# Review change log

## Outcome and changes

**No calculation, inclusion/exclusion or CSV correction was required.** The independently recalculated results match all 21 metric cells in `output/summary.md`, **October results**. The four CSV deliverables are byte-identical to the original submission.

- Preserved the original five deliverables at `original-submission/output/` and the original calculation script at `original-submission/work/reconcile.py`. Copy integrity was checked against SHA-256 hashes recorded in `original-submission/manifest.json`.
- Clarified `output/summary.md`, **Cleaning decisions and assumptions**: blank source amounts/unresolved products are unknown, while blank exclusion metadata can mean not applicable. Added an **Independent submission review** section. No financial figures changed.
- Added this change log and `output/review-checks.txt`. The independent review script is `work/review.py`; it derives eligibility directly from raw records rather than reusing the original script's row-ID exclusion map.

## Evidence and requirements reviewed

- **Source conflicts:** N1008 retains S10 at 180.00 because its update timestamp is later than S09's. S04 duplicates S03; R03 duplicates R02; A04 duplicates A03. Duplicate equality appropriately ignores source row IDs. Sources: `inputs/sales.csv`, S03–S04, S09–S10; `inputs/refunds.csv`, R02–R03; `inputs/ads.csv`, A03–A04; `inputs/policy.md`, **Northline Workshops accounting note — October 2026**.
- **Settlement and missing evidence:** pending R04 is excluded despite its populated date; R07 is outside October; R05 is unmatched; R06's amount remains blank. The policy's statement that refunds relate to October orders does not establish a valid paid order for N9999. No product or refund amount was guessed. Sources: `inputs/refunds.csv`, R04–R07; `inputs/sales.csv`, S01–S15; accounting note above.
- **Calculations and reconciliation:** independently checked all seven metrics for Workshop, Course and combined, including the weighted combined per-order figure. Checked 15 sales rows = 9 included + 6 excluded; 7 refund rows = 2 included + 5 excluded; 7 ad rows = 5 included + 2 excluded. Known amount controls reconcile, with the missing refund amount explicitly separate. Sources: `inputs/sales.csv`, S01–S15; `inputs/refunds.csv`, R01–R07; `inputs/ads.csv`, A01–A07.
- **File usability:** parsed all four CSV deliverables with Python's standard-library CSV reader; checked consistent field counts, unique headers, every normalized source field, all 29 original row IDs, the derived refund product joins, and exact exception coverage. Dates and known money values agree with the raw inputs under the policy. Full `updated_at` timestamps are retained to preserve version evidence. Users must filter cleaned files on `included_in_october=yes` before summing.
- **Assignment coverage and claims:** confirmed the five requested deliverables, required metrics, product/combined reporting, record and amount controls, source citations, provisional warnings, practical recommendation and distinction from business profit. The local review reproduced the substantive calculation/normalization checks. Hash checks establish preservation during the review and integrity of the archived copies, not external verification of source authenticity. The review obtained no new settlement evidence and performed no bank or external accounting checks.

## Revised handoff

Current deliverables:
- `output/cleaned-sales.csv`
- `output/cleaned-refunds.csv`
- `output/cleaned-ads.csv`
- `output/exceptions.csv`
- `output/summary.md`
- `output/change-log.md`
- `output/review-checks.txt`

All executed independent checks passed; see `review-checks.txt` for the actual check record. Remaining limitations are unchanged: obtain RF05's amount, verify RF04's paid-order/product linkage, and obtain missing business costs before assessing profit. Sources: `inputs/refunds.csv`, R05–R06; `inputs/policy.md`, **Northline Workshops accounting note — October 2026**. **Estimated human follow-up: 15–25 minutes once evidence is available**, excluding waiting time; this is an estimate, not a measured result.
