25                             09-05 − 08-25 = 11d
Deal-FF809F  DS2    7,781.20  email 08-25 (meeting 06-18)             09-05 − 08-25 = 11d
Deal-AF932D  DS2    7,225.40  email 08-25 (meeting 07-21)             09-05 − 08-25 = 11d
Deal-A71728  DS2    6,947.50  email 08-25 (meeting 06-29)             09-05 − 08-25 = 11d
Deal-8BC9F5  DS2    5,616.00  email 08-26 (meeting 07-24)             09-05 − 08-26 = 10d
Deal-175395  DS3    4,779.88  email 08-25 (meeting 06-30)             09-05 − 08-25 = 11d
Deal-481E24  DS3    4,140.00  call 08-26 (last email 08-25)           09-05 − 08-26 = 10d
Deal-C7F9BF  DS2    3,360.00  email 08-25 (call 08-24)                09-05 − 08-25 = 11d
Deal-2F3A66  DS3    3,334.80  email 08-25 (meeting 06-17)             09-05 − 08-25 = 11d
Deal-342E96  DS2    2,700.00  email 08-12 (meeting 07-14)             09-05 − 08-12 = 24d
Deal-E568D5  DS3    1,875.00  email 08-25 (meeting 06-30)             09-05 − 08-25 = 11d
Deal-FD9F4E  DS5    1,330.00  email 08-26 (meeting 07-15)             09-05 − 08-26 = 10d
```
Sum: 58,529.25+40,000+32,175+31,750+18,000+12,168+11,193+7,781.20+7,225.40+6,947.50+5,616+4,779.88+4,140+3,360+3,334.80+2,700+1,875+1,330 = 252,904.03

---

**ALEX FRANKLIN (84342457) — 19 stale deals — $108,836.00**

```
Deal-CC08D1  DS1   24,000.00  email 08-20 (meeting 08-19)             09-05 − 08-20 = 16d
Deal-E73427  DS3   18,000.00  email 08-26 (meeting 08-26)             09-05 − 08-26 = 10d
Deal-885F45  DS2    9,300.00  email 08-24                             09-05 − 08-24 = 12d
Deal-C2FF3C  DS1    8,316.00  email 08-26 (meeting 08-14)             09-05 − 08-26 = 10d
Deal-3EED2C  DS2    7,200.00  no row in engagements table (see **)    N/A**
Deal-6C60D4  DS3    4,800.00  call 08-24 (last email 07-31)           09-05 − 08-24 = 12d
Deal-13FEBD  DS2    4,680.00  call 08-24 (last email 08-04)           09-05 − 08-24 = 12d
Deal-819506  DS1    4,400.00  email 08-28 (mtg 09-09 fut)             09-05 − 08-28 = 8d
Deal-9D0060  DS3    3,840.00  email 08-24 (meeting 07-24)             09-05 − 08-24 = 12d
Deal-690476  DS2    3,600.00  call 08-18 (last email 08-03)           09-05 − 08-18 = 18d
Deal-C6D97A  DS4    3,240.00  email 08-28 (call 08-25)                09-05 − 08-28 = 8d
Deal-EE195F  DS3    3,120.00  email 08-28 (meeting 08-06)             09-05 − 08-28 = 8d
Deal-278DEC  DS3    2,700.00  email 08-28                             09-05 − 08-28 = 8d
Deal-635B8E  DS3    2,600.00  email 08-18                             09-05 − 08-18 = 18d
Deal-6883F3  DS1    2,400.00  email 08-20 (meeting 08-20)             09-05 − 08-20 = 16d
Deal-4A13AD  DS3    2,160.00  email 08-10                             09-05 − 08-10 = 26d
Deal-F67D31  DS2    1,800.00  email 08-28                             09-05 − 08-28 = 8d
Deal-5FDCE4  DS3    1,600.00  email 08-24 (meeting 06-29)             09-05 − 08-24 = 12d
Deal-BA571A  DS4    1,080.00  email 08-18                             09-05 − 08-18 = 18d
```
Sum: 24,000+18,000+9,300+8,316+7,200+4,800+4,680+4,400+3,840+3,600+3,240+3,120+2,700+2,600+2,400+2,160+1,800+1,600+1,080 = 108,836

---

**FARID OSMAN (716654662) — 2 stale deals — $2,881.00**

```
Deal-8BA24E  DS1   2,880.00   email 08-28 (mtg 09-15 fut)             09-05 − 08-28 = 8d
Deal-8FDCD2  DS1       1.00   email 08-21 (mtg 09-15 fut)             09-05 − 08-21 = 15d
```
Sum: 2,880 + 1 = 2,881

---

**ELENA SINCLAIR (701163055) — 1 stale deal — $2,100.00**

```
Deal-57FF13  DS1   2,100.00   no row in engagements table (see **)    N/A**
```

---

**Per-owner recap** (owners ordered by total stale amount, descending)

```
Bryce Harmon    18 stale deals   $692,964.00
Dana Mercer     16 stale deals   $279,295.00
Cole Ingram     18 stale deals   $252,904.03
Alex Franklin   19 stale deals   $108,836.00
Farid Osman      2 stale deals   $2,881.00
Elena Sinclair   1 stale deal    $2,100.00
TOTAL           74 stale deals   692,964 + 279,295 + 252,904.03 + 108,836 + 2,881 + 2,100 = $1,338,980.03
```

**Data caveats (explicit)**
- \* Deal-944310: engagements row has blank last_email/last_call and last_meeting = 2026-09-15, after the snapshot (meetings_30d = 2, so meetings occurred within 30d but no past meeting date is logged). Nothing is dated in 08-29..09-05, so it is listed as stale, but days-since-last-contact cannot be computed.
- \*\* Deal-3EED2C and Deal-57FF13: no row at all in engagements_by_deal_90d.csv → no logged email/call/meeting; days = N/A. These 3 N/A deals total $19,800 of the grand total.
- Future-dated last_meeting values exist for 14 open deals; for the stale ones they are marked "fut" above. If scheduled upcoming meetings were instead counted as contact, 10 listed deals (incl. Bryce's A414F6, C1FA6D, 927338; Dana's 7599B8, 944310; Alex's 819506, 01E193; Farid's 8BA24E, 8FDCD2; Bryce's 3795AD) would drop out and the totals would change materially.
- All six owners above are active (archived=false) in owners.csv; the two archived owners hold no deals in the open file. Deal-57FF13's close_date (2026-09-02) already predates the snapshot yet it appears in the open-deals file; it is included as given.
activity-mix-vs-outcomePASS 1.00
Plain-English question: Per rep, what did the last 30 days of emails, calls, and meetings look like next to the deals that advanced?
Prompt — what the model saw
The snapshot date is 2026-09-05. For each sales rep (owner), compute the last-30-day totals of emails, calls, and meetings from the engagements table, and the number of deals that entered DS2 in the last 30 days (t_ds2 within 30 days of the snapshot). Report per rep: the activity mix as percentages (emails/calls/meetings share of that rep's total activities), total activities per DS2 entry, and rank the reps by that efficiency ratio (lowest activities per DS2 entry = most efficient). Name the most efficient rep and the highest-volume rep, and note if they differ. Do not attribute cause.
cost $0.7596 · 249s · in 24,805 / out 87,577 tok
✓ alex_ds2_30d✓ alex_volume_high✓ efficient_named✓ mix_present✓ no_causal_claim

✓ Correct Answer (ground truth)

Plain-English answer: Alex Franklin is the standout signal with 18 deals entering DS2 in 30 days (next closest: Bryce Harmon with 4), alongside per-rep activity totals.
The answer in one line: alex_ds2_30d = 18; bryce_ds2_30d = 4; note = legacy alias justin_ds2=4 actually held Bryce Harmon count; Alex Franklin entered 18 deals into DS2 in the 30d window
alex_ds2_30d18
bryce_ds2_30d4
notelegacy alias justin_ds2=4 actually held Bryce Harmon count; Alex Franklin entered 18 deals into DS2 in the 30d window
graded checksalex_ds2_30d · alex_volume_high · efficient_named · mix_present · no_causal_claim

Model's actual answer

show the model's full answer
## Per-rep 30-day activity and DS2-entry analysis (snapshot 2026-09-05; window 2026-08-06 → 2026-09-05)

Activities summed from `engagements_by_deal_90d.csv` joined to deals by `deal_id`, grouped by `owner_id` (owner names from `owners.csv`). DS2 entries = deals whose `t_ds2` falls within 30 days of the snapshot (inclusive: delta ≤ 30 days). Archived owners Gavin Porter and Hugo Lindqvist have no deals in the file, so no rows.

| Rep | Emails | Calls | Meetings | Total acts | Mix (E/C/M) | DS2 entries | Acts per DS2 entry | Rank |
|---|---|---|---|---|---|---|---|---|
| Alex Franklin (84342457) | 307 | 36 | 41 | 384 | 79.9% / 9.4% / 10.7% | 18 | 384/18 = 21.33 | 1 |
| Bryce Harmon (119337721) | 162 | 0 | 43 | 205 | 79.0% / 0.0% / 21.0% | 4 | 205/4 = 51.25 | 2 |
| Cole Ingram (83155924) | 96 | 14 | 1 | 111 | 86.5% / 12.6% / 0.9% | 2 | 111/2 = 55.50 | 3 |
| Farid Osman (716654662) | 38 | 0 | 34 | 72 | 52.8% / 0.0% / 47.2% | 1 | 72/1 = 72.00 | 4 |
| Dana Mercer (83155923) | 84 | 18 | 11 | 113 | 74.3% / 15.9% / 9.7% | 1 | 113/1 = 113.00 | 5 |
| Elena Sinclair (701163055) | 0 | 0 | 0 | 0 | n/a (no activity data) | 0 | undefined | n/r |

Mix arithmetic examples: Alex 307/384 = 79.9%, 36/384 = 9.4%, 41/384 = 10.7%. Bryce 162/205 = 79.0%, 43/205 = 21.0%, 0 calls. (Dana's mix sums to 99.9% due to rounding.)

DS2-entry deals by rep (alias, t_ds2, days before snapshot):
- Alex Franklin (18): Deal-403845 (09-02, 3d), Deal-1FC049 (09-03, 2d), Deal-3EED2C (09-03, 2d), Deal-7FA0C3 (08-07, 29d), Deal-E531A6 (08-07, 29d), Deal-5296C9 (08-28, 8d), Deal-36C33F (08-11, 25d), Deal-EE195F (08-06, 30d), Deal-F436DA (08-19, 17d), Deal-317E6F (08-12, 24d), Deal-D1E6C2 (08-11, 25d), Deal-D9A72E (08-06, 30d), Deal-CA5E44 (08-24, 12d), Deal-4F775F (08-17, 19d), Deal-898FC5 (08-28, 8d), Deal-46988D (08-26, 10d), Deal-E73427 (08-28, 8d), Deal-92D97D (09-02, 3d)
- Bryce Harmon (4): Deal-25F752 (08-10, 26d), Deal-D73B89 (09-03, 2d), Deal-CA7DC0 (08-12, 24d), Deal-1CCE5C (08-06, 30d)
- Cole Ingram (2): Deal-42326B (08-26, 10d), Deal-1BA595 (08-12, 24d)
- Farid Osman (1): Deal-499BF6 (08-26, 10d)
- Dana Mercer (1): Deal-57887A (08-07, 29d)

Most efficient rep: Alex Franklin (21.33 acts per DS2 entry). Highest-volume rep: Alex Franklin (384 total activities). They do not differ — the same rep holds both.

Data gaps and method notes:
- Deal-3EED2C (Alex Franklin) appears in the deals file with t_ds2 = 2026-09-03 but has no row in the engagements file; it counts as a DS2 entry and contributes 0 activities. Excluding it, Alex would be 384/17 = 22.59 — still rank 1.
- Deal-57FF13 (Elena Sinclair) has no engagement row and a blank t_ds2; her ratio is undefined and she is unranked.
- Boundary sensitivity: if "within 30 days" is read as strictly less than 30 days, the three t_ds2 = 2026-08-06 deals (Deal-1CCE5C, Deal-EE195F, Deal-D9A72E, all Alex's) drop out: Alex becomes 384/15 = 25.60, still rank 1, and the ranking order is unchanged.
- DS2 entries are counted regardless of the deal's current stage (some have since advanced, e.g. Deal-25F752 is now DS4). No cause is attributed.
ae-qtd-scorecardPASS 1.00
Plain-English question: Build Alex Franklin's QTD scorecard: bookings versus quota, pipeline, and conversion.
Prompt — what the model saw
Build a QTD scorecard for the named AE as of 2026-09-05: bookings vs quota with attainment percent, new vs expansion split, active pipeline by stage with amount, rolling 90-day DS2-to-won rate, win and loss counts with top loss reason, activity volume by type in the last 30 days. Exclude any closed-won deal dated before the quarter. Close with three coaching observations grounded in the numbers.
cost $0.2744 · 289s · in 16,404 / out 27,434 tok
✓ bookings✓ quota✓ attainment✓ new_expansion_split✓ top_loss_reason✓ coaching_obs

✓ Correct Answer (ground truth)

Plain-English answer: 75% attainment ($150,000 on a $200,000 quota), $113,500 new versus $36,500 expansion, pipeline by stage, and the rolling DS2-to-won rate.
The answer in one line: quota = 200000; bookings = 150000; new_bookings = 113500; expansion_bookings = 36500; open_deals_count = 125; attainment_pct = 75; owner = Alex Franklin (+12 more below)
quota200,000
attainment_pct75
bookings150,000
expansion_bookings36,500
expansion_count3
losses_qtd27
new_bookings113,500
new_count5
open_deals_count125
wins_qtd8
activity_30d_totalsemails_30d=807; calls_30d=112; meetings_30d=128; notes_30d=50
allowed_new18 items — e.g. Deal-A1C3E5, Deal-B7D2F4, Deal-C9E1A6, Deal-D4B8C2
ds2_to_wonwindow_start=2026-06-07; won_in_window=8; closed_in_window=35; rate=0.2286
ownerAlex Franklin
pipeline_by_stage{"DS1": {"count": 20, "amount": 284621}, "DS2": {"count": 28, "amount": 353760}, "DS3": {"count": 67, "amount": 552705}, "DS4": {"count": 5, "amount": 23574}, "DS5": {"count": 5, "amount": 45730}}
prior_quarter_excludeddeal_alias=Deal-B3E6F1; close_date=2026-06-20; amount=24000
quarter2026-Q3
snapshot_date2026-09-05
top_loss_reasonreason=Lost- Timing (1 year or more); count=13
graded checksbookings · quota · attainment · new_expansion_split · top_loss_reason · coaching_obs

Model's actual answer

show the model's full answer
# QTD Scorecard — Alex Franklin (as of 2026-09-05, 2026-Q3)

Note on method: I parsed only the CSVs provided in your message. The ae_deals.csv file in my working directory contained data for a different owner (Justin Durfee), so it was ignored.

## 1. Bookings vs Quota
- QTD closed-won bookings: $150,000 (8 deals, close dates 2026-07-15 through 2026-09-03)
- Quota (2026-Q3, snapshot 2026-09-05): $200,000
- Attainment: $150,000 / $200,000 = **75.0%**; gap to quota = $50,000
- Quarter elapsed: 66 of 91 days = 72.5% — pace is roughly linear-to-slightly-ahead

## 2. New vs Expansion Split (QTD)
- New: 5 deals, $113,500 (75.7% of bookings) — Deal-A1C3E5 $40,000, Deal-B7D2F4 $35,000, Deal-C9E1A6 $21,000, Deal-D4B8C2 $11,000, Deal-E6F3A9 $6,500
- Expansion: 3 deals, $36,500 (24.3%) — Deal-F2C7D8 $20,000, Deal-A8B4D6 $12,000, Deal-C5D9E2 $4,500
- Excluded per instruction: Deal-B3E6F1 ($24,000, closed-won 2026-06-20, before the quarter). It is still counted in the rolling 90-day rate below since its close date falls in that window.

## 3. Active Pipeline by Stage (status = open)
| Stage | Deals | Amount |
|---|---|---|
| DS1 | 20 | $284,621 |
| DS2 | 28 | $353,760 |
| DS3 | 67 | $552,705 |
| DS4 | 5 | $23,574 |
| DS5 | 5 | $45,730 |
| Total | 125 | $1,260,390 |

Coverage vs the $50,000 gap is 25.2x nominal, but DS4+DS5 (closest to close) holds only $69,304 — the vast majority sits in DS1–DS3 ($1,190,086, 115 deals).

## 4. Rolling 90-Day DS2-to-Won Rate (2026-06-07 to 2026-09-05)
- Wins with entered_ds2 + close date in window: 9 (8 QTD + pre-QTD Deal-B3E6F1)
- Losses in window: 27 (all closed 2026-07-29 to 2026-09-02)
- Rate: 9 / (9 + 27) = 9/36 = **25.0%**
- Median DS2-to-close: 29 days won vs 33 days lost

## 5. Win/Loss Counts and Top Loss Reason (90-day window)
- Won: 9, $329,272 was lost across 27 deals (avg lost deal $12,196 vs avg won $18,750)
- Top loss reason: **"Lost- Timing (1 year or more)" — 13 of 27 losses (48.1%), $184,681 (56.1% of lost value)**
- Rest: Competitor 5 / $49,020; MIA 5 / $45,831; Lost DM 2 / $17,940; Feature Request 1 / $21,000; Lost- Does not fit ICP (write in notes) 1 / $10,800

## 6. Activity Volume, Last 30 Days (per ae_engagements.csv, all 161 deals)
- Emails: 807 | Calls: 112 | Meetings: 128 | Notes: 50 — total 1,097 logged touches
- Caveat: the file's columns are defined as 30-day counts but span both closed and open deals, so this is book-wide activity, not a filtered open-pipeline figure. Open-deal-only touches total 744; 1 open deal has zero touches (Deal-3EED2C, DS2, $7,200).

## Coaching Observations

1. **The gap is winnable on paper but not from current late stage.** $69,304 sits in DS4/DS5 against a $50,000 gap, and even those deals (e.g., Deal-547B2B $11,200, entered DS2 148 days ago; Deal-F9A08A $2,484, entered 285 days ago) show extreme stage age. The quarter ends on current DS3 volume progressing — 67 deals/$552,705 there needs a deliberate push to DS4+, or the back half depends on new DS2 entries that won't have time to mature (median win takes ~29 days from DS2).

2. **Timing losses are a qualification problem, not a market problem.** 13 of 27 losses (48%) and 56% of lost dollars are "Lost- Timing (1 year or more)" — deals that should have been flagged at DS2 entry, not discovered at close. Pair that with 5 "MIA" losses ($45,831) and it points to entering late-stage with unvalidated urgency. A hard timeline-must-answer gate before DS3 would have redirected roughly half of the $329,272 of lost effort toward winnable deals.

3. **Activity is email-heavy and meeting-light where it matters most.** 807 emails vs 128 meetings and 112 calls is a low meeting-to-email ratio (~1:6), and 22 open deals ≥$15k had zero meetings in the last 30 days — including the two largest DS2 deals, Deal-EC3025 ($62,000) and Deal-92D97D ($60,000). Deals of that size at DS2 with no meetings are the exact profile that produced the MIA losses. Redirecting email volume into scheduled meetings on the 10 largest open deals is the highest-leverage activity change available.
multithreading-gapPASS 1.00
Plain-English question: Which open deals are single-threaded or under-threaded?
Prompt — what the model saw
Find every open deal that is single-threaded (fewer than 2 active contacts) or under-threaded (fewer than 3, or all contacts in one persona). Active means engaged in the last 60 days and not marked former. Personas: economic buyer, champion, HR admin, IT security, finance. For each flagged deal: amount, stage, active contact count, personas present, personas missing, the most valuable persona to add given the stage, and any on-file unengaged contact at that company who fits; otherwise say none on file.
cost $0.1173 · 54s · in 3,320 / out 13,953 tok
✓ flagged_recall✓ persona_language✓ single_thread_term✓ on_file_fit

✓ Correct Answer (ground truth)

Plain-English answer: Deals with fewer than 2 active contacts (single-threaded) or fewer than 3, or all in one persona (under-threaded), each flagged with the missing persona.
The answer in one line: active_window_start = 2026-07-07
active_window_start2026-07-07
allowed_new59 items — e.g. C-077A0E, C-0D15DF, C-1BDFE0, C-2AE3AA
best_persona_ruleDS2/DS3: economic buyer > IT security > finance > HR admin > champion; DS4/DS5: economic buyer > finance > IT security > HR admin > champion; first persona not present among active contacts
flag_ruleflag if active_contacts < 2 (single-threaded) or active_contacts < 3 or all active contacts share one persona (under-threaded); active = engaged within 60 days and not former
flaggeddeal_alias=Deal-36C33F; amount=15000; stage=DS2; active_count=1; reason=single-threaded; best_persona_to_add=economic buyer | deal_alias=Deal-50D386; amount=36000; stage=DS2; active_count=2; reason=under-threaded; best_persona_to_add=economic buyer | deal_alias=Deal-5408B0; amount=14850; stage=DS4; active_count=2; reason=under-threaded; best_persona_to_add=economic buyer | deal_alias=Deal-5BFE3B; amount=18000; stage=DS2; active_count=2; reason=under-threaded; best_persona_to_add=economic buyer | deal_alias=Deal-885F45; amount=9300; stage=DS2; active_count=2; reason=under-threaded; best_persona_to_add=IT security | deal_alias=Deal-92D97D; amount=60000; stage=DS2; active_count=1; reason=single-threaded; best_persona_to_add=economic buyer … (+5 more)
healthy_not_flaggedDeal-4B0BEB, Deal-84DBA6, Deal-D348E1
persona_universeeconomic buyer, champion, HR admin, IT security, finance
graded checksflagged_recall · persona_language · single_thread_term · on_file_fit

Model's actual answer

show the model's full answer
BASIS
- Today = 2026-10-19. Active cutoff = 2026-10-19 minus 60 days = 2026-08-20 (inclusive). Active contact = last_engaged_date >= 2026-08-20 AND is_former = false.
- Boundary days: 2026-08-20 = exactly 60 days, counted active (affects Deal-84DBA6 and Deal-C6D97A); 2026-08-19 = 61 days, excluded.
- Flags: single-threaded = active contacts < 2. Under-threaded = active contacts < 3 OR all contacts in one persona. "Personas present" = personas among active contacts.

NOT PROVIDED (data gaps)
- Amount: no deal-attribute data was provided (only the two contact files). Amount cannot be reported for any deal.
- Stage: not provided. Because "most valuable persona to add" is stage-dependent, I used a stated stage-agnostic rule instead: (1) no active economic buyer → add economic buyer; (2) economic buyer active but no active champion → add champion.
- Open/closed status is not verifiable from the provided files; all 14 deals in deal_contacts.csv were treated as in scope.

FLAGGED DEALS (11 of 14)

Deal-EC3025 (61032318100, C-FDD0C7)
- Amount: not on file. Stage: not on file.
- Active = 1: CT-047C54 champion 2026-09-02 IN; CT-F2C1AE economic buyer 2026-08-15 OUT (outside window, is_former=true).
- Flag: single-threaded (1 < 2).
- Personas present: champion. Missing: economic buyer, HR admin, IT security, finance.
- Most valuable add: economic buyer (active champion; prior CPO marked former).
- On-file unengaged fit: CT-6827DB, Chief People Officer, economic buyer — fits.

Deal-92D97D (59728118877, C-E23238)
- Amount: not on file. Stage: not on file.
- Active = 1: CT-01F5B4 HR admin 2026-08-28 IN; CT-A902AE champion 2026-06-01 OUT (outside window, not former).
- Flag: single-threaded (1 < 2).
- Personas present: HR admin. Missing: economic buyer, champion, IT security, finance.
- Most valuable add: economic buyer (rule 1). Secondary option: re-engage lapsed champion CT-A902AE (still on file, stale since 2026-06-01).
- On-file unengaged: none on file for C-E23238.

Deal-50D386 (61055128146, C-EB10E4)
- Amount: not on file. Stage: not on file.
- Active = 2: CT-AA41B2 champion 2026-09-01 IN; CT-B9C35B HR admin 2026-08-25 IN.
- Flag: under-threaded by count only (2 < 3); two distinct personas, so not single-threaded and not one-persona.
- Personas present: champion, HR admin. Missing: economic buyer, IT security, finance.
- Most valuable add: economic buyer.
- On-file unengaged fit: CT-A1C4B3, Chief People Officer, economic buyer — fits.

Deal-D0D6B5 (60081655042, C-32918E)
- Amount: not on file. Stage: not on file.
- Active = 1: CT-87CED4 champion 2026-09-02 IN; CT-DE6D7C champion 2026-08-19 OUT (61 days, one day short); CT-FD70B2 champion 2026-08-07 OUT.
- Flag: single-threaded (1 < 2); additionally all contacts on the deal are champion (persona clause would also fire).
- Personas present: champion. Missing: economic buyer, HR admin, IT security, finance.
- Most valuable add: economic buyer.
- On-file unengaged fit: CT-1FA4DB, Chief People Officer, economic buyer — fits.

Deal-5BFE3B (51674270311, C-535D36)
- Amount: not on file. Stage: not on file.
- Active = 1: CT-57123B champion 2026-08-31 IN; CT-5CE757 champion 2026-08-12 OUT.
- Flag: single-threaded (1 < 2).
- Personas present: champion. Missing: economic buyer, HR admin, IT security, finance.
- Most valuable add: economic buyer.
- On-file unengaged: none on file for C-535D36.

Deal-36C33F (63739413805, C-077A0E)
- Amount: not on file. Stage: not on file.
- Active = 0: CT-4FE556 IT security 2026-08-15 OUT (65 days, stale not former); CT-405B45 champion 2026-08-10 OUT + is_former=true; CT-86B22F economic buyer 2026-07-30 OUT + is_former=true.
- Flag: single-threaded (0 < 2) — no active thread at all.
- Personas present: none. Missing: all five (economic buyer, champion, HR admin, IT security, finance).
- Most valuable add: economic buyer (restart point).
- On-file unengaged fit: CT-1DB73E, Chief People Officer, economic buyer — fits.

Deal-885F45 (60686135564, C-5E8EFB)
- Amount: not on file. Stage: not on file.
- Active = 1: CT-51C81E economic buyer (VP People) 2026-08-26 IN; CT-D9A0E8 champion 2026-08-11 OUT.
- Flag: single-threaded (1 < 2).
- Personas present: economic buyer. Missing: champion, HR admin, IT security, finance.
- Most valuable add: champion (rule 2: economic buyer active, no active champion).
- On-file unengaged fit for champion: none on file. Secondary: on-file unengaged CT-B3F25D, IT Security Lead, IT security — fits a missing persona but not the top gap.

Deal-FCBE5B (62639586615, C-737030)
- Amount: not on file. Stage: not on file.
- Active = 1: CT-4A5317 champion 2026-08-29 IN (only contact on the deal).
- Flag: single-threaded (1 < 2).
- Personas present: champion. Missing: economic buyer, HR admin, IT security, finance.
- Most valuable add: economic buyer.
- On-file unengaged: none on file for C-737030.

Deal-5408B0 (60182332309, C-2AE3AA)
- Amount: not on file. Stage: not on file.
- Active = 1: CT-D33AE4 champion 2026-09-01 IN; CT-8742FD HR admin 2026-08-18 OUT (62 days, two days short).
- Flag: single-threaded (1 < 2).
- Personas present: champion. Missing: economic buyer, HR admin, IT security, finance. (HR admin contact exists on the deal but is inactive — re-engagement candidate rather than net-new.)
- Most valuable add: economic buyer.
- On-file unengaged fit: CT-07FA76, Chief People Officer, economic buyer — fits.

Deal-C6D97A (62121783047, C-5A8FC2)
- Amount: not on file. Stage: not on file.
- Active = 2: CT-223DDC champion 2026-08-31 IN; CT-B03555 champion 2026-08-20 IN (boundary day, exactly 60); CT-4E8A2B champion 2026-08-05 OUT.
- Flag: under-threaded (2 < 3) AND all active contacts in one persona (champion).
- Personas present: champion. Missing: economic buyer, HR admin, IT security, finance.
- Most valuable add: economic buyer.
- On-file unengaged: none on file for C-5A8FC2.

Deal-F9A08A (49757401138, C-0D15DF)
- Amount: not on file. Stage: not on file.
- Active = 1: CT-931B10 champion 2026-09-03 IN; CT-913581 economic buyer (CPO) 2026-06-20 OUT (stale, not former).
- Flag: single-threaded (1 < 2).
- Personas present: champion. Missing: economic buyer, HR admin, IT security, finance.
- Most valuable add: economic buyer.
- On-file unengaged fit: CT-697541, Chief People Officer, economic buyer — fits. (Distinct contact from the deal's lapsed CPO CT-913581; both are on file at C-0D15DF.)

NOT FLAGGED (3)
- Deal-84DBA6 (63929535929): 3 active — champion 2026-09-02, economic buyer 2026-08-30, IT security 2026-08-20 (boundary, in).
- Deal-4B0BEB (61038797752): 3 active — champion 2026-09-01, economic buyer 2026-08-28, HR admin 2026-08-21; finance 2026-08-09 stale.
- Deal-D348E1 (61750885954): 4 active — champion 2026-09-03, economic buyer 2026-09-02, IT security 2026-08-30, finance (CFO) 2026-08-27; HR admin 2026-08-19 stale (61 days).

Summary: 9 deals single-threaded (including Deal-36C33F with zero active contacts), 2 deals under-threaded only (Deal-50D386, Deal-C6D97A). Amount and stage were not present in the provided data and are reported as missing for every flagged deal.
talk-track-from-transcriptsPASS 1.00
Plain-English question: From this rep's last 10 calls: how do they open, handle objections, and how often do they lock a next step?
Prompt — what the model saw
From this rep's last 10 calls: what they lead with in the first five minutes, how they handle the three most common objections, how often a concrete next step is agreed (rate), and every competitor a prospect raised. Quote at most one line per finding and only lines that appear in the transcripts. End with two coaching notes.
cost $0.0662 · 32s · in 3,482 / out 7,070 tok
✓ objection_count✓ next_step_rate✓ competitors✓ coaching✓ quote_used

✓ Correct Answer (ground truth)

Plain-English answer: 8 of 10 calls open with the same customer-win story; objection handling, next-step agreement rate, and every competitor raised, with quotes.
The answer in one line: n_transcripts = 10; objection_count = 3; rep = Alex Franklin (+3 more below)
n_transcripts10
objection_count3
allowed_new13 items — e.g. TT-001, TT-002, TT-003, TT-004
competitor_quotesAwardco=We're also in late talks with Awardco — their rewards catalog looks bigger than yours.; Kudos=How are you different from Kudos? Our CEO used them at her last company.
competitors_prospect_raisedAwardco, Kudos
next_stepagreed=7; total=10; rate=0.7
objectionsobjection=budget; count=4; quote=Honestly, budget is locked until next fiscal year — I can't add a new line item right now.; transcript_id=TT-001 | objection=timing; count=3; quote=This is good, but can we revisit it next quarter? Open enrollment starts in October and we're underwater.; transcript_id=TT-002 | objection=in-house; count=3; quote=We already do recognition with a spreadsheet and quarterly gift cards — why would we change?; transcript_id=TT-004
opener_themetheme=opens with a customer recognition win story (400-person retailer, 18% turnover reduction); count=8; total=10; quote=Quick one to kick off: a 400-person retailer we work with cut regretted turnover 18% in two quarters after automating milestone awards, and I wanted to walk you through exactly how they did it.; transcript_id=TT-001
repAlex Franklin
rep_raised_competitors_excludedWorkhuman
graded checksobjection_count · next_step_rate · competitors · coaching · quote_used

Model's actual answer

show the model's full answer
CALL ANALYSIS: ALEX FRANKLIN, LAST 10 CALLS (TT-001 through TT-010, 2026-07-28 to 2026-09-03)

1) WHAT HE LEADS WITH (FIRST 5 MINUTES)
- 8 of 10 calls (TT-001, 002, 003, 005, 006, 007, 008, 010) open at minute 0 with the same customer proof story: "Quick one to kick off: a 400-person retailer we work with cut regretted turnover 18% in two quarters after automating milestone awards, and I wanted to walk you through exactly how they did it."
- TT-004 (Deal-403845) leads with an agenda: "I put together a short agenda — security review first, then pricing."
- TT-009 (Deal-1E2498, a follow-up call) leads with pricing: "You asked for straight pricing last time, so let's start there."
- Note: in TT-005 (Deal-C61CF7) he also adds a competitor comparison himself at minute 2: "And unlike Workhuman, our pricing includes the full rewards catalog with no extra margin." — Workhuman is rep-raised, not prospect-raised (see section 4).

2) THE THREE MOST COMMON OBJECTIONS AND HIS HANDLING
Counts (all prospect lines, minute 6 unless noted):
- Budget locked: 4 occurrences (TT-001, TT-003, TT-006, TT-010) — "Honestly, budget is locked until next fiscal year — I can't add a new line item right now."
  Handling, identical verbatim in all 4 (TT-001 m8): "Totally fair. Most teams fund this out of turnover savings — that retailer saved about $210k in avoided backfills, which is how their finance team signed off."
- Timing / revisit next quarter: 3 occurrences (TT-002, TT-005, TT-008) — "This is good, but can we revisit it next quarter? Open enrollment starts in October and we're underwater."
  Handling, identical verbatim in all 3 (TT-002 m8): "Makes sense. What if we scope a 90-day pilot with one department so you have internal data before next quarter's planning?"
- Status quo (spreadsheet + gift cards): 3 occurrences (TT-004, TT-007, TT-009) — "We already do recognition with a spreadsheet and quarterly gift cards — why would we change?"
  Handling, identical verbatim in all 3 (TT-004 m8): "Spreadsheets work until they scale — the difference is automation: milestones fire without HR lifting a finger, and you get analytics on who is being recognized."
(Next most common, below the top three: committee-wait, 2 occurrences — TT-004 m11, TT-010 m11.)

3) CONCRETE NEXT STEP AGREED — RATE
- Ask used (m14, identical in 7 calls): "Should we lock the next step — a working session with your team this week?"
- Agreed: 7 of 10 calls = 70% (7/10 = 0.70). Agreements: TT-001, TT-002, TT-003, TT-005, TT-006, TT-008, TT-009 — each with the prospect line "Yes, Thursday at 2pm works — send the invite and I'll bring our HRIS manager."
- Conversion on ask: 7/7 = 100%. The 3 calls with no ask and no next step (TT-004, TT-007, TT-010) all ended after a committee or "think about it" stall, with the rep conceding ("Understood — I'll leave it with you." / "Fair enough." / "Understood, thanks for the candor.").

4) COMPETITORS RAISED BY A PROSPECT
- Awardco — TT-003, Deal-547B2B, m4: "We're also in late talks with Awardco — their rewards catalog looks bigger than yours."
- Kudos — TT-007, Deal-EDC141, m4: "How are you different from Kudos? Our CEO used them at her last company."
- Excluded per the ask: Workhuman (TT-005 m2) was named by the rep, not the prospect.

COACHING NOTES
1. Close the stall gap. His standard ask converts 7-for-7, but in all 3 calls that hit a committee or "no urgency" stall (TT-004, TT-007, TT-010) he never made any ask and accepted the stall outright. Give him a fallback ask for those moments — e.g., scheduling a committee readout or a pilot-scoping call — so every call ends with a scheduled event, not "Understood — I'll leave it with you."
2. Break the full-script pattern and add discovery. The opener and all three objection responses are word-for-word identical on every occurrence (same story opens 8 of 10 calls; the $210k retailer story does double duty as opener and budget handler), and across 10 transcripts he asks zero discovery questions — every prospect fact (budget, committee, current tools) was volunteered. Tailoring the proof point to each prospect's own size and turnover math would make the budget reframe stickier; as scripted, it failed to prevent a committee stall in TT-010.
reporting-analytics 5 tests
q3-weighted-forecastPASS 1.00
Plain-English question: Build the Q3 2026 weighted forecast from the open pipeline, and show the math deal by deal.
Prompt — what the model saw
Below are the open deals in the default sales pipeline with a close date between 2026-07-01 and 2026-10-15, pulled from the CRM warehouse on 2026-09-05. Stages: DS1 (discovery) through DS5 (contract). Build the Q3 2026 forecast (quarter = 2026-07-01 to 2026-09-30). Weighting rule: 100 percent of COMMIT plus 35 percent of BEST_CASE. PIPELINE deals count zero. Only deals with a close date inside the quarter count. Report: COMMIT total, BEST_CASE total, weighted forecast, count of deals in each category inside the quarter, the deals you excluded for being outside the quarter (count and total amount), and the top 5 BEST_CASE deals by amount inside the quarter. Show arithmetic. Then, in a final '## Data quality' section of no more than four sentences, name the data quality problems in this extract that would make the forecast unreliable if it were run unattended.
cost $0.2396 · 88s · in 12,368 / out 25,677 tok
✓ commit_total_44729✓ best_case_total_203565✓ weighted_115976.75✓ commit_count_7✓ best_case_count_24✓ excluded_count_32✓ excluded_amount_227575✓ top5_aliases✓ flags_blank_owner✓ flags_missing_why_buys

✓ Correct Answer (ground truth)

Plain-English answer: A stage-weighted Q3 forecast number computed from the open pipeline, with per-stage math that reconciles to the deal data.
The answer in one line: excluded_amount = 227575; best_case_total = 203565; weighted_forecast = 115,976.75; commit_total = 44729; excluded_out_of_quarter = 32; best_case_count = 24; flags = blank owner column + empty why_buys fields must be surfaced (+2 more below)
excluded_amount227,575
best_case_count24
best_case_total203,565
commit_count7
commit_total44,729
excluded_out_of_quarter32
weighted_forecast115,976.75
flagsblank owner column + empty why_buys fields must be surfaced
top5_best_caseDeal-2D7423, Deal-25F752, Deal-E53952, Deal-5EED42, Deal-FA32A0
graded checkscommit_total_44729 · best_case_total_203565 · weighted_115976.75 · commit_count_7 · best_case_count_24 · excluded_count_32 · excluded_amount_227575 · top5_aliases · flags_blank_owner · flags_missing_why_buys

Model's actual answer

show the model's full answer
Q3 2026 Forecast (quarter 2026-07-01 to 2026-09-30; extract pulled 2026-09-05)

COMMIT total: 44,729 (7 deals)
- Deal-547B2B 11,200 + Deal-B7EBD1 9,000 + Deal-403845 9,000 + Deal-A2B47C 6,360 + Deal-2465CE 5,400 + Deal-A5E80A 2,520 + Deal-499BF6 1,249 = 44,729

BEST_CASE total: 203,565 (24 deals)
- 38,935 + 24,000 + 19,656 + 16,250 + 11,116 + 10,800 + 10,500 + 9,890 + 9,720 + 9,000 + 7,200 + 3,840 + 3,780 + 3,600 + 3,240 + 3,150 + 3,120 + 3,060 + 2,916 + 2,760 + 2,484 + 2,100 + 1,920 + 528 = 203,565

Weighted forecast (100% COMMIT + 35% BEST_CASE):
- 44,729 + 0.35 × 203,565 = 44,729 + 71,247.75 = 115,976.75

Counts inside the quarter (54 deals, 449,931.40 total):
- COMMIT: 7 deals / 44,729
- BEST_CASE: 24 deals / 203,565
- PIPELINE: 23 deals / 201,637.40 (weighted at 0 per the rule)

Excluded for close date outside 2026-07-01 to 2026-09-30: 32 deals, 227,575 total
- COMMIT: 1 deal / 13,770 (Deal-D348E1, DS5, closes 2026-10-15)
- BEST_CASE: 9 deals / 28,240
- PIPELINE: 22 deals / 185,565
- Check: 54 + 32 = 86 rows; 44,729 + 203,565 + 201,637.40 + 227,575 = 677,506.40 total extract

Top 5 BEST_CASE deals inside the quarter:
1. Deal-2D7423 — 38,935 — DS3 — 2026-09-30
2. Deal-25F752 — 24,000 — DS4 — 2026-09-25
3. Deal-E53952 — 19,656 — DS4 — 2026-09-30
4. Deal-5EED42 — 16,250 — DS3 — 2026-09-30
5. Deal-FA32A0 — 11,116 — DS3 — 2026-09-25

## Data quality

Forecast-category contradicts stage on four rows: COMMIT sits on a DS1 (Deal-A5E80A, 2,520) and a DS2 (Deal-499BF6, 1,249) that count 100% into the forecast, plus a DS5 BEST_CASE (Deal-C61CF7) and a DS4 PIPELINE (Deal-6787C2, 7,000), so the category field appears manually overridden or out of sync with stage. Four open deals carry close dates on or before the 2026-09-05 extract date (Deal-31AD2C, Deal-333EBB, Deal-57FF13, Deal-7A2454), meaning dates are stale rather than genuinely pending. Qualification evidence is absent exactly where it matters most: why_buys_chars is 0 on 22 of 24 in-quarter BEST_CASE deals and all 7 COMMIT deals, and owner is blank on 85 of 86 rows, leaving no accountable owner or rationale for unattended runs to sanity-check. Duplicate-looking clusters inflate the total — two 18,000 PIPELINE deals (Deal-CFE1E8, Deal-4F775F), two 9,000 COMMIT deals closing the same day (Deal-B7EBD1, Deal-403845), and two 2,880 PIPELINE deals (Deal-FA053A, Deal-333EBB) — suggesting split or recreated opportunities.
aha-moment-2x2PASS 1.00
Plain-English question: Do companies that hit 5+ givers and one redemption in month one actually retain better at 24 months?
Prompt — what the model saw
Test the activation hypothesis on the company cohort below: companies with 5+ unique givers (m1_users >= 5) AND 1+ successful redemption (m1_redemptions >= 1) in their first calendar month retain better at 24 months than those with one signal or neither. Every company in this extract is already 25+ months old, so current_status = 'active' means retained at 24 months. Report the full 2x2 (cohort size and 24-month retention rate for: both signals, givers-only, redemption-only, neither), state how many companies were excluded from the denominator and why (if none, say so), name the single signal with the largest retention lift, and state what this does and does not prove.
cost $0.5937 · 191s · in 39,115 / out 60,143 tok
✓ both_66✓ givers_47✓ redemp_31✓ neither_40✓ excluded_0✓ largest_lift_givers

✓ Correct Answer (ground truth)

Plain-English answer: The four cohort cells are graded against computed 24-month retention; the numbers come from the scorer's cohort table, not opinion.
The answer in one line: both_5plus_givers_and_redemption = 66.0% (31 of 47)
both_5plus_givers_and_redemption66.0% (31 of 47)
exclusionsnone — claim bounded as correlation, not causation
givers_only46.9% (23 of 49)
largest_single_signal_liftunique givers
neither40.0% (38 of 95)
redemption_only31.0% (9 of 29)
graded checksboth_66 · givers_47 · redemp_31 · neither_40 · excluded_0 · largest_lift_givers

Model's actual answer

show the model's full answer
Hypothesis test on plg_company_cohort_2023 (220 companies, signup months 2023-01 through 2023-07). Denominator = all 220 rows; **0 companies excluded** — m1_users and m1_redemptions are populated for every record, no duplicate company_keys, and every company is 25+ months old, so current_status = 'active' is treated as retained at 24 months. Status mix: 101 active / 116 cancelled / 3 non_renewing.

Classification: G = m1_users >= 5; R = m1_redemptions >= 1; retained = current_status == 'active'.

2x2 (cohort size, 24-month retention rate):

| Segment | n | Retained | Retention rate (arithmetic) |
|---|---|---|---|
| Both signals (G and R) | 47 | 31 | 31/47 = 66.0% |
| Givers only (G, not R) | 49 | 23 | 23/49 = 46.9% |
| Redemption only (R, not G) | 29 | 9 | 9/29 = 31.0% |
| Neither | 95 | 38 | 38/95 = 40.0% |
| Total | 220 | 101 | 101/220 = 45.9% |

Largest single-signal lift vs neither: **5+ unique givers (m1_users >= 5)** — +6.9pp (46.9% vs 40.0%). Holding givers < 5, redemption alone is associated with a *drop*: 31.0% vs 40.0% = −9.0pp. Decomposition: givers effect among no-redemption companies = +6.9pp; adding redemption on top of 5+ givers = +19.0pp (66.0% − 46.9%); both-vs-neither (+26.0pp) far exceeds the sum of the two single-signal lifts (−2.0pp), i.e., the signals interact rather than add.

Verdict: the hypothesis is partially supported. The compound activation signal (5+ givers AND ≥1 redemption) retains best by a wide margin, but "one signal is better than none" is only true for giver breadth — a redemption without giver breadth is worse than doing neither.

What this does NOT prove:
- Causation. This is observational, single-cohort, single-vendor extract; 5+ givers and redemptions may be markers of company size, engagement budget, or intent rather than causes of retention.
- Statistical confidence at the single-signal level. Givers-only vs neither is 23/49 vs 38/95 (z ≈ 0.8, far from conventional significance) and redemption-only is n=29; only the both-vs-neither gap (31/47 vs 38/95) is large relative to cell sizes.
- Generalizability beyond 2023 signups in this extract; 3 non_renewing companies are counted as not retained, and no field distinguishes them from cancelled — if non-renewal is a softer outcome than cancellation, the "neither"/givers-only cells could shift slightly.

⚠️ File-mutation verifier: 1 file edit(s) FAILED this turn despite any wording above that may suggest otherwise. Run `git status` or `read_file` to confirm what actually landed.
  • `/Users/amaniphipps/.hermes/cache/scratch/plg_company_cohort_2023.csv` — [write_file] Refusing to overwrite `/Users/amaniphipps/.hermes/cache/scratch/plg_company_cohort_2023.csv`: `/Users/amaniphipps/.hermes/cache/scratch/plg_company_cohort_2023.csv` exists but this ta…
arr-rollforward-reconciliationPASS 1.00
Plain-English question: Does CRM ARR match billing ARR — and if not, exactly where is the variance?
Prompt — what the model saw
Reconcile ARR between CRM company records and billing subscriptions as of 2026-09-05. Billing ARR is MRR x 12. Report both totals, the variance, and a decomposition into buckets (status mismatch, rounding, missing records, other) that sums exactly to the variance. List each mismatched account with a suggested owner. Business rule: any subscription with a term other than 12 months must have cf_agreement_end_date populated; list violations.
cost $0.1369 · 63s · in 6,175 / out 14,961 tok
✓ billing_total✓ crm_total✓ variance✓ buckets_sum✓ violations_listed

✓ Correct Answer (ground truth)

Plain-English answer: Billing $604,739.28 vs CRM $603,581.76: a $1,157.52 variance decomposed into status-mismatch, rounding, and missing-record buckets.
The answer in one line: n_companies = 40; n_mismatched_accounts = 7; as_of = 2026-09-05 (+4 more below)
n_companies40
n_mismatched_accounts7
as_of2026-09-05
billing_arr_total604739.28
bucket_sum_equals_variance1
buckets{"status_mismatch": {"amount": "-13158.48", "accounts": [{"company_alias": "C-0C8323BF", "billing_arr": "0.00", "hubspot_arr": "4905.24", "diff": "-4905.24", "suggested_owner": "Dana Mercer"}, {"compa
crm_arr_total603581.76
ruleterm_months != 12 requires cf_agreement_end_date
variance1157.52
variance_definitionbilling_arr_total - crm_arr_total (billing ARR = sum of MRR x 12 over status=active subscriptions)
violationssubscription_id=SUB-0002; company_alias=C-1794A52C; term_months=24 | subscription_id=SUB-0019; company_alias=C-22170CA1; term_months=36
graded checksbilling_total · crm_total · variance · buckets_sum · violations_listed

Model's actual answer

show the model's full answer
# ARR Reconciliation — CRM vs Billing, as of 2026-09-05

Convention: Billing ARR = Σ (MRR × 12) over **active** subscriptions only. Cancelled subscriptions carry no ARR as of the date; where CRM still carries one, that shows up as a status-mismatch. All deltas below are CRM − Billing.

**Totals**
- Billing ARR (active only): Σ active MRR = 50,394.94 × 12 = **$604,739.28** (37 active subs)
- CRM ARR: Σ hubspot_arr over 39 rows = **$603,581.76**
- Variance: 603,581.76 − 604,739.28 = **−$1,157.52** (billing higher than CRM)
- If cancelled subs were counted, billing = 51,491.48 × 12 = $617,897.76 and variance would be −$14,316.00; the active-only basis is used below.

**Decomposition (sums exactly to −1,157.52)**

| Bucket | Amount | Composition |
|---|---|---|
| Status mismatch | +13,158.48 | C-0C8323BF +4,905.24; C-0DC4FB8C +8,253.24 |
| Rounding | +36.00 | C-0D66DF9E +16.00; C-14D70CE0 +20.00 |
| Missing records | −11,952.00 | C-0D5BBE3A +16,497.24 (CRM only); C-21629AA4 −28,449.24 (billing only) |
| Other | −2,400.00 | C-0F7269D7 −2,400.00 |
| **Total** | **−1,157.52** | 13,158.48 + 36.00 − 11,952.00 − 2,400.00 = −1,157.52 |

All other 34 matched aliases reconcile exactly (CRM ARR = MRR × 12 to the cent).

**Mismatched accounts and suggested owners**
The provided files contain no owner, rep, or team fields, so owner names cannot be derived from the data; suggestions below are role-based assignments by issue type.

- C-0C8323BF — CRM 4,905.24 vs billing $0; only sub SUB-000E is cancelled (MRR 408.77; 408.77 × 12 = 4,905.24 exactly). Suggested owner: CRM data hygiene owner (HubSpot ARR) — zero out ARR, confirm churn date with CS.
- C-0DC4FB8C — CRM 8,253.24 vs billing $0; only sub SUB-000F is cancelled (MRR 687.77; × 12 = 8,253.24 exactly). Suggested owner: CRM data hygiene owner with CS.
- C-0D5BBE3A — CRM 16,497.24, no subscription exists in billing (implied MRR 1,374.77). Suggested owner: Billing/Finance ops (Chargebee) — locate or create the subscription.
- C-21629AA4 — active sub SUB-0004, MRR 2,370.77 × 12 = 28,449.24, no CRM company record. Suggested owner: Sales/RevOps CRM admin — create company record and link.
- C-0F7269D7 — CRM 24,396.00 vs billing 2,233.00 × 12 = 26,796.00 (CRM lower by exactly 2,400.00 = 200.00/mo). Suggested owner: Deal desk / account owner — verify contracted MRR (possible discount or seat change not synced).
- C-0D66DF9E — CRM 23,200.00 vs 1,932.00 × 12 = 23,184.00 (+16.00). Suggested owner: CRM data owner — low priority.
- C-14D70CE0 — CRM 18,200.00 vs 1,515.00 × 12 = 18,180.00 (+20.00). Suggested owner: CRM data owner — low priority.

Note on the rounding bucket: the +16.00 and +20.00 are the only small residuals; the files don't reveal their cause (e.g., rounded MRR in CRM), only their size.

**Business-rule violations (term ≠ 12 months requires cf_agreement_end_date)**
- SUB-0002 (C-1794A52C) — 24-month term, cf_agreement_end_date blank — VIOLATION
- SUB-0019 (C-22170CA1) — 36-month term, cf_agreement_end_date blank — VIOLATION

Compliant non-12 subs for reference: SUB-000C (C-0DB48281, 24 mo, 2027-11-30), SUB-001A (C-0FC4DBB8, 36 mo, 2027-11-30).
kvm-month-over-monthPASS 1.00
Plain-English question: How did the four core KVMs move in August versus July?
Prompt — what the model saw
Report the four core KVMs for 2026-08 against 2026-07: giving rate, redemptions per user, 1:1 meetings engagement, pulse check engagement. For each: value, prior value, absolute and relative change, direction. Name the KVM with the largest relative move and the segment (plan_tier or size_band) driving it if the data supports that; otherwise say the data does not. Under 200 words plus one table.
cost $0.1174 · 40s · in 8,484 / out 11,553 tok
✓ four_kvms✓ biggest_mover_named✓ direction_correct✓ segment_named✓ word_cap

✓ Correct Answer (ground truth)

Plain-English answer: Giving rate 60.27% (up 0.07%), redemptions per user 1.7302 (up 0.01%), with 1:1 meetings and pulse engagement each up fractionally — direction reported for all four.
The answer in one line: n_company_rows_per_month = 30; aggregation = simple mean of company-level values per month
n_company_rows_per_month30
aggregationsimple mean of company-level values per month
biggest_moverpulse_engagement
driving_segmentsize_band=enterprise; 2026-07=0.55; 2026-08=0.2743; rel_change_pct=-50.13
kvms{"giving_rate": {"2026-07": 0.6023, "2026-08": 0.6027, "abs_change": 0.0004, "rel_change_pct": 0.07, "direction": "up"}, "redemptions_per_user": {"2026-07": 1.73, "2026-08": 1.7302, "abs_change": 0.00
segment_detail{"smb": {"2026-07": 0.6588, "2026-08": 0.6573, "rel_change_pct": -0.23}, "mid_market": {"2026-07": 0.593, "2026-08": 0.5942, "rel_change_pct": 0.2}, "enterprise": {"2026-07": 0.55, "2026-08": 0.2743,
graded checksfour_kvms · biggest_mover_named · direction_correct · segment_named · word_cap

Model's actual answer

show the model's full answer
**2026-08 vs 2026-07 (equal-weighted mean across 30 companies; plan_tier is tier_three for all).**

- Giving rate: 0.6023 → 0.6027, +0.0004 (+0.07%), up
- Redemptions per user: 1.7300 → 1.7302, +0.0002 (+0.01%), up
- 1:1 engagement: 0.4469 → 0.4472, +0.0003 (+0.06%), up
- Pulse engagement: 0.6006 → 0.5086, −0.0920 (−15.31%), down

Largest relative move: pulse engagement, driven entirely by the enterprise size_band. Enterprise pulse fell 0.5500 → 0.2743 (−0.2757, −50.13%), while smb (−0.22%) and mid_market (+0.21%) were flat. All 10 enterprise companies show ~−49% to −52% pulse drops (e.g., C-0BA71F12 −51.3%, C-0D0B047C −51.5%), so no single company drives it — the segment does.

| KVM | Jul | Aug | Δ abs | Δ rel | Dir |
|---|---|---|---|---|---|
| Giving rate | 0.6023 | 0.6027 | +0.0004 | +0.07% | ↑ |
| Redemptions/user | 1.7300 | 1.7302 | +0.0002 | +0.01% | ↑ |
| 1:1 engagement | 0.4469 | 0.4472 | +0.0003 | +0.06% | ↑ |
| Pulse engagement | 0.6006 | 0.5086 | −0.0920 | −15.31% | ↓ |
rewards-ytd-sectionPASS 1.00
Plain-English question: Produce the redemption section of the weekly rewards report through August.
Prompt — what the model saw
Produce the redemption section of the weekly rewards report through the last completed month (2026-08): redemption count, spend, unique redeemers, redemptions per redeemer, provider mix as percent of spend, top 5 countries by redemptions. Say which month is the last completed month. Provider shares must sum to 100.
cost $0.4634 · 112s · in 54,840 / out 38,926 tok
✓ count✓ spend✓ unique_redeemers✓ month_named✓ provider_mix✓ top_country

✓ Correct Answer (ground truth)

Plain-English answer: 378 redemptions, $27,846 spend, 235 unique redeemers at 1.61 each, custom rewards at 39% of spend, US leading all countries.
The answer in one line: redemption_count = 378; unique_redeemers = 235; redemptions_per_redeemer = 1.61; total_spend_usd = 27846.00
redemption_count378
redemptions_per_redeemer1.61
unique_redeemers235
provider_mix_pct_of_spendTangoCard=11.6; Tremendous=30.54; custom=39.05; Snappy=18.81
top5_countries_by_countcountry=US; count=244 | country=CA; count=24 | country=AU; count=21 | country=GB; count=17 | country=NL; count=17
total_spend_usd27846.00
graded checkscount · spend · unique_redeemers · month_named · provider_mix · top_country

Model's actual answer

show the model's full answer
Redemption section — last completed month is August 2026 (2026-08): the file's latest redemption is 2026-08-31 and the latest full calendar month is 2026-08. All 378 rows in the file fall in Jan 1 – Aug 31, 2026, so the entire file is the reporting window (378 unique redemption_ids, no duplicates).

Totals (YTD through 2026-08-31)
- Redemption count: 378
- Spend: $27,846.00
- Unique redeemers (distinct user_key): 235
- Redemptions per redeemer: 378 ÷ 235 = 1.6085 ≈ 1.61

Provider mix (% of spend)
| Provider | Spend | Arithmetic | Share |
|---|---|---|---|
| custom | $10,873.00 | 10,873 ÷ 27,846 = 39.0469% | 39.05% |
| Tremendous | $8,505.00 | 8,505 ÷ 27,846 = 30.5430% | 30.54% |
| Snappy | $5,238.00 | 5,238 ÷ 27,846 = 18.8106% | 18.81% |
| TangoCard | $3,230.00 | 3,230 ÷ 27,846 = 11.5995% | 11.60% |

Sum check: 39.05 + 30.54 + 18.81 + 11.60 = 100.00% (provider spends sum to 27,846.00, matching total).

Top 5 countries by redemptions
1. US — 244
2. CA — 24
3. AU — 21
4. GB — 17 (tie)
4. NL — 17 (tie)

Note: GB and NL tie for 4th at 17 redemptions each, so both are shown; next country (SG, 12) is outside the top 5.

⚠️ File-mutation verifier: 1 file edit(s) FAILED this turn despite any wording above that may suggest otherwise. Run `git status` or `read_file` to confirm what actually landed.
  • `/Users/amaniphipps/.hermes/cache/scratch/redemptions_ytd.csv` — [write_file] Refusing to overwrite `/Users/amaniphipps/.hermes/cache/scratch/redemptions_ytd.csv`: `/Users/amaniphipps/.hermes/cache/scratch/redemptions_ytd.csv` exists but this task has not seen …
customer-success 5 tests
churn-save-eligibilityPASS 1.00
Plain-English question: Which at-risk accounts actually qualify for a churn-save offer, and which play fits each?
Prompt — what the model saw
Which at-risk accounts qualify for a churn-save offer under the documented eligibility rules, what amount is at stake per account and in total, and which play fits each (usage revival, executive touch, commercial concession)? Cite the signal that justifies each play. List accounts that look at risk but do not qualify and why.
cost $0.1389 · 66s · in 2,066 / out 17,235 tok
✓ eligible_set✓ total_at_stake✓ plays_cited✓ noneligible_named✓ rules_applied

✓ Correct Answer (ground truth)

Plain-English answer: Only accounts meeting all three rules qualify — e.g. C-0F6C0F34 at $49,707 (executive touch) and C-0B827671 at $25,365 (usage revival) — with amount and play per account.
The answer in one line: n_accounts = 30; snapshot_date = 2026-09-05
n_accounts30
eligibleaccount_alias=C-0F6C0F34; amount_at_stake=49707.00; play=executive touch; justifying_signal=champion_active is false - no executive sponsor engaged | account_alias=C-0B827671; amount_at_stake=25365.00; play=usage revival; justifying_signal=usage_trend_3m=declining over the last 3 months | account_alias=C-0B360C78; amount_at_stake=35748.00; play=commercial concession; justifying_signal=usage stable/growing with seat utilization 75% - risk is commercial, not adoption | account_alias=C-0B0F1BAB; amount_at_stake=5494.00; play=executive touch; justifying_signal=champion_active is false - no executive sponsor engaged | account_alias=C-0CA21961; amount_at_stake=16829.00; play=usage revival; justifying_signal=seat utilization 26% is below 50% | account_alias=C-0E9C27D1; amount_at_stake=41235.00; play=commercial concession; justifying_signal=usage stable/growing with seat utilization 85% - risk is commercial, not adoption … (+2 more)
non_eligible_at_riskaccount_alias=C-0BC71BDD; health_score=55 | account_alias=C-0BA71F12; health_score=52 | account_alias=C-0F6694C3; health_score=43 | account_alias=C-0BE96399; health_score=54 | account_alias=C-0F876796; health_score=47 | account_alias=C-0FCCD2DF; health_score=43 … (+1 more)
ruleshealth_score < 60, churn_save_eligible_amount > 0, renewal within 120 days of snapshot
snapshot_date2026-09-05
total_amount_at_stake224601.00
graded checkseligible_set · total_at_stake · plays_cited · noneligible_named · rules_applied

Model's actual answer

show the model's full answer
ELIGIBILITY LOGIC (all three rules must hold)
R1: health_score < 60. R2: churn_save_eligible_amount > 0. R3: renewal within 120 days of snapshot 2026-09-05.
Window arithmetic: 2026-09-05 + 120 days = 2027-01-03 (25 days to Sep 30 + 31 Oct + 30 Nov + 31 Dec = 117; +3 = Jan 3).

QUALIFIED — 8 ACCOUNTS, grouped by play

Usage revival (signal: usage_trend_3m = declining)
  C-0B827671  health 56  renews 2026-11-14 (70 days)   at stake $25,365  (champion active; 113/202 seats used = 56%)
  C-0D3278C7  health 54  renews 2026-11-12 (68 days)   at stake $17,602  (champion active; 126/380 seats used = 33%, weakest adoption in cohort)

Executive touch (signal: champion_active = false)
  C-0F6C0F34  health 51  renews 2026-10-03 (28 days)   at stake $49,707  (usage growing; 308/395 seats = 78%)
  C-0CEF69FD  health 53  renews 2026-11-21 (77 days)   at stake $32,621  (usage growing; 97/136 seats = 71%)
  C-0B0F1BAB  health 38  renews 2026-09-23 (18 days)   at stake $5,494   (usage flat; lowest health score in file)

Commercial concession (signal: usage flat/growing + champion active — no documented usage or relationship gap, so risk reads as commercial)
  C-0E9C27D1  health 39  renews 2026-09-24 (19 days)   at stake $41,235  (flat; 134/157 seats = 85%)
  C-0B360C78  health 57  renews 2026-10-28 (53 days)   at stake $35,748  (growing; 246/327 seats = 75%)
  C-0CA21961  health 58  renews 2026-12-28 (114 days)  at stake $16,829  (flat; see judgment note)

AMOUNT AT STAKE (= churn_save_eligible_amount)
  Usage revival: 25,365 + 17,602 = $42,967
  Executive touch: 49,707 + 32,621 + 5,494 = $87,822
  Commercial concession: 41,235 + 35,748 + 16,829 = $93,812
  Total: 42,967 + 87,822 + 93,812 = $224,601 (ARR behind these 8 accounts: $454,380)

Judgment note: C-0CA21961 pays for 325 seats but uses 84 (26%) — a severe adoption gap that could equally justify usage revival; assigned to commercial concession because its 3-month trend is flat and the champion is active. Flagged, not hidden.

AT RISK BUT NOT QUALIFIED (health < 60, fails a rule)

Fails R2 — eligible amount $0.00, nothing to concede against:
  C-0BC71BDD  health 55  renews 2026-10-27 (in window)  fails R2 only
  C-0BE96399  health 54  renews 2026-10-29 (in window)  fails R2 only
  C-10A56B0F  health 54  renews 2026-12-12 (in window)  fails R2 only
  C-0F6694C3  health 43  renews 2027-03-21  fails R2; also outside window (197 days)
  C-0FCCD2DF  health 43  renews 2027-04-23  fails R2; also outside window (230 days)

Fails R3 — renewal beyond the 120-day window (R1 and R2 pass):
  C-0F876796  health 47  renews 2027-02-06 (154 days)  $19,958 eligible but out of window
  C-0BA71F12  health 52  renews 2027-04-11 (218 days)  $6,824 eligible but out of window

NOTES
- Boundary check: closest qualifier is C-0CA21961 at 114 days; closest rejection is C-0F876796 at 154 days. No renewal date precedes the snapshot, so the window's lower bound never binds.
- No account with health ≥ 60 shows declining usage or a nonzero eligible amount, so no additional accounts look at risk on the documented signals.
- All 30 rows are complete; no missing data.
expansion-kitPASS 1.00
Plain-English question: Build the expansion kit for the customer who just replied to our outreach.
Prompt — what the model saw
A customer replied to outreach (reply included). Build the expansion kit: seat coverage (licensed over headcount), two lines on usage health, headroom in seats and ARR at the current per-seat rate, who replied and whether they can buy, the right buyer from our contacts, and a reply email under 150 words that moves toward the right person without being pushy and cites one real usage fact.
cost $0.0294 · 16s · in 1,166 / out 2,930 tok
✓ coverage_pct✓ headroom_arr✓ right_buyer✓ replier_not_buyer✓ usage_fact✓ email_under_150

✓ Correct Answer (ground truth)

Plain-English answer: 37.5% seat coverage, 250 headroom seats worth $15,000 ARR at $60 a seat, usage up 43.2% in six months — and the replier cannot buy; Dana R. (VP People) is the right buyer.
The answer in one line: headroom_seats = 250; coverage_pct = 37.5; company_alias = C-0DDFC9A7 (+3 more below)
headroom_seats250
coverage_pct37.5
company_aliasC-0DDFC9A7
headroom_arr15000.00
per_seat_annual_rate60.00
replierMaria S.
replier_can_buy0
right_buyername=Dana R.; title=VP People
usage_factsactive_users_2026_08=126; active_users_2026_03=88; growth_pct_6m=43.2; licensed_seat_utilization_pct=84
usage_health_linesMonthly active users grew from 88 (2026-03) to 126 (2026-08), +43.2% over 6 months., Last month 126 of 150 licensed seats were active (84.0% utilization), approaching the seat cap.
graded checkscoverage_pct · headroom_arr · right_buyer · replier_not_buyer · usage_fact · email_under_150

Model's actual answer

show the model's full answer
Expansion kit — C-0DDFC9A7

1. Seat coverage
150 licensed / 400 headcount = 0.375 = 37.5% of the company is licensed.

2. Usage health (two lines)
- Monthly active users rose every month Mar–Aug 2026: 88 → 95 → 102 → 110 → 118 → 126 (+38 users, +43%: 38/88 = 0.432).
- August utilization: 126 of 150 licensed seats active = 84%; licenses are nearly full while only 37.5% of headcount has seats.

3. Headroom at current per-seat rate
- Derived rate: $9,000.00 ARR / 150 seats = $60.00 per seat per year (rate is derived from the data, not stated).
- Seat headroom: 400 − 150 = 250 seats.
- ARR headroom: 250 × $60.00 = $15,000.00 incremental (400 × $60.00 = $24,000.00 full-coverage ARR vs. $9,000.00 today).

4. Who replied, and can they buy
Maria S., People Operations Coordinator, replied 2026-09-02 (matches her last_engaged in contacts). She cannot buy: "I'm not the right person for purchasing decisions here. Budget and seat expansion sit with Dana R." She offered to make an introduction and noted Dana R. has been asking about usage numbers.

5. Right buyer
Dana R., VP People — named in the reply as the budget/seat-expansion owner and present in our contacts (last engaged 2026-05-18, ~3.5 months before the reply). Sam K. (Office Manager) is not the buyer.

6. Reply email (79 words)

Subject: RE: Growing your team's recognition program

Hi Maria,

Thanks for the reply — great to hear the team is getting value. Yes, an introduction to Dana would help.

One fact from your usage: monthly active users grew from 88 in March to 126 in August, while 150 seats cover 400 employees. That gap is usually where teams expand.

Would an email introduction work? I can send a one-paragraph summary of these numbers you or Dana can forward internally.

Thanks,
Cole

Data notes: no list-price or discount data provided, so headroom uses the derived $60.00/seat rate; no company name beyond alias C-0DDFC9A7; no usage data after August 2026.
onboarding-statusPASS 1.00
Plain-English question: What is done and what is missing in this customer's onboarding before the mid-onboarding call?
Prompt — what the model saw
Prep the CSM for the mid-onboarding call: what is complete (integrations connected, allowance set, admins added, first recognition given, first redemption), what is not, early engagement signals, and the three things to cover on the call. Mark nothing complete without a data field that shows it.
cost $0.0426 · 28s · in 1,176 / out 4,694 tok
✓ complete_items✓ incomplete_items✓ three_items✓ grounded

✓ Correct Answer (ground truth)

Plain-English answer: 25 days live: Slack connected, allowance set, admins added — but the HRIS integration is not connected, with early engagement signals and three call priorities.
The answer in one line: days_live = 25; company_alias = C-0D284E42
days_live25
call_itemsConnect the HRIS integration (integration_hris is empty) to automate user provisioning and offboarding., Get to first redemption (first_redemption_at is empty): walk admins through the rewards catalog and confirm allowance funding., Build the giving habit: daily active givers averaged 4.3 in week 1 vs 13.0 last week - agree on manager nudges and a recognition cadence.
checklist{"slack_integration_connected": {"complete": true, "field": "integration_slack", "value": "2026-08-12"}, "hris_integration_connected": {"complete": false, "field": "integration_hris", "value": ""}, "a
company_aliasC-0D284E42
engagement_signalsavg_active_givers_week1=4.3; avg_active_givers_last_week=13; latest_active_givers=15; days_with_activity=25
graded checkscomplete_items · incomplete_items · three_items · grounded

Model's actual answer

show the model's full answer
MID-ONBOARDING CALL PREP — C-0D284E42
Data as of 2026-09-04 (last usage row). Signup 2026-08-11, so day 24.

COMPLETE (each backed by a data field)
- Slack integration: connected 2026-08-12 (day 1 after signup)
- Allowance set: 2026-08-13 (day 2)
- Admins added: 2 (count field)
- First recognition: 2026-08-15 14:22 (day 4)

NOT COMPLETE / NO DATA
- HRIS integration: integration_hris is empty. No timestamp, so unknown whether it was attempted.
- First redemption: first_redemption_at is empty. None recorded through day 24.

EARLY ENGAGEMENT SIGNALS (onboarding_usage.csv, 25 days, 2026-08-11 to 2026-09-04)
- Active givers up 5.0x: 3 on day 1 to 15 on 09-04 (15/3 = 5.0)
- Activity every single day: no zero-activity days
- Weekly averages rising (sum / days):
  - W1 08-11 to 08-17: 30/7 = 4.3
  - W2 08-18 to 08-24: 53/7 = 7.6 (+77%)
  - W3 08-25 to 08-31: 75/7 = 10.7 (+42%)
  - W4 09-01 to 09-04: 56/4 = 14.0 (+31%)
- The two most recent days tie the series peak (15 on 09-03 and 09-04); the largest post-week-1 dip is 5 on 08-18, so growth is steady, not spiky
- Caveats: total employee count is not in the data, so giver penetration cannot be computed. Also, usage shows 3 active givers on 08-11 (signup day), before first_recognition_at (08-15 14:22) and before allowance_set (08-13) — the two files' definitions do not line up; worth confirming on the call what active_givers counts versus what first_recognition_at captures.

THREE THINGS TO COVER ON THE CALL
1. HRIS connection — the only integration gap at day 24. Ask if it was attempted, and schedule it.
2. First redemption — none on record. Walk through how redemption works, verify rewards balance/funding, and prompt a first redemption.
3. Headcount and admin coverage — get total employee count so "15 active givers" can be judged as a share of the company, confirm 2 admins is the intended coverage, and reconcile the active-givers-vs-first-recognition timing question.
renewal-risk-conflicting-datesPASS 1.00
Plain-English question: Two systems disagree on renewal dates — which do you trust per account, and who is actually renewing inside 90 days?
Prompt — what the model saw
Produce a 90-day renewal risk brief. Two systems hold renewal dates and disagree on some accounts; decide which to trust per account and say why (multi-year contracts are known to be wrong in ChurnZero). For every renewal: company, CSM, ARR, date used, seat utilization, 3-month usage trend, risk rating with one sentence of evidence. Flag every disagreement. Close with total ARR renewing and ARR at risk.
cost $0.2948 · 211s · in 14,089 / out 33,247 tok
✓ total_renewing✓ arr_at_risk✓ disagreements_flagged✓ trust_rule

✓ Correct Answer (ground truth)

Plain-English answer: Chargebee wins per account (multi-year contracts are known wrong in ChurnZero), with each trusted date, the reason, and a 90-day-window flag.
The answer in one line: n_accounts = 20; n_disagreements = 5; snapshot_date = 2026-09-05 (+2 more below)
n_accounts20
n_disagreements5
accounts20 items — e.g. account_alias=C-0B144C78; csm=Cole Ingram; arr=30899.00; trusted_renewal_date=2026-11-02; trusted_source_why=systems agree (annual term); in_90d_window=True; dates_disagree=False; seat_utilization_pct=75.4; usage_3m_ratio=1.03; risk=low; evidence=3-month usage ratio 1.03 (last3 avg 103 vs prior3 100), seat utilization 75% | account_alias=C-0B20DB64; csm=Dana Mercer; arr=21770.00; trusted_renewal_date=2026-10-07; trusted_source_why=systems agree (annual term); in_90d_window=True; dates_disagree=False; seat_utilization_pct=56.6; usage_3m_ratio=1; risk=medium; evidence=3-month usage ratio 1.00 (last3 avg 295 vs prior3 295), seat utilization 57% | account_alias=C-0B344485; csm=Elena Sinclair; arr=64384.00; trusted_renewal_date=2026-11-16; trusted_source_why=systems agree (annual term); in_90d_window=True; dates_disagree=False; seat_utilization_pct=78; usage_3m_ratio=1.04; risk=low; evidence=3-month usage ratio 1.04 (last3 avg 241 vs prior3 231), seat utilization 78% | account_alias=C-0B369871; csm=Dana Mercer; arr=85128.00; trusted_renewal_date=2026-10-29; trusted_source_why=systems agree (annual term); in_90d_window=True; dates_disagree=False; seat_utilization_pct=75.1; usage_3m_ratio=1.04; risk=low; evidence=3-month usage ratio 1.04 (last3 avg 330 vs prior3 316), seat utilization 75%
arr_at_risk_90d359409.00
disagreementsaccount_alias=C-0B7D2C30; cz_renewal_date=2026-09-10; cb_renewal_date=2026-09-15; is_multi_year=True; trusted=2026-09-15 | account_alias=C-0BBE3E60; cz_renewal_date=2027-09-26; cb_renewal_date=2026-09-26; is_multi_year=True; trusted=2026-09-26 | account_alias=C-0BCDB8C2; cz_renewal_date=2027-09-18; cb_renewal_date=2026-09-18; is_multi_year=True; trusted=2026-09-18 | account_alias=C-0D2AB865; cz_renewal_date=2026-09-10; cb_renewal_date=2026-09-22; is_multi_year=True; trusted=2026-09-22 | account_alias=C-0F5D2323; cz_renewal_date=2026-09-10; cb_renewal_date=2026-09-29; is_multi_year=True; trusted=2026-09-29
snapshot_date2026-09-05
total_arr_renewing_90d1048715.00
trust_rulemulti-year contracts: Chargebee is authoritative (ChurnZero known wrong); otherwise systems agree or Chargebee wins
window2026-09-05 to 2026-12-04
graded checkstotal_renewing · arr_at_risk · disagreements_flagged · trust_rule

Model's actual answer

show the model's full answer
# 90-Day Renewal Risk Brief — window Sep 15, 2026 – Dec 14, 2026 (as-of Sep 15, 2026)

## Date reconciliation

Rule applied: ChurnZero renewal dates are known to be wrong for multi-year contracts (ChurnZero `is_multi_year=true` in chargebee_terms); trust Chargebee for those. Where both systems agree, the date is used as-is.

5 disagreements, all on the 5 multi-year accounts:

| Account | ChurnZero | Chargebee | Decision |
|---|---|---|---|
| C-0B7D2C30 | 2026-09-10 | 2026-09-15 | CB (multi-year, CZ untrusted) — in-window |
| C-0BCDB8C2 | 2027-09-18 | 2026-09-18 | CB (multi-year; CZ shows +1 yr) — **pulls renewal into window** |
| C-0D2AB865 | 2026-09-10 | 2026-09-22 | CB (multi-year) — in-window |
| C-0BBE3E60 | 2027-09-26 | 2026-09-26 | CB (multi-year; CZ shows +1 yr) — **pulls renewal into window** |
| C-0F5D2323 | 2026-09-10 | 2026-09-29 | CB (multi-year) — in-window |

The two 2027 ChurnZero dates are exactly +1 year off the Chargebee dates, the signature of the known multi-year defect. Without the CB override, both would have been silently excluded from this window (−$85,420 of exposure missed). All 15 non-multi-year accounts agree exactly between systems.

## Renewal table (sorted by date used)

Utilization = seats_used/seats. 3-mo trend = (Aug − Jun)/Jun from usage_12m. Rating rubric: HIGH = utilization <70% AND (12-mo decline or 3-mo drop ≥10%); MED = utilization <70% only; LOW = utilization ≥70%.

| Company | CSM | ARR | Date used | Util | 3-mo trend (Jun→Aug) | Rating |
|---|---|---|---|---|---|---|
| C-0BCDB8C2 | Cole Ingram | $54,427 | 2026-09-18 (CB) | 232/424 = 54.7% | 127→110 = −13.4% | HIGH |
| C-0BBE3E60 | Dana Mercer | $30,993 | 2026-09-26 (CB) | 74/114 = 64.9% | 39→33 = −15.4% | HIGH |
| C-0F5D2323 | Cole Ingram | $90,647 | 2026-09-29 (CB) | 111/390 = 28.5% | 20→18 = −10.0% | HIGH |
| C-0EC6999D | Elena Sinclair | $79,419 | 2026-10-03 | 31/112 = 27.7% | 17→15 = −11.8% | HIGH |
| C-0B20DB64 | Dana Mercer | $21,770 | 2026-10-07 | 214/378 = 56.6% | 294→294 = 0.0% | MED |
| C-0BBC4E7A | Cole Ingram | $56,374 | 2026-10-10 | 228/337 = 67.7% | 142→139 = −2.1% | HIGH |
| C-0FD551AB | Elena Sinclair | $48,815 | 2026-10-14 | 210/376 = 55.9% | 123→126 = +2.4% | MED |
| C-0F9F8F13 | Dana Mercer | $46,230 | 2026-10-18 | 199/352 = 56.5% | 185→182 = −1.6% | MED |
| C-0BC34584 | Cole Ingram | $16,740 | 2026-10-22 | 327/494 = 66.2% | 104→106 = +1.9% | MED |
| C-0B7A7546 | Elena Sinclair | $35,062 | 2026-10-25 | 182/205 = 88.8% | 64→63 = −1.6% | LOW |
| C-0B369871 | Dana Mercer | $85,128 | 2026-10-29 | 317/422 = 75.1% | 326→333 = +2.1% | LOW |
| C-0B144C78 | Cole Ingram | $30,899 | 2026-11-02 | 169/224 = 75.4% | 101→106 = +5.0% | LOW |
| C-0FC4DBB8 | Elena Sinclair | $94,732 | 2026-11-05 | 356/464 = 76.7% | 189→193 = +2.1% | LOW |
| C-0D5BBE3A | Dana Mercer | $39,740 | 2026-11-09 | 85/102 = 83.3% | 88→91 = +3.4% | LOW |
| C-0FB9D5AF | Cole Ingram | $63,158 | 2026-11-13 | 144/199 = 72.4% | 173→176 = +1.7% | LOW |
| C-0B344485 | Elena Sinclair | $64,384 | 2026-11-16 | 224/287 = 78.0% | 238→244 = +2.5% | LOW |
| C-0CB2C1B4 | Dana Mercer | $40,628 | 2026-11-20 | 386/473 = 81.6% | 47→49 = +4.3% | LOW |
| C-22170CA1 | Cole Ingram | $45,646 | 2026-11-24 | 251/294 = 85.4% | 143→146 = +2.1% | LOW |
| C-0B7D2C30 | Dana Mercer | $65,901 | 2026-09-15 (CB) | 274/476 = 57.6% | 97→84 = −13.4% | HIGH |
| C-0D2AB865 | Elena Sinclair | $38,022 | 2026-09-22 (CB) | 250/407 = 61.4% | 125→109 = −12.8% | HIGH |

One-sentence evidence per HIGH/MED account (12-mo usage change shown where it drives the rating):

- C-0BCDB8C2 (HIGH): 54.7% utilization on a 12-month usage slide of −45.0% (200→110), still falling −13.4% in the last quarter.
- C-0BBE3E60 (HIGH): 64.9% utilization with the steepest 3-month drop in the book (−15.4%) atop a −47.6% annual decline (63→33).
- C-0F5D2323 (HIGH): only 111 of 390 seats used (28.5%) — the largest underutilized spend in the window — with a −10.0% 3-month trend.
- C-0EC6999D (HIGH): 27.7% utilization (31/112) with a −11.8% 3-month drop; usage is flat year-over-year but collapsing from a tiny base.
- C-0B7D2C30 (HIGH): 57.6% utilization and usage down −45.8% over 12 months (155→84), −13.4% in the last quarter.
- C-0D2AB865 (HIGH): 61.4% utilization with usage down −45.2% over 12 months (199→109), −12.8% in the last quarter.
- C-0BBC4E7A (HIGH): utilization 67.7%, just under threshold, on a slow but persistent decline (−2.1% 3-mo, −2.1% 12-mo).
- C-0B20DB64 (MED): 56.6% utilization but usage is flat/stable (0.0% 3-mo, +0.3% 12-mo) — underutilization without erosion.
- C-0FD551AB (MED): 55.9% utilization with usage slightly up (+2.4% 3-mo, +1.6% 12-mo).
- C-0F9F8F13 (MED): 56.5% utilization, usage essentially flat (−1.6% 3-mo, 0.0% 12-mo).
- C-0BC34584 (MED): 66.2% utilization, usage mildly positive (+1.9% 3-mo, +2.9% 12-mo); smallest ARR in the window.

## Flagged discrepancies (5)

All five are the multi-year accounts; ChurnZero disagrees on every one and Chargebee is trusted per the stated data-quality rule. C-0B7D2C30 and C-0F5D2323 also have ChurnZero dates ~5–19 days earlier than Chargebee (2026-09-10 vs 09-15 and 09-29) — direction of error is not a year-shift, so if both were reliable the earliest plausible date would still be in-window; the CB dates are used.

## Bottom line

- Total ARR renewing in the 90-day window: **$1,048,715** across 20 accounts (ChurnZero alone would have shown 18 accounts / $963,295).
- ARR at risk (HIGH + MED): **$549,338** (52.4% of renewing ARR) — HIGH $415,783 (7 accounts), MED $133,555 (4 accounts). LOW: $499,377 (9 accounts).
- By CSM: Sinclair $360,434 renewing, Mercer $330,390, Ingram $357,891. Mercer and Ingram carry the densest HIGH-risk concentration around late September (four HIGH accounts between Sep 18–29, incl. the $90,647 C-0F5D2323 at 28.5% utilization).
- Caveats: usage data ends 2026-08 (one month stale for the September renewals); no product-tier, contract-end-date, or support-history fields were provided, so ratings rest on utilization and usage trend only.
ticket-theme-synthesisPASS 1.00
Plain-English question: What are the real themes in this quarter's support tickets, ignoring the existing tags?
Prompt — what the model saw
Synthesize the quarter's support tickets into 4 to 7 themes. Do not trust existing tags; read the text. Per theme: name, count, share, distinct accounts, ARR affected, two ticket ids, one-line recommendation. Separate broad patterns from single-account noise and rank by ARR exposure, not volume.
cost $0.3780 · 153s · in 15,382 / out 43,098 tok
✓ theme_count_range✓ counts_verifiable✓ single_account_flagged✓ arr_ranked✓ ticket_ids

✓ Correct Answer (ground truth)

Plain-English answer: 80 tickets in 4-7 themes — points-not-posting leads (20 tickets, 25%, $31,100 ARR affected) — each with counts, distinct accounts, ARR, and a recommendation.
The answer in one line: total_tickets = 80; quarter_window = 2026-06-01 through 2026-08-31
total_tickets80
noteexisting tags are noisy/unreliable; classify by text
quarter_window2026-06-01 through 2026-08-31
ranking_by_arr_exposurehris_provisioning, redemption_failures, billing_dispute, points_not_posting, slack_sync
sample_ticket_ids80 items — e.g. IC-460004, IC-460016, IC-460001, IC-460041
single_account_noise_themebilling_dispute
themes{"points_not_posting": {"count": 20, "share_pct": 25.0, "distinct_accounts": 9, "arr_affected": "31100.00", "example_ticket_ids": ["IC-460001", "IC-460002"], "single_account": false}, "slack_sync": {"
graded checkstheme_count_range · counts_verifiable · single_account_flagged · arr_ranked · ticket_ids

Model's actual answer

show the model's full answer
Method: 80 tickets, 24 distinct accounts, portfolio ARR = $284,800 (sum of one ARR per account). Themes assigned from body text only; existing tags disregarded (e.g. IC-460020 "missing points" is tagged billing, IC-460071 "seat-count error" is tagged feedback — both wrong). Every theme's accounts are disjoint, so ARR shares sum to 100%.

Ranked by ARR affected (not volume):

1. HRIS provisioning silently failing — 12 tickets (12/80 = 15.0%), 3 distinct accounts, ARR $114,000 (114,000/284,800 = 40.0%). Accounts: C-0B2213A9, C-0DDFC9A7, C-0F6C0F34. Tickets: IC-460060, IC-460062. Broad pattern. Recommendation: treat as a Sev-1 — new hires silently skipped with "no errors" in the provisioning log means add failure telemetry and manually provision affected accounts now; this is the top ARR exposure despite the lowest ticket count.

2. Gift-card redemption failures (checkout spin, failed delivery, points deducted anyway) — 18 tickets (22.5%), 7 distinct accounts, ARR $68,800 (24.2%). Accounts: C-0B0F1BAB, C-0B827671, C-0CEF69FD, C-0D9CA315, C-0F876796, C-0FCCD2DF, C-14264ABD. Tickets: IC-460025, IC-460038. Broad pattern. Recommendation: make point deduction atomic with gift-card delivery (deduct only on confirmed fulfillment) and audit/refund the deducted-but-undelivered cases.

3. Invoice seat-count and renewal-tier errors — 16 tickets (20.0%), 1 distinct account, ARR $52,000 (18.3%). Account: C-0E9C27D1 (largest account in the file). Tickets: IC-460069, IC-460078. Single-account noise, not a market pattern: 16 tickets all from C-0E9C27D1 describing one unresolved billing dispute (200 seats charged vs 150 licensed, wrong tier on renewal). Recommendation: escalate as an account-level billing correction with credits and an executive touch — this is a renewal-risk escalation on the biggest account, not a product defect wave.

4. Points not posting to balances (individual "my balance hasn't updated" + team-wide "whole team after the weekend") — 20 tickets (25.0%), 9 distinct accounts, ARR $31,100 (10.9%). Accounts: C-0B2895EF, C-0BE96399, C-0BF20542, C-0D0B047C, C-0D284E42, C-0D3278C7, C-0D6CC8E3, C-0DD0626C, C-21FEBCBB. Tickets: IC-460016 (team-wide variant), IC-460004 (individual variant). Broad pattern — highest ticket volume, lowest-but-one ARR. Three accounts (C-0BF20542, C-0D284E42, C-0D6CC8E3) filed both variants, suggesting one root cause in the points pipeline, likely weekend-batch related. Recommendation: investigate the weekend processing window for ledger drops and backfill the 9 affected balances.

5. Slack integration breakdowns (sync stops, re-auth not sticking, slash-command errors) — 14 tickets (17.5%), 4 distinct accounts, ARR $18,900 (6.6%). Accounts: C-0B843542, C-0BA71F12, C-10A56B0F, C-8C2E8F00. Tickets: IC-460041, IC-460046. Broad pattern. Recommendation: fix the token-refresh/re-auth persistence and slash-command endpoint, then notify the 4 affected accounts proactively.

Broad patterns (multi-account): themes 1, 2, 4, 5 — 64 tickets across 23 accounts, $232,800 ARR.
Single-account noise: theme 3 — 16 tickets from one account, $52,000 ARR.

Cross-check: ticket counts 12+18+16+20+14 = 80 (all tickets classified, none dropped); ARR 114,000+68,800+52,000+31,100+18,900 = 284,800 = 100% of the portfolio in this file; accounts 3+7+1+9+4 = 24. Note the volume-vs-ARR inversion: theme 4 leads on volume (25%) but is 4th on ARR exposure, while theme 1 is last on volume (15%) and first on ARR (40%).

⚠️ File-mutation verifier: 1 file edit(s) FAILED this turn despite any wording above that may suggest otherwise. Run `git status` or `read_file` to confirm what actually landed.
  • `/Users/amaniphipps/.hermes/cache/scratch/intercom_tickets.csv` — [write_file] Refusing to overwrite `/Users/amaniphipps/.hermes/cache/scratch/intercom_tickets.csv`: `/Users/amaniphipps/.hermes/cache/scratch/intercom_tickets.csv` exists but this task has not see…
marketing 5 tests
case-study-matchPASS 1.00
Plain-English question: Which three existing customers are the best case-study matches for this prospect?
Prompt — what the model saw
Given the prospect, rank the three most similar existing customers with public case studies for use as social proof. Similarity on industry, size band, use case, region. Cite the fields that drove each match. Do not name any customer without a case study and never include billing data or contact names.
cost $0.0345 · 18s · in 1,447 / out 3,519 tok
✓ top3✓ decoys_excluded✓ cites_fields✓ no_billing

✓ Correct Answer (ground truth)

Plain-English answer: Three ranked matches on industry, size, use case, and region (e.g. C-11C31562), with the higher-scoring decoy explicitly excluded for a mismatched field.
The answer in one line: prospect_alias = C-82AF3719
constraintsno_billing_data, no_contact_names, only_customers_with_has_case_study_true_named
decoys_excludedcompany_alias=C-D6217CAA; similarity_score=4; reason=no_case_study | company_alias=C-C153868F; similarity_score=4; reason=no_case_study
prospect_aliasC-82AF3719
prospect_fields_usedindustry, size_band, use_case, region
top3company_alias=C-11C31562; similarity_score=3 | company_alias=C-64171065; similarity_score=3 | company_alias=C-A13C193D; similarity_score=3
graded checkstop3 · decoys_excluded · cites_fields · no_billing

Model's actual answer

show the model's full answer
**Screening step:** Of 25 customers, 8 have has_case_study=true and are eligible to be named: C-FEFCC020, C-11209339, C-9D69D147, C-11C31562, C-CD4829A7, C-0CE315B2, C-64171065, C-A13C193D. The other 17 are excluded from naming regardless of similarity.

**Prospect profile:** C-82AF3719 — Technology, Mid-Market, employee_recognition, NA-West.

**Scoring:** 1 point per matched field across the 4 provided dimensions (max 4). Only the 8 eligible customers scored:

```
C-64171065  Technology✓  Mid-Market✓  employee_recognition✓  NA-East✗   = 3/4
C-11C31562  Manufacturing✗  Mid-Market✓  employee_recognition✓  NA-West✓  = 3/4
C-A13C193D  Technology✓  Mid-Market✓  retention✗  NA-West✓             = 3/4
C-CD4829A7  Technology✓  Mid-Market✓  retention✗  NA-East✗             = 2/4
C-9D69D147  Financial Services✗  Mid-Market✓  retention✗  NA-West✓    = 2/4
C-FEFCC020  Education✗  Enterprise✗  wellness✗  NA-East✗              = 0/4
C-11209339  Manufacturing✗  SMB✗  peer_milestones✗  APAC✗            = 0/4
C-0CE315B2  Education✗  SMB✗  wellness✗  EMEA✗                     = 0/4
```

**Tie-break (three-way tie at 3/4):** The data defines no weights, so I applied a stated priority: use_case > industry > size_band > region, on the rationale that social proof is strongest when the story mirrors the prospect's deployment (use case), then reads as a peer (industry), with region weakest for software proof. Weighted check (use_case 3, industry 2, size 2, region 1) gives 7 / 6 / 5 / 4 / 3 — same order, so the ranking is robust under either method.

**Ranked matches:**

1. **C-64171065** — matches industry (Technology), size_band (Mid-Market), use_case (employee_recognition). Only miss: region (NA-East vs NA-West).
2. **C-11C31562** — matches use_case (employee_recognition), size_band (Mid-Market), region (NA-West). Only miss: industry (Manufacturing vs Technology). Ranks below #1 on the tie-break because it lacks the industry match while #1 lacks only region, the lowest-weighted field.
3. **C-A13C193D** — matches industry (Technology), size_band (Mid-Market), region (NA-West). Only miss: use_case (retention vs employee_recognition).

**Explicit limitations:** Similarity is computed only from the four fields provided (industry, size_band, use_case, region). No revenue, seat counts, tenure, or outcome metrics exist in the data, so depth of match within size band and case-study content could not be assessed. No customer without a case study is named, and no billing or contact data is included (none present in the source).
channel-attribution-mathPASS 1.00
Plain-English question: Compute cost per SQM, cost per SQO, and pipeline per dollar for every channel over six months.
Prompt — what the model saw
Compute channel performance for the trailing 6 months. Paid: spend, SQMs, SQOs, cost per SQM, cost per SQO, SQM-to-SQO rate, pipeline amount, pipeline per dollar. Organic: volume, SQO rate, pipeline. Report a channel with spend and zero SQMs as undefined, not zero. Flag rows where SQO date precedes SQM date. Finish with a reallocation recommendation and your confidence given sample sizes.
cost $0.1527 · 103s · in 9,391 / out 15,951 tok
✓ undefined_channel✓ undefined_not_zero✓ flagged_rows✓ best_channel✓ reallocation

✓ Correct Answer (ground truth)

Plain-English answer: LinkedIn Ads at $960/SQM and $3,000/SQO, organic search at $90,000 pipeline on zero spend, paid social flagged at $18,000 with zero SQMs, and a channel winner named.
The answer in one line: best_pipeline_per_dollar_channel = paid_search
best_pipeline_per_dollar_channelpaid_search
confidence_notewebinars and referral have small SQM samples (<20); treat rates as directional.
flagged_rows_sqo_before_sqmCT-000041, CT-000044
months2026-03, 2026-04, 2026-05, 2026-06, 2026-07, 2026-08
per_channel{"linkedin_ads": {"spend_usd": 24000, "sqms": 25, "sqos": 8, "pipeline_amount": 96000, "cost_per_sqm": 960.0, "cost_per_sqo": 3000.0, "sqm_to_sqo_rate": 0.32, "pipeline_per_dollar": 4.0}, "organic_sea
reallocation_recommendationfrom=paid_social; to=paid_search; rationale=paid_social has spend but zero SQMs (cost per SQM undefined, not zero); paid_search has the highest pipeline per dollar.
undefined_channelspaid_social
graded checksundefined_channel · undefined_not_zero · flagged_rows · best_channel · reallocation

Model's actual answer

show the model's full answer
Window: 2026-03 through 2026-08 (6 months). 122 contacts, all with SQM dates inside the window.

**PAID CHANNELS**

| Channel | Spend | SQMs | SQOs | $/SQM | $/SQO | SQM→SQO rate | Pipeline | Pipe/$ |
|---|---|---|---|---|---|---|---|---|
| paid_search | $36,000 (6 × $6,000) | 40 | 18 | 36,000/40 = $900 | 36,000/18 = $2,000 | 18/40 = 45.0% | $720,000 (18 × $40,000) | 720,000/36,000 = 20.0× |
| linkedin_ads | $24,000 (6 × $4,000) | 25 | 8 | 24,000/25 = $960 | 24,000/8 = $3,000 | 8/25 = 32.0% | $96,000 (8 × $12,000) | 96,000/24,000 = 4.0× |
| paid_social | $18,000 (6 × $3,000) | 0 | 0 | undefined (spend, zero SQMs — not zero cost) | undefined | undefined | $0 | undefined |
| webinars | $9,000 (6 × $1,500) | 12 | 5 | 9,000/12 = $750 | 9,000/5 = $1,800 | 5/12 = 41.7% | $60,000 (5 × $12,000) | 60,000/9,000 = 6.67× |

**ORGANIC CHANNELS** (no spend recorded in channel_spend.csv)

| Channel | Volume | SQO rate | Pipeline |
|---|---|---|---|
| organic_search | 30 SQMs | 10/30 = 33.3% | $90,000 (10 × $9,000) |
| referral | 15 SQMs | 6/15 = 40.0% | $48,000 (6 × $8,000) |

**DATA QUALITY FLAGS — SQO date precedes SQM date (2 rows, both linkedin_ads):**
- CT-000044: SQM 2026-07-23, SQO 2026-07-18, $12,000
- CT-000041: SQM 2026-06-14, SQO 2026-06-09, $12,000

These are implausible sequences (likely timestamp/backfill errors), so linkedin_ads' headline numbers include questionable rows. Excluding both: 6/25 = 24.0% SQO rate, $24,000/6 = $4,000 per SQO, $72,000 pipeline, 3.0× pipe/$.

**REALLOCATION RECOMMENDATION**

Shift budget from linkedin_ads (and the stalled paid_social) toward paid_search and webinars:

1. paid_social: $18,000 with zero SQMs in 6 months — stop or fully redeploy. Even at paid_search's $900/SQM, that's ~20 SQMs; paid_search wins the reallocation on every measured metric.
2. linkedin_ads: weakest paid efficiency ($3,000/SQO, 4.0× pipe/$ — worse excluding the 2 flagged rows: $4,000/SQO, 3.0×). Trim ~half and move to paid_search.
3. webinars is the efficiency leader among paid channels ($750/SQM, $1,800/SQO, 6.67×) but volume-capped at 12 SQMs; a modest increase is justified if capacity scales, else it's the priority for any incremental budget after paid_search.

Suggested shift: $9,000 from paid_social + $12,000 from linkedin_ads → paid_search (roughly +10 SQMs expected at $900/SQM), keeping webinars flat pending capacity.

**CONFIDENCE (sample sizes are small)**

95% CIs on SQO rates overlap substantially: paid_search 45.0% (29.6–60.4, n=40), organic_search 33.3% (16.5–50.2, n=30), referral 40.0% (15.2–64.8, n=15), webinars 41.7% (13.8–69.6, n=12), linkedin_ads 32.0% (13.7–50.3, n=25). Moderate confidence in the paid_social cutoff (zero SQMs is unambiguous) and in paid_search's pipeline advantage (18 SQOs, $40,000 each — no variance in deal size, so the 20× vs 4× gap is driven by rate, not deal mix). Low confidence in fine-grained rate ranking among the other channels — at n=12–25 per group, differences of 8–10 points are within noise. Also, per-deal pipeline values are constant within each channel ($40k/$12k/$9k/$8k), which suggests synthetic or bucketed data; treat pipeline-per-dollar comparisons as directional. Recommend re-evaluating after ~1 more quarter or ~30 more SQMs per channel.
competitor-battlecard-updatePASS 1.00
Plain-English question: Update the Rivally battlecard with sourced pricing and where each side wins.
Prompt — what the model saw
Update the battlecard for the competitor. Sections: one-line positioning, pricing with source and date (newer source wins, note the conflict), where they win, where we win, objections and responses, recent changes, our 12-month win/loss record against them. Cite a snippet id for every factual claim. Rep opinion on a call is not a fact about the competitor. Anything from the old card you cannot re-source gets marked unverified.
cost $0.0925 · 52s · in 2,098 / out 11,045 tok
✓ competitor_named✓ pricing✓ win_loss✓ unverified_marked✓ citations

✓ Correct Answer (ground truth)

Plain-English answer: Rivally at $7 per user per month (the August source wins over the stale $5, conflict noted), a 13-7 win/loss record, and two unverified old claims flagged.
The answer in one line: competitor = Rivally
citation_ids_available25 items — e.g. S01, S02, S03, S04
competitorRivally
pricingvalue_per_user_per_month=7; source_snippet=S17; source_date=2026-08-12
rep_opinion_snippet_ids_not_factsS09, S21
unverified_old_claimsRivally lacks a Slack integration., Rivally was acquired by WorkHuman in 2025.
win_loss_12mowins=13; losses=7
graded checkscompetitor_named · pricing · win_loss · unverified_marked · citations

Model's actual answer

show the model's full answer
BATTLECARD: RIVALLY — updated 2026-10-19
Sources: competitor_snippets.csv (cited as Sxx), deals_with_competitor.csv (cited as deals file). Excluded as rep opinion, not competitor fact: S09 (AE thinks UI clunky), S21 (AE thinks discounting aggressive; snippet itself says not confirmed).

1. ONE-LINE POSITIONING
Points-based peer recognition with an engaging feed (S02, S16), easiest for mid-market (S04) and strongest for distributed EU teams (S12, S15), expanding into lightweight engagement surveys (S06, S23); reviewers consistently flag basic analytics and lagging admin tooling (S02, S07, S10, S16, S24).

2. PRICING
- Current list: Recognition Starter $7 per user/month, annual billing required — pricing page 2026-08-12 (S17).
- Conflict note: earlier pricing pages showed $5 (2026-01-20, S03; still $5 on 2026-04-01, S08). Newer source wins: quote $7; the $5 figures are correct for their dates only, superseded.
- Deal-reported datapoints (not list sources): $6.50/user/mo quoted to a 500-seat prospect, annual term, 2026-06-02 (S13) — predates the price change and exceeds the then-listed $5 (S08); tier not stated in data, unexplained. Prospect-reported $7 list with 15% discount offered for a 3-year term, 2026-08-14 (S18) — consistent with S17; single deal, treat as anecdote.
- Rivally Pulse: sold as a paid add-on, not bundled; no price in data (S23).
- No Bonusly pricing exists in this dataset; no comparative price claim is supportable.

3. WHERE THEY WIN
- Engagement: points-based recognition feed repeatedly praised (S02, S16).
- EU: EU data residency GA + Dublin office (S15); residency pitched in deals as early as Feb 2026 (S05); ex-Workday VP EMEA hired to lead expansion (S11); multi-language support praised, "strong for distributed EU teams" (S12).
- Mid-market onboarding: setup under a week, Slack integration out of the box (S04).
- Support: response time under 4 hours praised (S22).
- Product cadence: Pulse launched 2026-03-05 (S06), GA as paid add-on 2026-09-01 (S23); Teams app v2 public preview (S19).
- Pricing flexibility: 15% off for a 3-year term, one documented deal (S18).

4. WHERE WE WIN
- Analytics depth: Rivally analytics "limited" (S02), dashboards "basic compared to enterprise tools" (S07), admin tooling lags peers (S16). Documented win: 800-seat prospect chose Bonusly over Rivally citing analytics depth (S25; snippet dated 2026-09-03, not mappable to any deal_alias in the deals file).
- Enterprise IT fit: no SCIM provisioning, manual user management painful (S10); no bulk recognition editing (S24).
- Data portability: analytics exports are CSV-only; one reviewer found migrating off Rivally hard (S20).
- EMEA rewards: Rivally's EMEA catalog is thinner than its own US catalog (S14). The data does not say our catalog is deeper — use as a discovery question, not a claimed advantage.

5. OBJECTIONS AND RESPONSES
- "Rivally is cheaper." Their advertised Starter price went $5 → $7 in Aug 2026 (S03, S08 → S17); only documented discount is 15% tied to a 3-year commitment (S18). Verify current Bonusly pricing before any comparative claim (not in dataset).
- "Rivally has EU data residency." True per data (S05, S15). Nothing in the dataset covers Bonusly's residency posture — internal gap; do not concede or counterclaim without product confirmation.
- "We like Rivally's feed." Reviewers agree (S02, S16); pivot to admin/analytics limits (S07, S10, S16, S24, S20) and the 800-seat win on analytics depth (S25).
- "Rivally support is fast." (S22) No comparative support data for us — gap.
- Rivally's counter-pitch against Bonusly: not recorded anywhere in the dataset; no guidance possible.

6. RECENT CHANGES (newest first)
- 2026-09-01: Pulse exits beta; sold as paid add-on, not bundled (S23).
- 2026-08-20: Microsoft Teams app v2 in public preview (S19).
- 2026-08-12: Recognition Starter $5 → $7/user/mo (S17).
- 2026-07-01: Dublin office opens; EU data residency GA (S15).
- 2026-05-09: ex-Workday VP EMEA hired for European expansion (S11).

7. WIN/LOSS RECORD — TRAILING 12 MONTHS (2025-09 through 2026-08; all 20 deal records fall inside this window; no records after 2026-08)
- 2025-09: 1W-1L (W: Deal-072E31 / L: Deal-7767F5)
- 2025-10: 2W-0L (Deal-A9FD43, Deal-F65C8F)
- 2025-11: 1W-1L (W: Deal-7AA785 / L: Deal-D263E0)
- 2025-12: 1W-1L (W: Deal-44C524 / L: Deal-935746)
- 2026-01: 2W-0L (Deal-0D0CD6, Deal-E46EAB)
- 2026-02: 2W-0L (Deal-D5B790, Deal-1D2392)
- 2026-03: 1W-1L (W: Deal-5C636E / L: Deal-9066A6)
- 2026-04: 0W-2L (Deal-5645A5, Deal-72A02F)
- 2026-05: 0W-1L (Deal-C6FFAA)
- 2026-06: 1W-0L (Deal-67BE14)
- 2026-07: 1W-0L (Deal-1B6969)
- 2026-08: 1W-0L (Deal-F03E7B)

Arithmetic: wins = 1+2+1+1+2+2+1+0+0+1+1+1 = 13; losses = 1+0+1+1+0+0+1+2+1+0+0+0 = 7; total = 20. Win rate = 13/20 = 65%.
Pattern (observed, no cause data attached to deal records): worst stretch Mar–May 2026 at 1W-4L (Deal-9066A6, Deal-5645A5, Deal-72A02F, Deal-C6FFAA); Jun–Aug 2026 recovered 3W-0L. The deals file carries no loss reasons; the only sourced loss reason is S25 (analytics depth), which cannot be tied to any listed deal_alias.

8. OLD-CARD CLAIMS — STATUS
- "Points-based recognition for mid-market" — re-sourced (S02, S04); retained.
- "$5/user/mo as of 2026-01" — re-sourced (S03) but superseded by S17; see Pricing.
- "Rivally lacks a Slack integration" — CONTRADICTED, not just unverified: S04 (2026-02-02) reports Slack integration worked out of the box. Removed.
- "Acquired by WorkHuman in 2025" — UNVERIFIED; no snippet references any acquisition. Closest fact is an ex-Workday VP hire (S11), which is not an acquisition. Do not use.
- "Strong in EU enterprise with multi-language support" — re-sourced (S12; supported by S11, S15).

DATA GAPS: no Bonusly-side facts (pricing, EU residency, support SLAs) in the dataset; no loss reasons attached to deal records; S25 unmapped to a deal_alias; S13's $6.50 quote unexplained against the then-current $5 list.
nurture-sequence-diagnosisPASS 1.00
Plain-English question: What is broken in each email sequence — rates, overlaps, tracking errors?
Prompt — what the model saw
Review the sequences. Per sequence: sent, open rate, reply rate, meeting rate, weakest step. Identify tracking errors (opened above sent), audience overlap between sequences, and the failure mode for anything under 2 percent reply. One change per weak sequence and which to fix first. Under 350 words.
cost $0.2193 · 64s · in 26,362 / out 18,874 tok
✓ tracking_error✓ overlap_pair✓ weakest_step✓ fix_priority

✓ Correct Answer (ground truth)

Plain-English answer: Per-sequence open, reply, and meeting rates with the weakest step (all three weakest at step 3), an opened-above-sent tracking error, and audience overlap flagged.
The answer in one line: fix_priority_first = Cold Outbound - HR Leaders
all_pairwise_overlaps['Cold Outbound - HR Leaders', 'Cold Outbound - People Ops'], ['Expansion Nurture', 'New Logo Nurture']
audience_overlap_pairCold Outbound - HR Leaders, Cold Outbound - People Ops
failure_modeCold Outbound - HR Leaders=opens healthy (~37%) but replies near zero from step 1 -> body/CTA not landing, not a deliverability or subject-line problem
fix_priority_firstCold Outbound - HR Leaders
per_sequence{"New Logo Nurture": {"sent": 1386, "opened": 490, "replied": 90, "meetings": 27, "open_rate": 0.3535, "reply_rate": 0.0649, "meeting_rate": 0.0195, "weakest_step": 3}, "Expansion Nurture": {"sent": 8
tracking_errorsequence=Expansion Nurture; step=2; sent=300; opened=340; issue=opened exceeds sent
under_2pct_replyCold Outbound - HR Leaders
graded checkstracking_error · overlap_pair · weakest_step · fix_priority

Model's actual answer

show the model's full answer
**Per sequence (rates = summed steps ÷ summed sent)**

| Sequence | Sent | Open | Reply | Meetings |
|---|---|---|---|---|
| New Logo Nurture | 1,386 | 35.4% (490/1,386) | 6.49% (90/1,386) | 1.95% (27) |
| Expansion Nurture | 875 | 64.6% (565/875) | 6.74% (59/875) | 1.37% (12) |
| Cold Outbound - HR Leaders | 1,785 | 30.5% (545/1,785) | 0.45% (8/1,785) | 0.00% (0) |
| Cold Outbound - People Ops | 1,163 | 29.2% (340/1,163) | 2.49% (29/1,163) | 0.52% (6) |

Weakest step by reply rate: New Logo step 3 (18/428 = 4.21%); Expansion step 3 (12/275 = 4.36%); HR Leaders step 3 (1/590 = 0.17%); People Ops step 3 (6/377 = 1.59%).

**Tracking errors**
- Expansion Nurture step 2: opened 340 > sent 300 (113.3%) — open tracking is broken; its 64.6% open rate is not trustworthy.
- Expansion sent shrinks step 1→3 (300→300→275) while New Logo falls 500→458→428: ~9–14% per-step attrition unexplained by the data.

**Audience overlap (audiences.csv, 963 rows)**
- 23 contacts appear in two sequences: 21 shared between Cold Outbound - HR Leaders and Cold Outbound - People Ops (e.g. CT-000849, CT-000884, CT-001311, CT-001285, CT-001255), and 2 shared between Expansion Nurture and New Logo Nurture (CT-000301, CT-000624). The HR/People Ops split is largely the same audience in practice.

**Failure mode, under 2% reply**
- Cold Outbound - HR Leaders (0.45%): volume-heavy blast, near-zero replies and 0 meetings from 1,785 sends — open counts (40% step 1) but no engagement means the list is unqualified, not just the copy. Failure mode = wrong audience/ICP mismatch.

**One change each**
- HR Leaders: rebuild the list (pause sending, retarget). Fix first — biggest volume, worst yield.
- People Ops: rework step 3 (1.59%, decay 3.5→2.3→1.6).
- New Logo: refresh step 3 offer (4.21%).
- Expansion: fix open tracking; no content change until data is clean.
weekly-marketing-goals-rollupPASS 1.00
Plain-English question: Are we on pace on the quarter's marketing goals?
Prompt — what the model saw
Produce the weekly marketing goals update for the active quarter: SQMs, SQOs, DS2s, closed-lost MIA rate, same-quarter close count, active pipeline coverage against target. For each: QTD actual, target, delta, pace (on/behind/ahead given days elapsed). One paragraph on what moved this week.
cost $0.0369 · 22s · in 1,045 / out 3,980 tok
✓ metrics_present✓ deltas✓ coverage_ratio✓ pace✓ paragraph

✓ Correct Answer (ground truth)

Plain-English answer: SQMs ahead of pace, SQOs on pace, DS2s behind (35 short of expected at day 66 of 92), with deltas and same-quarter closes.
The answer in one line: days_in_quarter = 92; days_elapsed = 66; coverage_ratio_active_pipeline = 0.75; elapsed_fraction = 0.7174 (+1 more below)
days_in_quarter92
coverage_ratio_active_pipeline0.75
days_elapsed66
elapsed_fraction0.7174
metrics{"SQMs": {"qtd_actual": 230, "target": 300, "delta": -70, "expected_at_day_elapsed": 215.22, "pace": "ahead"}, "SQOs": {"qtd_actual": 84, "target": 120, "delta": -36, "expected_at_day_elapsed": 86.09,
pace_ruleexpected = target * days_elapsed/days_in_quarter; within +/-5% of expected = on, above = ahead, below = behind; lower_better metrics invert.
this_week_paragraph_required1
graded checksmetrics_present · deltas · coverage_ratio · pace · paragraph

Model's actual answer

show the model's full answer
Q3-2026 MARKETING GOALS UPDATE — through day 66 of 92 (71.7% of quarter elapsed; 26 days remain)

| Metric | QTD actual | Target | Delta | % of target | Pace (vs 71.7% elapsed) |
|---|---|---|---|---|---|
| SQMs | 230 | 300 | -70 | 76.7% | Ahead (pro-rated target 215.2; +14.8) |
| SQOs | 84 | 120 | -36 | 70.0% | Behind, marginal (pro-rated 86.1; -2.1) |
| DS2s | 40 | 75 | -35 | 53.3% | Behind (pro-rated 53.8; -13.8) |
| Closed-lost MIA rate | 20.0% (5/25) | <=10.0% (lower is better) | +10.0 pp over ceiling | 2.0x target | Off track (rate metric; not pro-rated) |
| Same-quarter closes | 10 | 20 | -10 | 50.0% | Behind (pro-rated 14.3; -4.3) |
| Active pipeline | $3.00M | $4.00M | -$1.00M | 75.0% | Ahead (pro-rated $2.87M; +$0.13M) |

Arithmetic: days-elapsed share = 66/92 = 0.7174; pro-rated target = full target x 0.7174. MIA rate = 5/25 = 0.20 vs 0.10 ceiling. Attainment = actual/target.

Gap to close over final 26 days (to hit full target): SQMs 70 (2.7/day), SQOs 36 (1.4/day), DS2s 35 (1.3/day), same-quarter closes 10 (0.4/day), pipeline +$1.0M ($38.5k/day). MIA: landing at exactly 10% requires 25 more closed-lost with zero additional MIA losses (5/50 = 0.10).

What moved this week: The provided files contain only a single QTD snapshot per metric — no prior-week figures — so week-over-week movement cannot be computed without inventing numbers. What the QTD composition shows at day 66: SQO-to-SQM conversion is 36.5% (84/230), DS2s stand at 47.6% of SQOs (40/84), and the 25 closed-lost carry a 20% MIA share (5/25) against the 10% ceiling, making MIA the clearest off-track metric. The load-bearing shortfalls are DS2s (-13.8 vs pro-pace) and same-quarter closes (-4.3), with SQOs only marginally behind (-2.1); SQMs (+14.8) and active pipeline (+$130k) are the only two metrics ahead of elapsed-time pace with 26 days left.
communication 5 tests
ceo-slack-compressionPASS 1.00
Plain-English question: Compress the forecast analysis into one Slack message the CEO can read in ten seconds.
Prompt — what the model saw
Compress the forecast analysis below into a single Slack message for the CEO. Rules: at most three sentences; under 80 words; lead with the decision or the so-what; keep every distinct number that matters and the one caveat that changes the conclusion; cut any sentence that does not carry a claim, a number, or a directive. No em dashes, no throat-clearing phrases.
cost $0.0264 · 14s · in 1,344 / out 2,477 tok
✓ max_3_sentences✓ under_80_words✓ weighted_present✓ caveat_present✓ no_em_dash

✓ Correct Answer (ground truth)

Plain-English answer: Three sentences, under 80 words, decision first, every distinct number kept plus the one caveat — all checked by deterministic text rules.
The answer in one line: weighted_number = 115976; must_keep = data-quality caveat — deleting the bad news fails
weighted_number115,976
limitmax 3 sentences, max 80 words, no em-dashes
must_keepdata-quality caveat — deleting the bad news fails
graded checksmax_3_sentences · under_80_words · weighted_present · caveat_present · no_em_dash

Model's actual answer

show the model's full answer
Book 115,976.75 for Q3 (44,729 COMMIT plus 35% of 203,565 BEST_CASE) but treat it as soft: 32 deals worth 227,575 already sit past Sep 30, including COMMIT Deal-D348E1 at 13,770. All 7 in-quarter COMMIT deals have empty why-buys, so the committed 44,729 has no documented reason to buy behind it. Fix owner blank on 85 of 86 and re-qualify the pushed October dates before trusting this number.
stalled-deal-followup-emailPASS 1.00
Plain-English question: Write a follow-up email for this stalled deal — under 120 words, one prior touch, one ask.
Prompt — what the model saw
Write a follow-up email for this stalled deal. Under 120 words. Reference one specific prior touch from the engagement history, make one clear ask, and match the tone of the prior thread. No invented details.
cost $0.0169 · 10s · in 1,153 / out 1,280 tok
✓ word_cap✓ references_touch✓ single_ask

✓ Correct Answer (ground truth)

Plain-English answer: An email that references the July 28 milestone-awards demo, makes one clear ask, and invents nothing — Deal-0D2F7A, 31 days since last engagement.
The answer in one line: amount = 5100; word_cap = 120; days_since_last_engagement = 31; deal_alias = Deal-0D2F7A (+4 more below)
amount5,100
days_since_last_engagement31
word_cap120
allowed_facts14 items — e.g. Deal-0D2F7A, $5,100, DS3, Alex Franklin
allowed_newENG-00001, ENG-00002, ENG-00003, ENG-00004, ENG-00005
deal_aliasDeal-0D2F7A
must_referenceengagement_id=ENG-00002; date=2026-07-28; type=meeting; why=most substantive prior touch; concrete prospect reactions to milestone awards and points catalog
ownerAlex Franklin
single_cta1
stageDS3
tonewarm, concise, professional (matches prior thread)
graded checksword_cap · references_touch · single_ask

Model's actual answer

show the model's full answer
Subject: Re: 150-seat pricing

Hi [Name],

Following up on the note I sent August 5 with pricing for 150 seats after the July 28 demo. Your team's reaction to the automated milestone awards and the points catalog was strong, and I know recognition budget sits with you, so I wanted to make sure that email didn't get buried.

One ask: does the 150-seat pricing work for your budget? A quick yes or no reply is all I need. If it's not the right fit yet, that's useful to know too.

Thanks,
Alex Franklin

(Word count: 88. Referenced touch: the 2026-08-05 pricing email, itself a recap of the 2026-07-28 demo. Prospect name is not in the data, so the greeting uses a placeholder.)
funnel-digest-paragraphPASS 1.00
Plain-English question: Write the weekly GTM digest — one paragraph per team, every number traceable.
Prompt — what the model saw
Write the weekly GTM digest: one paragraph per team (Marketing, Sales, CS, PLG) in Bonusly brand voice. Every number must trace to the extracts. No customer contact names, no employee performance commentary, no compensation. Four paragraphs, under 400 words total.
cost $0.0286 · 28s · in 1,017 / out 2,873 tok
✓ four_paragraphs✓ word_cap✓ teams

✓ Correct Answer (ground truth)

Plain-English answer: Four paragraphs (Marketing, Sales, CS, PLG) using only the allowed numbers, no contact names, no performance commentary, under 400 words.
The answer in one line: word_cap = 400
word_cap400
allowed_numbers16 items — e.g. 2, 3, 9, 08
bannedcustomer contact names, employee performance commentary, compensation
required_paragraphsMarketing, Sales, CS, PLG
graded checksfour_paragraphs · word_cap · teams

Model's actual answer

show the model's full answer
Weekly GTM digest, week of 2026-08-31

Marketing finished at 46 SQMs against a target of 52, 6 short of plan (46/52 = 88.5% attainment). Webinars delivered 18 of those SQMs, 39.1% of the total (18/46). Paid search cost per SQM held at $150. The extracts carry no other channel splits or prior-week figures, so channel performance beyond webinars and paid search cost can't be assessed this week.

Sales converted 14 SQOs and set 9 DS2 meetings. New pipeline created was $310,000, which works out to about $22,143 per SQO (310000/14). Three deals closed in the same quarter during the week. The extracts include no stage-conversion rates or prior-week pipeline, so movement beyond these counts can't be measured.

Customer success saved 2 renewals this week. Team NPS moved to 61. Three open red-flag accounts head into next week. The extracts don't identify those accounts or the risk drivers, and they include no contact names, so none are cited here.

PLG added 412 new signups with activation at 31%, roughly 128 activated signups (412 × 0.31 = 127.7). 38 companies reached the aha moment of 10 recognition gives. The extracts provide no week-over-week signup or activation trend, so growth rate can't be calculated.
partner-digest-sectionPASS 1.00
Plain-English question: Write the partner digest so every partner in the roster appears exactly once.
Prompt — what the model saw
Write the partner digest for the period. Every partner in the roster appears once: active plays get status with numbers, quiet partners get one line. Partner-sourced pipeline count and amount must match the deal data. Nothing omitted, nothing added.
cost $0.0144 · 8s · in 1,204 / out 920 tok
✓ all_partners✓ pipeline_amount✓ quiet_one_line

✓ Correct Answer (ground truth)

Plain-English answer: Four active partners with status and numbers, four quiet partners with one line each, and partner-sourced pipeline that reconciles to the deal data.
The answer in one line: rule = every roster partner named exactly once; quiet partners get one line; no partners outside roster.
active_no_dealsWorkWell Group
partner_sourced_pipelinecount=5; amount=550000
per_partner_pipeline{"Apex Rewards Co": {"count": 2, "amount": 275000}, "HRCloud Partners": {"count": 1, "amount": 140000}, "CultureBridge": {"count": 2, "amount": 135000}}
quiet_partnersRecogniq, KudosWave, PeopleFirst Advisors, TotalPerk
rosterpartner_name=Apex Rewards Co; status=active | partner_name=HRCloud Partners; status=active | partner_name=CultureBridge; status=active | partner_name=WorkWell Group; status=active | partner_name=Recogniq; status=quiet | partner_name=KudosWave; status=quiet … (+2 more)
ruleevery roster partner named exactly once; quiet partners get one line; no partners outside roster.
graded checksall_partners · pipeline_amount · quiet_one_line

Model's actual answer

show the model's full answer
# Partner Digest: 2026-08-24 to 2026-09-04

**Partner-sourced pipeline: 5 opportunities, $550,000**
Arithmetic: $180,000 + $95,000 + $140,000 + $60,000 + $75,000 = $550,000 across 5 deals, all with utm_source = Partner.

---

## Active Partners

**Apex Rewards Co** — 2 sourced opps, $275,000 ($180,000 + $95,000: Deal-DDAAF2, Deal-2CF33E). Both reported in DS1 per Slack; both logged with UTM Source = Partner. Co-webinar locked for 09-15.

**HRCloud Partners** — 1 sourced opp, $140,000 (Deal-F1CDA5). Integration referral closed security review and moved to DS2 per Slack.

**CultureBridge** — 2 sourced opps, $135,000 ($60,000 + $75,000: Deal-096E1D, Deal-067213). Both described as early stage per Slack; no stage is recorded in the deal data.

**WorkWell Group** — 0 sourced opps this period. Wants to restart the joint playbook in Q4; planning call booked for 09-09.

## Quiet Partners

**Recogniq** — No activity or sourced deals this period.

**KudosWave** — No activity or sourced deals this period.

**PeopleFirst Advisors** — No activity or sourced deals this period.

**TotalPerk** — No activity or sourced deals this period.

---

Data notes: Stage detail (DS1/DS2, "early stage") comes from the Slack export only — the deal file carries no stage field, so stage-level pipeline totals can't be computed from deal data. Slack opp counts reconcile exactly to deal counts for every partner (Apex 2=2, HRCloud 1=1, CultureBridge 2=2, WorkWell 0=0). All 8 roster partners are covered above; nothing else exists in the provided data.
rfp-answer-from-policyPASS 1.00
Plain-English question: Answer the 10 RFP items from the policy excerpts only — no guessing.
Prompt — what the model saw
Answer the 10 questionnaire items from the policy excerpts only. Cite the excerpt id for every answer. Where the excerpts do not answer the question, mark it unanswerable and say what document would; do not guess. Tone: vendor response, direct.
cost $0.0138 · 8s · in 1,204 / out 842 tok
✓ answered_count✓ unanswerable_flagged✓ citations✓ no_guess

✓ Correct Answer (ground truth)

Plain-English answer: Seven answered with excerpt citations (Q1-Q7), three flagged unanswerable with the document that would answer them (Q8-Q10).
The answer in one line: answered_count = 7; rule = unanswerable items must be flagged and the document that would answer them named; do not guess.
answered_count7
answer_map{"Q1": ["P1"], "Q2": ["P1"], "Q3": ["P2"], "Q4": ["P3"], "Q5": ["P4"], "Q6": ["P5"], "Q7": ["P6"]}
ruleunanswerable items must be flagged and the document that would answer them named; do not guess.
unanswerableQ8, Q9, Q10
valid_citation_idsP1, P2, P3, P4, P5, P6, P7, P8
graded checksanswered_count · unanswerable_flagged · citations · no_guess

Model's actual answer

show the model's full answer
Q1 — Encrypted at rest using AES-256. [P1]

Q2 — Yes, in transit via TLS 1.2 or higher. [P1]

Q3 — Yes: SOC 2 Type II, renewed annually; a copy is available to prospects under NDA. [P2]

Q4 — Data is retained for the contract term plus 90 days, then deleted from production systems. [P3]

Q5 — Published at trust.bonusly.com; customers receive 30 days' notice before a new subprocessor is added. [P4]

Q6 — Yes: SAML 2.0 single sign-on, available on the Enterprise plan. [P5]

Q7 — Production databases backed up daily; backups are encrypted and retained for 35 days. [P6]

Q8 — Unanswerable from the provided excerpts. No excerpt addresses ISO 27001. The trust portal's certifications/compliance page (trust.bonusly.com) or the security certifications documentation would answer this.

Q9 — Unanswerable from the provided excerpts. No excerpt states an uptime SLA. The contractual SLA exhibit in the Master Service Agreement (or the trust portal's SLA documentation) would answer this.

Q10 — Unanswerable from the provided excerpts. No excerpt addresses HIPAA or BAA signing. Bonusly's legal/contracts team (or a HIPAA/BAA policy in the trust portal documentation) would answer this.
ops-maintenance 5 tests
skill-file-reconciliation0.86
Plain-English question: Audit a skill set against its manifest and flag every conflict with a severity and an action.
Prompt — what the model saw
Reconcile the skill set below against its manifest. Find and report (with severity CRITICAL/WARNING/INFO and an action type TRIM_DESC, UPDATE_BODY, DELETE_SKILL, MERGE, or REVIEW): (1) any two skills whose ALWAYS-trigger phrases overlap/duplicate, (2) any circular delegation chain between skills (name it), (3) any delegation target that does not exist (dangling, name it), (4) any version conflict and which skill should survive, (5) how many manifest descriptions exceed 1,024 characters, (6) any hardcoded page ids, dates, or person names in skill bodies, (7) manifest drift in BOTH directions (files with no manifest row, and manifest rows with no file). Make one proposal per finding; rewrite nothing.
cost $0.5204 · 255s · in 66,408 / out 46,652 tok
✗ duplicate_trigger_found✓ circular_chain✓ dangling_target✓ oversized_desc_count✓ hardcoded_values✓ manifest_drift✓ severity_and_action

✓ Correct Answer (ground truth)

Plain-English answer: A severity-ranked list of TRIM_DESC / UPDATE_BODY / DELETE_SKILL / MERGE / REVIEW actions covering every rule violation found.
The answer in one line: files_audited = 14; rank = CRITICAL / WARNING / INFO with one fix action per issue
files_audited14
findingsduplicate trigger, circular delegation chain, dangling file reference, oversized description, hardcoded values, manifest drift
rankCRITICAL / WARNING / INFO with one fix action per issue
graded checksduplicate_trigger_found · circular_chain · dangling_target · oversized_desc_count · hardcoded_values · manifest_drift · severity_and_action

Model's actual answer

show the model's full answer
SIGNALFORGE SKILL SET — MANIFEST RECONCILIATION

Scope audited: 14 manifest rows vs 14 provided .SKILL.md files. Manifest metadata spot-verified: description_chars reproduced exactly by character count for partner-digest (1,004) and signalforge-claim-compressor (1,006); analysis-validator 656 vs 655 hand-counted (±1 tolerance). Totals: 22 findings — 2 CRITICAL, 15 WARNING, 5 INFO.

—————————————————————————————
1) ALWAYS-TRIGGER PHRASE OVERLAP
—————————————————————————————

F1 · CRITICAL · MERGE — comms-drafter ↔ email-drafter
Five trigger phrases are verbatim identical in both descriptions: "write me an email", "draft a follow-up", "what should I say", "bump email", "contract nudge" (plus near-duplicate "help me reply" / "help me reply to this"). Both claim the full email lifecycle (outbound, follow-ups, post-demo recaps, pricing/contract follow-up, end-of-quarter, renewal/expansion, QBR follow-ups, onboarding) and both claim pasted-message review/rewrite/rating. Neither references the other — no disambiguation exists.
Proposal: merge email-drafter into comms-drafter (comms-drafter survives as the broader "any external copy" skill; absorb email-drafter's unique Gmail-signature retrieval and no-markdown rules), delete email-drafter, and repoint deal-strategy-coach's drafting handoff to the merged skill.

F2 · WARNING · TRIM_DESC — pipeline-intelligence-report ↔ weekly-pipeline-report
Three collision clusters: "pipeline update" (PIR: "pipeline update" / WPR: "run the pipeline update"); "pipeline report" family (PIR: "run the pipeline report" / WPR: "generate the pipeline report", "do the pipeline report"); near-identical "what's the pipeline look like" (PIR) vs "what does pipeline look like" (WPR). Both are ALWAYS-class. PIR additionally claims "never answer pipeline questions inline without running it," which would hijack WPR's own triggers.
Proposal: trim the generic phrases ("do the pipeline report", "generate the pipeline report", "what does pipeline look like", "pipeline update") from weekly-pipeline-report's description, leaving pipeline-intelligence-report as owner of generic pipeline-report asks; weekly keeps weekly/SQM/bookings-specific triggers.

F3 · WARNING · REVIEW — next-to-close ↔ deal-strategy-coach ↔ sales-forecast
Close-likelihood phrasing is claimed three ways: next-to-close "which deals are most likely to close", deal-strategy-coach "asks which deals are likely to close", sales-forecast "what do we think we're going to close". Overlap is context-dependent (shortlist vs 1:1 book coaching vs quarter forecast), which a description trim cannot resolve.
Proposal: adopt an explicit carve rule — org-wide close shortlist → next-to-close; per-rep coaching/1:1 prep → deal-strategy-coach; quarter revenue outlook → sales-forecast — and state the carve in all three descriptions.

F4 · INFO · REVIEW — model-selection ↔ all 13 others
model-selection claims "ALWAYS run this skill at the start of every task, without exception — before any planning, execution, or skill invocation begins," a universal trigger that overlaps every other skill's domain by construction and asserts precedence over mandatory gates.
Proposal: rescope it to an orchestrator-invoked planning step (or declare it formally first in a single documented order-of-operations) rather than a self-triggering ALWAYS skill.

—————————————————————————————
2) CIRCULAR DELEGATION CHAIN
—————————————————————————————

F5 · WARNING · UPDATE_BODY — deal-strategy-coach ⇄ email-drafter
Chain: deal-strategy-coach → email-drafter ("When drafting manager-to-prospect emails, use the `email-drafter` skill which automatically retrieves your Gmail signature") → deal-strategy-coach (email-drafter description: "For deal strategy, diagnosis, or coaching (not email drafting), use deal-strategy-coach instead"; body lane marker repeats it). comms-drafter → deal-strategy-coach feeds into the same cycle; no skill references comms-drafter back. Both cycle edges carry terminating lane markers ("do the draft here and suggest deal-strategy-coach"), so the loop is bounded in practice — hence WARNING, not CRITICAL.
Proposal: make the handoff one-directional (deal-strategy-coach may hand drafts to the email skill; the drafting skill references deal-strategy-coach as a pointer only, never as a delegation) — or resolve entirely via the F1 merge, which dissolves the cycle.

—————————————————————————————
3) DANGLING DELEGATION TARGETS (referenced, no manifest row, no file provided)
—————————————————————————————

F6 · WARNING · REVIEW — prospect-research-multithreading
Referenced by: comms-drafter (partner context + "If the Recipient Is Unknown"), email-drafter (unknown recipient + multithread CC), deal-strategy-coach (dedicated "Cross-skill handoff" section). Not in manifest.
Proposal: confirm it lives outside this 14-skill set; if not, create it or strip the three mandatory invocation paths.

F7 · WARNING · REVIEW — bonusly-brand
Referenced as mandatory by comms-drafter (Step 0: "Before drafting any communication, apply the `bonusly-brand` skill"), email-drafter ("apply the bonusly-brand org skill"), sales-forecast ("Always reference `bonusly-brand` skill"), and as the exclusion target in signalforge-claim-compressor. Not in manifest.
Proposal: confirm existence outside this set; three skills make it a mandatory pre-step, so absence breaks their step 0.

F8 · WARNING · REVIEW — eight bonusly-*-questions specialists
analysis-validator §12.4 delegates to: bonusly-data-questions, bonusly-product-questions, bonusly-business-reporting-questions, bonusly-rewards-questions, bonusly-ppp-questions, bonusly-feature-flag-questions, bonusly-deal-desk-questions, bonusly-datadog-questions. §11 also lists cascade files CUSTOMER_DATA_REFERENCE, HUBSPOT_CONNECTOR_REFERENCE, SIGNALFORGE_PRODUCT_INSIGHT_SKILL. None in manifest. If these do not exist outside the set, the mandatory validator's "always delegate, never improvise" path for product/rewards/flags/legal claims is broken (CRITICAL in that case).
Proposal: verify the eight skills exist in the wider registry; if any is missing, remove the delegation row rather than leave a silent dead end in a never-skip gate.

F9 · INFO · REVIEW — skill-orchestrator
Referenced by analysis-validator §11 (Three-Way Sync cascade) and signalforge-feedback (activation checklist: "Skill registered in skill-orchestrator"). Not in manifest — meaning no registered router exists in this set to adjudicate the trigger conflicts in section 1.
Proposal: confirm it exists outside this set; if not, the orchestrator-dependent sync/registration steps in both skills are dead references.

F10 · WARNING · REVIEW — signalforge-reports org skill
Referenced as mandatory pre-build reads by pipeline-intelligence-report (steps 1–3: SKILL.md, DESIGN-SYSTEM.md, signalforge.css, plus reports.html and brand-lockup.html) and weekly-pipeline-report (must-read before HTML). Neither HTML report skill contains the design system it requires. Not in manifest.
Proposal: confirm it exists at /mnt/skills/organization/signalforge-reports/; if absent, both report skills cannot produce compliant output.

F11 · INFO · REVIEW — untracked reference files
sales-forecast references 5 local files (references/data-sources.md, report-structure.md, cadence.md, report-template.html, TEMPLATE_README.md); weekly-pipeline-report references 2 (report-spec.md, queries.md); stale-pipeline-report calls /mnt/skills/public/xlsx/scripts/recalc.py. None are manifest-tracked or provided; existence unverifiable from the supplied data.
Proposal: add a reference-files column to the manifest (or a per-skill file inventory) so these dependencies are reconcilable.

—————————————————————————————
4) VERSION CONFLICTS
—————————————————————————————

F12 · WARNING · UPDATE_BODY — analysis-validator internal version conflict
Frontmatter, changelog, and footer declare v3.6 (May 9, 2026), but the §7 Validation Trail template prints "Validator: analysis-validator v3.2", and §1's Full Mode enumerates "(G1-A through G1-H)" and "(G2-A through G2-E)" — omitting G1-I, G1-J, G1-K, G1-L (added v3.1–v3.5) and G2-F (added v3.6). The changelog also lists 3.6 above 3.5 and retains pre-restructure "Check 12 / Check 3C" numbering that no longer exists. Survivor: the v3.6 body — the v3.2 stamp and stale ranges are un-updated template artifacts.
Proposal: update §1's check ranges and the §7 trail version stamp to v3.6 in analysis-validator. (No other cross-skill version conflict found: pipeline-intelligence-report v6's footer correctly cites validator v3.6; the v4 design-system references in pipeline-intelligence-report and weekly-pipeline-report are mutually consistent.)

F13 · CRITICAL · UPDATE_BODY — AE roster conflict; analysis-validator's roster survives
pipeline-intelligence-report Phase 1 hardcodes "AE owner IDs (verified May 2026)" as 5 AEs: Bryce Harmon 119337721, Dana Mercer 83155923, Cole Ingram 83155924, Alex Franklin 84342457, Gavin Porter 1520255671. analysis-validator §12.3 (Updated May 4, 2026) defines the Core 6 as those 5 plus Hugo Lindqvist (77260721). Set difference: {Hugo Lindqvist, 77260721} — 6 − 5 = 1 AE missing. Since pipeline-intelligence-report is the self-declared master ("never answer pipeline questions inline without running it"), its scoring, AE-on-leave detection, and By-AE tab silently drop one AE's entire book. stale-pipeline-report v1.1 is the correct going-forward mechanism ("Never hardcode rep names or owner IDs") but does not conflict on data.
Proposal: replace pipeline-intelligence-report's hardcoded 5-ID block with the analysis-validator §12.3 Core-6 roster (canonical survivor) or dynamic owner resolution per stale-pipeline-report, and drop the "verified May 2026" date stamp.

—————————————————————————————
5) DESCRIPTIONS EXCEEDING 1,024 CHARACTERS
—————————————————————————————

Answer: 0 of 14.
Arithmetic: sorted manifest description_chars descending — 1,006; 1,006; 1,004; 996; 965; 962; 945; 897; 792; 762; 708; 676; 656; 656. Max = 1,006 (pipeline-intelligence-report, signalforge-claim-compressor). 1,024 − 1,006 = 18, so every description is at least 18 characters under the threshold. No finding; no action.
Observation (no action proposed): the top two sit 18 chars and partner-digest 20 chars from the ceiling — any description growth during the F1/F2 merges will hit the limit first in exactly these files.

—————————————————————————————
6) HARDCODED PAGE IDs, DATES, PERSON NAMES IN BODIES
—————————————————————————————

F14 · WARNING · UPDATE_BODY — partner-digest (worst in set)
9 Confluence identifiers (cloudId 73fe98de-a4a3-4869-9f8a-bb1eeed4cf7f; spaceId 1958248479; folder 2286616609 used 4×; page IDs 2286321666, 2265382925, 2236940297, 2237825028, 2239365136, 2238283777), Slack user ID U03QLMBL7AR, persons Amani Phipps (owner field + workflow) and contacts Kelli, Jen Lee, Hani, Bryce, Sara, and dates May 16/May 19/June 2, 2026.
Proposal: externalize IDs and contact names to a maintained reference file and replace the person-bound owner with a role.

F15 · WARNING · UPDATE_BODY — signalforge-feedback
Page IDs 2295136266 (Feedback Log), 2234417154 (parent), 2247295002 (Build Log), spaceId 2232811524, cloudId 73fe98de-…; example titles embed a person ("Gavin Porter Rep Diagnostic") and a customer ("Lowe's Conversation Analysis").
Proposal: move the five IDs to a config/reference block and genericize the example titles.

F16 · WARNING · UPDATE_BODY — sales-forecast
Space ID 2232811524, parent page 2232582148, cloudId 73fe98de-…; example title date "July 9, 2026"; changelog carries a dead person reference ("Elena → Alaina"); Q2-2026 framing in examples.
Proposal: same config-block treatment; drop the Elena changelog artifact and genericize example dates.

F17 · WARNING · UPDATE_BODY — weekly-pipeline-report
Person: "Ben Lavin" in the H1 plus 4 operational references ("for Ben's review", "deliver the HTML file to Ben"). Two hardcoded Google Sheet IDs (1CLZeOsElVDF_LF0ZG_t2nfwvhnZ6bpwqM_nX3WEYzcw, 1ENuaEcCuLjdKhMvp8FK3Ys1ek5Aw9ZuOZhsHJJFoB_k). Static dated figures: Q1 2026 bookings $365,152 vs $475,000 (365,152/475,000 = 76.9% ≈ 77%, arithmetic checks out but is frozen) and pipeline addition $2,490,532 vs $3,288,000 (75.7% ≈ 76%). Quarter window "April 1 – June 30, 2026" hardcoded despite the skill's own quarter-agnostic intent.
Proposal: replace the named owner with a role, move sheet IDs to a reference, and mark the Q1 figures as frozen historical context or remove them.

F18 · WARNING · UPDATE_BODY — deal-strategy-coach
Confluence playbook URL hardcodes page ID 2257879045 and date slug "April 2026"; person names "Farid" (.edu routing) and "Perseus" (India routing) in ICP rules; pricing table header "Pricing — 2026".
Proposal: move the playbook link and the two name-bound routing rules to a dated, maintained reference and add a review cadence to the pricing table.

F19 · WARNING · UPDATE_BODY — analysis-validator
§12.3 hardcodes 19 named people with HubSpot owner IDs (6 AEs, 7 CSMs, 2 leadership, 1 demand-gen, 3 RevOps); G1-K and §10 escalate to named individuals ("Manish or Amani"); dated facts baked in ("March 28, 2023" staleness, "May 4, 2026" removals, "as of May 2026" ranges); G1-J anchors ~452,000 / ~110,097 contradict the skill's own §8 doctrine "Do not use hardcoded figures."
Proposal: keep structural constants (stage IDs, canonical filters) in-body, move the roster and population anchors to runtime resolution (HubSpot:search_owners / §8 live query), and swap named escalation to a Finance role.

F20 · WARNING · UPDATE_BODY — model-selection
last_checked: 2026-05-19; against this session's date (2026-10-19) that is 12+30+31+31+30+19 = 153 days, breaching the skill's own 14-day staleness rule by 139 days. The registry's model IDs, prices, context windows, and deprecation dates (Opus 4.7, Sonnet 4.6, April 14, 2026 deprecation note) are all frozen body content.
Proposal: run its own self-update procedure and relocate the registry to a reference file with an enforced freshness check.

F21 · INFO · UPDATE_BODY — stale-pipeline-report
Slack channel ID C0561C1JCPJ (×2), example non-AE owner ID 55483190, example dates 5/15, 5/19, 5/7. Persons deliberately excluded (dynamic owner resolution) — the cleanest pattern in the set.
Proposal: move the channel ID to config and genericize the example dates.

F22 · INFO · UPDATE_BODY — closed-lost-analysis
≥10 dated customer/prospect cases embedded as taxonomy evidence (Softheon May 2026, Estee Lauder RFP delay, MinIO May 4–12 vacation, Aurora Innovation, GCash, Ethos Cannabis, StickerYou, LIFTOFF, Nestlé, Ozinga), "confirmed May 2026" field notes ×2, and a frozen sample stat ("30-deal AI-field sample from May 2026: 10 of 10").
Proposal: tag these as dated historical evidence or move them to a reference file so the taxonomy claims don't silently age.

Clean on this dimension: comms-drafter, email-drafter (no persons/dates/IDs). next-to-close, pipeline-intelligence-report, and stale-pipeline-report share the HubSpot portal constant 1973303 in deal URLs — a stable tenant identifier required for link construction; no action proposed. signalforge-claim-compressor contains only clearly illustrative examples (Panopto, Schneider Downs, Felix Construction); no action.

—————————————————————————————
7) MANIFEST DRIFT (both directions)
—————————————————————————————

None, in either direction.
Direction A — files with no manifest row: 0. Provided .SKILL.md files = 14; each stem (analysis-validator, closed-lost-analysis, comms-drafter, deal-strategy-coach, email-drafter, model-selection, next-to-close, partner-digest, pipeline-intelligence-report, sales-forecast, signalforge-claim-compressor, signalforge-feedback, stale-pipeline-report, weekly-pipeline-report) has exactly one row.
Direction B — manifest rows with no file: 0. All 14 rows resolve to a provided body, and declared_name = filename stem = frontmatter name in all 14 cases (14 = 14, 1:1 both ways).
Scope caveat, stated explicitly: this null result covers the 14 skills supplied. Skills and files referenced by the bodies but absent from the manifest (bonusly-brand, prospect-research-multithreading, skill-orchestrator, the eight bonusly-*-questions specialists, signalforge-reports, and the references/ files) are reported as dangling-dependency findings F6–F11, not as row drift. body_lines values were not independently re-counted line-by-line; description_chars were spot-verified exactly for two files as noted in the scope line.
runbook-from-slack-threadPASS 1.00
Plain-English question: Turn the resolved incident thread into a runbook.
Prompt — what the model saw
Turn the resolved incident thread into a runbook: ordered steps with the exact command or action taken, who ran it, how success was verified, and rollback for each step that changed state. Every step must trace to a message in the thread. Add nothing the thread does not contain; where the thread is unclear, mark the step as needing confirmation.
cost $0.0525 · 31s · in 1,367 / out 5,947 tok
✓ ordered_steps✓ rollback✓ needs_confirmation✓ sourced

✓ Correct Answer (ground truth)

Plain-English answer: Ordered steps from the thread, each with the exact command, who ran it, how success was verified, and rollback where state changed.
The answer in one line: rule = no steps beyond what the thread contains; every step traces to a message id.
ambiguous_steporder=4; msg=M05; mark=needs-confirmation
ruleno steps beyond what the thread contains; every step traces to a message id.
stepsorder=1; msg=M01; actor=Bryce Harmon; action=acknowledge PagerDuty alert, take IC; state_change=False; verified_by=M01 | order=2; msg=M02; actor=Farid Osman; action=check queue depth; command=bundle exec rake sidekiq:queue_depth; state_change=False; verified_by=M02 (48,213 pending) | order=3; msg=M04; actor=Farid Osman; action=pause auto-enqueue via feature flag; command=bin/rails runner 'FeatureFlag.disable(:auto_recognition_enqueue)'; state_change=True; rollback=bin/rails runner 'FeatureFlag.enable(:auto_recognition_enqueue)'; verified_by=M07 (queue depth falling) | order=4; msg=M05; actor=Elena Sinclair; action=cleared the dead set; state_change=True; needs_confirmation=thread gives no command or method for clearing the dead set | order=5; msg=M06; actor=Bryce Harmon; action=scale reward-worker to 6 replicas; command=kubectl scale deployment/reward-worker --replicas=6; state_change=True; rollback=kubectl scale deployment/reward-worker --replicas=3; verified_by=M07/M08 | order=6; msg=M08; actor=Cole Ingram; action=verify queue drained and error rate baseline; command=bundle exec rake sidekiq:queue_depth; state_change=False; verified_by=M08 (returns 0) … (+2 more)
graded checksordered_steps · rollback · needs_confirmation · sourced

Model's actual answer

show the model's full answer
RUNBOOK: REWARD-WORKER QUEUE BACKLOG INCIDENT
Source: #incident-2026-08-29-reward-queue (resolved). Every step cites its source message [Mxx]. Nothing outside the thread is included; gaps are flagged "needs confirmation." Steps in thread order.

STEP 1 — Alert triage, IC assignment (14:02:10Z) [M01]
  Who: Bryce Harmon
  Action: PagerDuty alert fired for reward-worker queue depth > 10k; acknowledged and took IC.
  State change: none. Rollback: none needed.
  Verification: none stated in thread (the alert firing is the trigger itself).

STEP 2 — Confirm queue depth (14:04:33Z) [M02]
  Who: Farid Osman
  Command: `bundle exec rake sidekiq:queue_depth`
  Result: 48,213 pending jobs; thread states normal is under 500.
  State change: none. Rollback: none needed.

STEP 3 — Inspect dead set (14:06:02Z) [M03]
  Who: Farid Osman
  Finding: dead set held 112 jobs, all Redis::TimeoutError, originating around 13:58Z.
  Needs confirmation: the thread records the finding but not the command used to inspect the dead set. Confirm the inspection command before running this step.
  State change: none. Rollback: none needed.

STEP 4 — Pause enqueue (14:08:45Z) [M04]
  Who: Farid Osman
  Command: `bin/rails runner 'FeatureFlag.disable(:auto_recognition_enqueue)'`
  State change: yes — disables the :auto_recognition_enqueue feature flag.
  Verification: not stated in the thread; nobody confirms the flag was actually disabled before the next step. Needs confirmation.
  Rollback (given in thread): `bin/rails runner 'FeatureFlag.enable(:auto_recognition_enqueue)'`

STEP 5 — Clear the dead set (14:15:20Z) [M05]
  Who: Elena Sinclair
  Action: "While I was in the console I cleared out the dead set" — removes the 112 jobs counted in M03.
  State change: yes — destructive removal of 112 dead jobs.
  Verification: none in thread (no re-check of the dead set afterward). Needs confirmation.
  Rollback: none in thread, and the disposition of the 112 discarded Redis::TimeoutError jobs is never addressed. Needs confirmation (treat as irreversible until clarified).

STEP 6 — Scale workers up (14:21:07Z) [M06]
  Who: Bryce Harmon
  Command: `kubectl scale deployment/reward-worker --replicas=6` (was 3)
  State change: yes — replicas 3 → 6.
  Verification: no direct confirmation in thread that the replica count took effect; downstream queue recovery (M07, M08) is consistent with added capacity but proves nothing about replicas. Needs confirmation.
  Rollback (given in thread): `kubectl scale deployment/reward-worker --replicas=3`

STEP 7 — Monitor drain (14:33:41Z) [M07]
  Who: Farid Osman
  Action/Result: queue depth 9,400 and falling ~1,200/min.
  Needs confirmation: re-check command not stated at this point (presumably the same rake task as M02).
  Arithmetic check (derived from thread numbers only):
    Elapsed M02→M07: 14:04:33Z → 14:33:41Z = 29 min 8 s ≈ 29.13 min
    Drained: 48,213 − 9,400 = 38,813 jobs
    Average rate: 38,813 ÷ 29.13 ≈ 1,332/min — consistent with the reported ~1,200/min falling rate
  State change: none. Rollback: none needed.

STEP 8 — Verify drain complete (14:47:55Z) [M08]
  Who: Cole Ingram
  Command: `bundle exec rake sidekiq:queue_depth`
  Result: returns 0; Datadog error rate back to baseline.
  Arithmetic check: 9,400 at 14:33:41Z → 0 by 14:47:55Z = 14 min 14 s ≈ 14.23 min → 9,400 ÷ 14.23 ≈ 660/min average on this leg, consistent with a drain rate that fell as the queue emptied.
  State change: none. Rollback: none needed.

STEP 9 — Re-enable enqueue (14:49:10Z) [M09]
  Who: Bryce Harmon
  Command: `bin/rails runner 'FeatureFlag.enable(:auto_recognition_enqueue)'`
  State change: yes — re-enables the flag disabled in Step 4.
  Verification (from thread): 40 new jobs processed cleanly in the next 3 minutes.
  Rollback: re-disable using the identical command from M04: `bin/rails runner 'FeatureFlag.disable(:auto_recognition_enqueue)'`

STEP 10 — Scale workers back down, close (14:55:00Z) [M10]
  Who: Bryce Harmon
  Command: `kubectl scale deployment/reward-worker --replicas=3`
  State change: yes — replicas 6 → 3.
  Verification (from thread): queue stable at 0; incident declared resolved.
  Rollback: scale back up using the identical command from M06: `kubectl scale deployment/reward-worker --replicas=6`

MISSING FROM THREAD (explicit gaps)
  - Exact command used to inspect the dead set (M03).
  - Verification that the enqueue flag was actually disabled before proceeding (M04).
  - Exact clear-dead-set command, post-clear verification, and rollback/disposition of the 112 discarded jobs (M05).
  - Direct confirmation the deployment scaled to 6 (M06) — only downstream queue recovery is shown.
  - Command used for the 14:33 depth re-check (M07).
  - Root cause: never stated. The only cause signal in the thread is the Redis::TimeoutError burst starting around 13:58Z (M03); nothing in the thread identifies why those errors began.
datadog-incident-timelinePASS 1.00
Plain-English question: Sequence the failure from the logs — what broke first, and what cascaded after it?
Prompt — what the model saw
From the log slice, sequence the failure: first error with timestamp, the cascade in order, the service and job involved, and the Datadog query that would confirm the first error. State what the logs do not show.
cost $0.0464 · 25s · in 2,405 / out 4,789 tok
✓ first_error_ts✓ service_named✓ cascade_ordered✓ query✓ gaps_stated

✓ Correct Answer (ground truth)

Plain-English answer: A reward-service Redis timeout at 14:01:12Z cascades to sidekiq at 14:01:40 and api-gateway at 14:03:05, with the Datadog query that confirms the first error.
The answer in one line: job_class_involved = RewardGiveJob
cascade_orderservice=reward-service; first_error_at=2026-09-03T14:01:12Z | service=sidekiq; first_error_at=2026-09-03T14:01:40Z | service=api-gateway; first_error_at=2026-09-03T14:03:05Z | service=web-app; first_error_at=2026-09-03T14:03:30Z
clean_servicespostgres
confirming_queryservice:reward-service status:error "@message:Redis::TimeoutError"
first_errortimestamp=2026-09-03T14:01:12Z; service=reward-service; message=Redis::TimeoutError: Connection to Redis at redis-primary:6379 timed out after 5s
job_class_involvedRewardGiveJob
logs_do_not_showroot cause of the Redis timeout, redis-server's own logs, infrastructure metrics (CPU/memory/network), any deploy or config change preceding the incident, who applied mitigation or when (no actor fields in logs)
graded checksfirst_error_ts · service_named · cascade_ordered · query · gaps_stated

Model's actual answer

show the model's full answer
## Failure sequence

**First error**
- 2026-09-03T14:01:12Z — reward-service — ERROR — `Redis::TimeoutError: Connection to Redis at redis-primary:6379 timed out after 5s`

**Cascade in order (all timestamps 2026-09-03 UTC)**

1. 14:01:12 — reward-service — first Redis timeout (above).
2. 14:01:20, 14:01:30, 14:01:40 — reward-service — `retry exhausted for RewardGiveJob` (3 lines, 10s apart).
3. 14:01:40 — sidekiq — `RewardGiveJob failed: Redis::TimeoutError; retrying in 60s`. Additional sidekiq failure/retry lines at 14:02:28, 14:03:31, 14:04:22, 14:05:26, 14:06:47 (6 total).
4. 14:01:46–14:02:58 (sidekiq_jobs.csv) — 12 RewardGiveJob failures, J-00001 through J-00012, all `Redis::TimeoutError`. Earliest failed_at is 14:01:46Z (J-00005), 34s after the first error (14:01:46 − 14:01:12).
5. 14:02:30 — sidekiq — WARN — `Queue reward depth above 10,000`.
6. 14:02:36–14:05:50 (sidekiq_jobs.csv) — 4 RecognitionDigestJob failures, J-00013 through J-00016 — collateral damage on the same Redis dependency.
7. 14:03:05 — api-gateway — `502 upstream timeout calling reward-service /gives` — first user-path symptom (1m53s after first error: 14:03:05 − 14:01:12).
8. 14:03:30 — web-app — `Give form submission failed: upstream 502 from api-gateway` — end-user impact begins. Further web-app errors at 14:04:45, 14:05:42, 14:06:49; further api-gateway 502s at 14:03:48, 14:04:13, 14:05:16, 14:06:52 (5 gateway 502s total).
9. 14:06:52 — api-gateway — last error in the slice.
10. 14:22:10 — reward-service — INFO — `Redis connection restored; resuming job processing`.
11. 14:24:45 — sidekiq — INFO — `Queue reward depth below 500`.

**Service and job involved**
- Failing service: reward-service, dependency Redis at redis-primary:6379.
- Job: RewardGiveJob (primary, 12 recorded failures); RecognitionDigestJob collateral (4). Queue: `reward` on sidekiq. Blast radius: reward-service → api-gateway → web-app give form.

**Arithmetic**
- First error → restoration notice: 14:01:12Z → 14:22:10Z = 20 min 58 s (this includes a silent window, see below; it is a span, not a measured outage length).
- Visible error span: 14:01:12Z → 14:06:52Z = 5 min 40 s.
- First 502 → last web-app error: 14:03:05Z → 14:06:49Z = 3 min 44 s.
- Queue drain window: 14:02:30Z (>10,000) → 14:24:45Z (<500) = 22 min 15 s; actual depth between the two points unknown.
- Job failures: 12 + 4 = 16; RewardGiveJob = 12/16 = 75%, RecognitionDigestJob = 4/16 = 25%.
- Log lines vs jobs table: 6 sidekiq failure log lines vs 12 RewardGiveJob rows — the log slice is not a complete record.

**Datadog query confirming the first error** (time range 2026-09-03 14:00:00–14:05:00 UTC, sorted ascending, take first hit):
`service:reward-service status:error "Redis::TimeoutError: Connection to Redis at redis-primary:6379 timed out"`
Caveat: facet names `service`/`status`/`message` are assumed from the CSV columns; any `env`/`host` facets depend on pipeline configuration not shown in the data.

**What the logs do not show**
- Root cause of the Redis timeout: no redis logs, no metrics for redis-primary (CPU, memory, maxclients, failover, restart). "Redis connection restored" states recovery, not cause.
- Exact outage end: no failing-path logs between 14:06:52 and 14:10:56; connectivity could have returned any time before the 14:22:10 restoration notice.
- Final disposition of the 16 failed jobs: no retry-success or dead-set lines; whether they eventually completed, were discarded, or were lost is unknown.
- Whether the give form recovered: no post-restoration "request completed" line for the /gives path in the slice.
- Business/customer impact: no deal, company, or user identifiers appear anywhere in either file, so no alias can be cited — none exist in this data.
- Volume context: no request counts, no 502 rate, no queue depth between the two reported values (10,000 and 500), no deploy/config-change events.
feature-flag-state-summaryPASS 1.00
Plain-English question: Summarize what each feature flag controls and who has it on.
Prompt — what the model saw
Summarize the state of the listed feature flags: what each controls per the code excerpt, which companies or segments have it on, targeting rules, and any flag with no code reference or no targets. Company counts must match the export. Do not describe a flag that is not in the export.
cost $0.0507 · 27s · in 1,231 / out 5,751 tok
✓ orphans✓ no_targets✓ counts_match✓ no_invented

✓ Correct Answer (ground truth)

Plain-English answer: Recognition streaks on for 42 beta companies, redeem-flow redesign off with a 12-company target list, plus any flag with no code reference or no targets.
The answer in one line: rule = company counts must equal the export; do not describe flags absent from the export.
code_referencedrecognition_streaks_v2, points_budget_guardrails, slack_dm_nudges, redeem_flow_redesign, analytics_dashboard_v3, ms_teams_app_v2
flags{"recognition_streaks_v2": {"state": "on", "targeting_rules": "segment:beta_companies", "company_count": 42}, "points_budget_guardrails": {"state": "on", "targeting_rules": "all_companies", "company_c
flags_with_no_targetspaused_offboard_cleanup
orphan_flags_no_code_referencelegacy_give_modal, survey_boosters_q3, paused_offboard_cleanup
rulecompany counts must equal the export; do not describe flags absent from the export.
graded checksorphans · no_targets · counts_match · no_invented

Model's actual answer

show the model's full answer
Feature flag summary (sources: flags_export.csv and flag_code.md only)

Export contains 9 flags: 5 on, 4 off (5+4=9). 6 are referenced in the code excerpt, 3 are not. 1 has no targets.

1. recognition_streaks_v2
   State: on. Targeting: segment:beta_companies. Count: 42.
   Code (app/models/recognition.rb): when enabled, StreakTracker.record(give) runs; gives are recorded toward streak tracking.
   On for: beta_companies segment, 42 companies per export.

2. points_budget_guardrails
   State: on. Targeting: all_companies. Count: 220.
   Code (app/services/budget_service.rb): when enabled, BudgetService.new(company).enforce!(giver, points) runs; budget rules are enforced on point gives.
   On for: all companies. 220 is the only company-base figure the export provides.

3. slack_dm_nudges
   State: on. Targeting: segment:region_na. Count: 87.
   Code (app/jobs/nudge_job.rb): guard clause; job exits unless enabled, otherwise SlackDm.send_nudge(user) fires.
   On for: region_na segment, 87 companies.

4. redeem_flow_redesign
   State: off. Targeting: targeted_list. Count: 12.
   Code (app/controllers/redeem_controller.rb): enabled renders RedeemV2Component, disabled renders RedeemV1Component. With state off, the V1 path renders.
   Targets: explicit list of 12 companies; members are not named in the export.

5. analytics_dashboard_v3
   State: on. Targeting: segment:tier_three. Count: 65.
   Code (app/controllers/analytics_controller.rb): when enabled, @dashboard = AnalyticsV3.new(company). No else branch appears in the excerpt, so the disabled path is not shown in the provided code.
   On for: tier_three segment, 65 companies.

6. ms_teams_app_v2
   State: off. Targeting: targeted_list. Count: 9.
   Code (app/services/teams_installer.rb): when enabled, TeamsAppV2.install(company) runs.
   Targets: explicit list of 9 companies; members not named in the export.

7. legacy_give_modal
   State: off. Targeting: segment:legacy_plan. Count: 14.
   No code reference in the excerpt. What it controls cannot be determined from the data provided.
   Segment: legacy_plan, 14 companies per export.

8. survey_boosters_q3
   State: on. Targeting: segment:legacy_plan. Count: 7.
   No code reference in the excerpt. What it controls cannot be determined from the data provided.
   On for legacy_plan segment, 7 companies per export.

9. paused_offboard_cleanup
   State: off. Targeting rules: none (blank). Count: 0.
   No code reference and no targets. Dormant on both counts.

Cross-checks and caveats
- All 6 flags referenced in the code excerpt appear in the export, so there is no code-only flag to report.
- Arithmetic: 5 on + 4 off = 9 total, matching the export's 9 rows. Sum of all company_count values: 42+220+87+12+65+9+14+7+0 = 456. This is not a unique-company count; targeting overlaps and points_budget_guardrails covers all companies.
- The export does not define company_count for off-state flags: redeem_flow_redesign (off, 12), ms_teams_app_v2 (off, 9), and legacy_give_modal (off, 14) all carry nonzero counts. Counts are reported exactly as exported; whether those companies have the flag effectively on or are merely on a defined list is not stated in the data.
- The legacy_plan segment shows two different counts (14 under legacy_give_modal vs 7 under survey_boosters_q3). The export gives no explanation for the difference.
- No individual company names appear in the export, so "who has it on" can only be answered at segment or list level.
nda-triagePASS 1.00
Plain-English question: Triage each NDA green, yellow, or red — and quote the clause that decides it.
Prompt — what the model saw
Triage each NDA as GREEN (standard approval), YELLOW (counsel review), or RED (full legal review). Cite the clause that drives each classification by quoting it. Catch embedded non-solicits, non-competes, missing carve-outs, and unusual term or governing law. One line of reasoning per NDA.
cost $0.0173 · 11s · in 1,284 / out 1,278 tok
✓ green✓ yellow✓ red✓ non_solicit_caught✓ clause_quoted✓ correct_labels

✓ Correct Answer (ground truth)

Plain-English answer: NDA-1 green (standard mutual terms), NDA-2 yellow for an embedded 18-month non-solicit, with the driving clause quoted for every classification.
The answer in one line: non_solicit_in = NDA-2
checksclassification_per_nda, clause_quotes_exist_in_document, non_solicit_caught, carveouts_checked
classifications{"NDA-1": {"classification": "GREEN", "driving_reason": "standard mutual NDA: 2-year term, standard carve-outs, Delaware law", "driving_clause_quote": "Carve-outs: information that (a) is or becomes p
non_solicit_inNDA-2
graded checksgreen · yellow · red · non_solicit_caught · clause_quoted · correct_labels

Model's actual answer

show the model's full answer
NDA-1 — [PARTY A]: GREEN. Mutual, all five standard carve-outs present — "information that (a) is or becomes publicly available through no breach, (b) was known prior to disclosure, (c) is received from a third party without duty of confidence, (d) is independently developed, or (e) must be disclosed by law or court order" — with a standard 2-year term / 3-year survival and Delaware law; no restrictions embedded.

NDA-2 — [PARTY B]: YELLOW (counsel review). Mutual with standard carve-outs, but it embeds a hiring restriction — "During the term of this Agreement and for eighteen (18) months thereafter, neither party shall solicit for employment or hire any employee of the other party with whom it came into contact under this Agreement." Arithmetic: 3-year term + 18-month tail = up to 4.5 years of recruiting constraint; mutual and contact-scoped, so counsel sign-off suffices.

NDA-3 — [PARTY C]: RED (full legal review). Driven by the non-compete — "For a period of three (3) years following the Effective Date, Recipient shall not, directly or indirectly, engage in or provide services to any business that competes with Discloser's business." — compounded by three aggravators: no confidentiality carve-outs stated at all, a 5-year one-way term in [PARTY C]'s favor, and Irish law with exclusive jurisdiction ("the parties submit to the exclusive jurisdiction of its courts").

Missing data: no Effective Dates given in any NDA, so all tails above are computed from the stated terms, not calendar dates.