All deliverables are written and verified. Summary of results and handoff:

## October 2026 results (CAD, provisional)

| | Paid orders | Gross | Applied refunds | Net revenue | Ad spend | Contribution after ads | Net rev/order |
|---|---|---|---|---|---|---|---|
| Workshop | 5 | 1,920.00 | 240.00 | 1,680.00 | 300.00 | 1,380.00 | 336.00 |
| Course | 4 | 900.00 | 60.00 | 840.00 | 150.00 | 690.00 | 210.00 |
| **Combined** | **9** | **2,820.00** | **300.00** | **2,520.00** | **450.00** | **2,070.00** | **280.00** |

**Provisional** because refund RF05 (refunds.csv R06, order N1011) is settled in October with a missing amount — preserved as empty, never zeroed, per inputs/policy.md. Key decisions: S04/R03/A04 exact duplicates collapsed; S09 superseded by S10 (latest updated_at, gross 180.00); pending/cancelled orders and pending refunds not counted; S11, S15, R07, A07 outside Oct 1–31; R05 flagged (order N9999 doesn't exist). Recommendation in summary.md: fix the two refund exceptions before closing the month; Workshop drives 66.7% of contribution, but contribution per ad dollar is identical (4.60) for both lines, so test spend shifts rather than reallocate blindly; 2,070.00 is contribution after ads only — COGS, fees, salaries and overhead are absent, so it is not profit.

## Handoff

**Output files** (raw inputs untouched):
- `output/cleaned-sales.csv` (13 rows), `output/cleaned-refunds.csv` (6), `output/cleaned-ads.csv` (6) — normalized dates/money, source row IDs, included flag + reason
- `output/exceptions.csv` — 13 row-level exceptions with policy citations
- `output/summary.md` — results, reconciliation, normalization notes, recommendation, assumptions
- `output/reconcile.py` — reproducible generator (stdlib only)

**Checks actually performed** (via `reconcile.py` assertions and a separate `verify.py`, both run locally, all passed):
- Row conservation: 15 sales = 9 included + 1 duplicate + 1 superseded + 4 excluded; 7 refunds = 2 + 1 + 0 + 4; 7 ads = 5 + 1 + 0 + 1
- All dates YYYY-MM-DD, all money two-decimal; R06 amount preserved empty
- Totals independently recomputed from the written cleaned CSVs and matched against hand-derived expectations
- Applied refunds ≤ paid gross per order (RF01 240 ≤ 480; RF02 60 ≤ 240)
- Spot-checks of dedup, conflict resolution, DD/MM parsing (S02, S07, R02), currency parsing (A06), and inclusion/exclusion flags

**Limitations:** results are provisional pending RF05's amount (Course net revenue bounded 480.00–840.00 given the paid-gross ceiling); RF04's orphan order N9999 may indicate a missing sales record — not resolvable from supplied inputs; RF05's ceiling could not be checked; no profitability costs supplied. No site/ inputs exist in this workspace, so only Task 1 was applicable.

**Estimated human finish time:** ~20–30 minutes (obtain RF05 amount, investigate N9999, re-run `reconcile.py`) — an estimate, not a measured result.