MRR by month and churn for the first half
The request
The task request supplied to the adapter, alongside the workspace below. Runtime instructions, authorized tools and skills also affect the system. Containerized runs withhold host-side references; local mode is not an isolation boundary.
The board pack needs our MRR month by month for the first half and what we lost to churn. Build it as mrr.xlsx with live formulas and write memo.md alongside it. Marcus's email says how he wants the subscriptions counted.
Workspace
- email_from_finance.txt
- opening_book_dec2025.xlsx
- subscription_events_2026.csv
Expected deliverables
- memo.md
- mrr.xlsx
Recorded acceptance checks
The table uses check names and outcomes from the frozen campaign ledger. Acceptance requires all required predicates to pass; a predicate may itself contain a tolerance or coverage threshold. Its scope is part of the measurement contract.
| Recorded check | Type | Required | Failed / observed |
|---|---|---|---|
| mrr.xlsx exists | file_exists | yes | 0 / 6 |
| live formulas | xlsx_has_formulas | yes | 0 / 6 |
| no error cells | xlsx_no_errors | yes | 0 / 6 |
| March ending MRR | xlsx_value_present | yes | 1 / 6 |
| June ending MRR | xlsx_value_present | yes | 1 / 6 |
| churned MRR in May | xlsx_value_present | yes | 2 / 6 |
| churned MRR for the half | xlsx_value_present | yes | 1 / 6 |
| memo carries the June MRR and the churn figure | text_numbers_present | yes | 0 / 6 |
| memo names the month the churn landed in | text_sentence_matches | yes | 1 / 6 |
Current task declaration (9 checks)
From task.yaml. Original runner grading and frozen retrospective scoring are separate. Read the frozen source for the rules behind the reported results.
- mrr.xlsx exists —
file_exists, required - live formulas —
xlsx_has_formulas, required - no error cells —
xlsx_no_errors, required - March ending MRR —
xlsx_value_present, required - June ending MRR —
xlsx_value_present, required - churned MRR in May —
xlsx_value_present, required - churned MRR for the half —
xlsx_value_present, required - memo carries the June MRR and the churn figure —
text_numbers_present, required - memo names the month the churn landed in —
text_sentence_matches, required
Recorded results, desk release
| System | Repetitions | Frozen / 3 | Original / 3 | Median minutes | Mean model estimate (USD) |
|---|---|---|---|---|---|
| Proto / DeepSeek V4.1 Flash | 1 | 1 | 10.1 | 0.1212 | |
| Codex / GPT-5.6 sol | 1 | 1 | 8.9 | 0.9275 |
These are repeated executions on the released fixture, not independent new business cases. Cost is captured model usage priced under the recorded assumptions, excluding infrastructure and human review. Per-attempt check verdicts, raw and frozen verdicts, usage, receipt and artifact hashes for this task are in results/latest/attempts.jsonl under "task": "mrr-report".
Recorded results, desk comparison, 28 September 2026
| System | Repetitions | Frozen / 3 | Original / 3 | Median minutes |
|---|---|---|---|---|
| Proto / GPT-6 sol | 1 | 1 | 3.7 | |
| Codex / GPT-6 sol | 2 | 2 | 4.0 | |
| Proto / DeepSeek V4.1 Flash | 1 | 1 | 5.3 |
A separate campaign under the same frozen scorer, not pooled with the release. Conditions and data.