# Change log — revision 2 (self-review of the first submission)

The first submission is preserved unchanged in `output/original-submission/` (summary.md, four CSVs, reconcile.py).

## Result of the review

No figure changed. The four data files (cleaned-sales.csv, cleaned-refunds.csv, cleaned-ads.csv, exceptions.csv)
are byte-identical to the first submission (`verify.py` step 5). All corrections are to wording in summary.md and to
what is claimed about checks.

## Corrections

| # | Where | First submission | Problem | Revision 2 |
|---|---|---|---|---|
| 1 | summary.md §5, recommendation 3 | "Workshop carries the larger refund exposure" | Not supported. Workshop's known refunds are 720.00 of 1920.00 (37.5%, including pending RF03 and November-settled RF06). Course is 60.00 plus RF05, whose amount is unknown; at the 360.00 ceiling Course would be 420.00 of 900.00 (46.7%). | States the known Workshop amount, the Course ceiling, and that the higher refund rate cannot be determined until RF05 is known. |
| 2 | summary.md §4, item 2 (RF04 / refunds.csv R05) | Flagged the unmatched order N9999 only | Did not point out that this conflicts with policy.md's statement "every refund here relates to an October order". | Conflict stated; noted as unresolved; flag-and-exclude rule followed as before. |
| 3 | summary.md §5, "This is not profit" paragraph | "Sales are also attributed to ads only at product level, as policy.md instructs" | Misstates the policy, which assigns ad spend to products ("October ads are all attributable to the product named in each row"); it says nothing about attributing sales to ads. | Reworded to say ad spend is assigned by product and the inputs do not show which sales the ads produced. |
| 4 | summary.md §5, recommendation 1 | Only RF05 named as able to lower net revenue per ad dollar | RF04 (120.00) could also lower either product if it is traced to an October paid order. | Sentence added. |
| 5 | reconcile.py comment | "raw inputs untouched: re-read and compare row counts" | Overstated: the assertion only checks expected row counts. | Comment corrected. A real comparison of the three input CSVs against the listing in the assignment prompt was added in `verify.py` step 4 and passed. |
| 6 | Verification script | Lived in work/, not delivered | The first handoff cited a check whose script was not in output/. | Extended and delivered as `output/verify.py`. |
| 7 | summary.md header | — | — | Revision marker pointing to this file. |

## Checks run for this revision (all by `output/verify.py`, exit code 0)

1. Headline figures recomputed from the cleaned CSVs alone and matched to the summary.md table.
2. Cleaned files: unique order/refund/ad IDs; sales all paid and dated in October; refunds all settled in October,
   matched to an included paid order and not above its gross; ads dated in October; dates YYYY-MM-DD; money two
   decimals; products only Workshop/Course. R06 amount blank in exceptions.csv.
3. Every source row_id is in exactly one of a cleaned file or exceptions.csv (15 = 9 + 6; 7 = 2 + 5; 7 = 5 + 2).
4. inputs/sales.csv, refunds.csv, ads.csv are identical to the listing in the assignment prompt (ignoring a
   trailing newline).
5. Data files identical to the first submission; summary.md differs in 5 lines (the corrections above).

## Not done

- No check against a bank, payment processor or ad platform; none is available in the inputs.
- RF05's amount and RF04's order remain unresolved; Course and combined results remain provisional.
- summary.md was checked as Markdown text only, not in a rendered view.
