# Northline Workshops — October 2026 reconciliation

## Basis and conclusion

All amounts below are CAD, tax excluded, for October 1–31, 2026 inclusive. Paid order gross less eligible October-settled refunds is net revenue; net revenue less October ad spend is contribution after advertising. These are the exercise's definitions, not a cash-flow statement. Source: `inputs/policy.md`, **Northline Workshops accounting note — October 2026**.

**Calculable combined net revenue is 2,520.00 and contribution after advertising is 2,070.00. Both are provisional:** settled refund RF05 has no amount (`inputs/refunds.csv`, R06). No zero has been substituted. RF04 cannot be applied because its order N9999 is absent (`inputs/refunds.csv`, R05; `inputs/sales.csv`, S01–S15).

## Results

| Metric | Workshop | Course | Combined |
|---|---:|---:|---:|
| Paid order count | 5 | 4 | 9 |
| Gross sales | 1,920.00 | 900.00 | 2,820.00 |
| Applied refunds (deduction) | 240.00 | 60.00 | 300.00 |
| Net revenue | 1,680.00 | 840.00* | 2,520.00* |
| Advertising spend | 300.00 | 150.00 | 450.00 |
| Contribution after advertising | 1,380.00 | 690.00* | 2,070.00* |
| Net revenue per paid order | 336.00 | 210.00* | 280.00* |

\* Provisional because R06's missing amount belongs to Course order N1011. Workshop is calculable without that missing amount. All results remain subject to investigating the excluded unmatched refund R05.

Calculation evidence:
- Workshop sales: `inputs/sales.csv`, S01, S03, S06, S12, S14; Course sales: S02, S07, S10, S13. Latest S10 replaces S09 for N1008.
- Applied refunds: `inputs/refunds.csv`, R01 (Workshop, 240.00) and R02 (Course, 60.00), linked through N1003 and N1006 respectively.
- Workshop ads: `inputs/ads.csv`, A01, A03, A06 (150.00 + 100.00 + 50.00); Course ads: A02, A05 (90.00 + 60.00).
- Net revenue = gross − applied refunds. Contribution = net revenue − ads. Net revenue per paid order = net revenue / paid order count; fully refunded orders would still be paid orders. Combined per-order revenue uses 9 orders, not an unweighted average of the product averages. Definitions follow `inputs/policy.md`, **Northline Workshops accounting note — October 2026**.

If R06 is resolved to amount x with no other changes, Course net revenue becomes 840.00 − x and contribution 690.00 − x; combined net revenue becomes 2,520.00 − x and contribution 2,070.00 − x. Course and combined per-order revenue become (840.00 − x)/4 and (2,520.00 − x)/9, respectively. This is a sensitivity expression, not an estimate of x.

## Record and amount reconciliation

Cleaned files retain **every original row and source row_id**, including excluded records, so they are audit tables rather than included-only transaction extracts. Filter `included_in_october=yes` to calculate the results. Every excluded row has a matching entry in `exceptions.csv` with a reason and policy citation.

| Source | Raw rows | Included | Excluded | Raw known amount | Excluded known amount | Included amount |
|---|---:|---:|---:|---:|---:|---:|
| `inputs/sales.csv`, S01–S15 | 15 | 9 | 6 | 4,140.00 | 1,320.00 | 2,820.00 |
| `inputs/refunds.csv`, R01–R07 | 7 | 2 | 5 | 960.00 + unknown | 660.00 + unknown | 300.00 |
| `inputs/ads.csv`, A01–A07 | 7 | 5 | 2 | 630.00 | 180.00 | 450.00 |
| Total records | 29 | 16 | 13 | — | — | — |

The raw amounts above include duplicate and otherwise ineligible rows; they are only control totals, not business results. Refund unknown refers solely to R06, not a zero amount. Amounts across the three sources are not additive measures of performance.

### Excluded sales
Source: `inputs/sales.csv`.
- S04: duplicate of S03, 480.00.
- S05: pending, 120.00; S08: cancelled, 240.00.
- S09: older conflicting N1008 record, 120.00. Retain S10's 180.00 because its `updated_at` is later.
- S11: September order, 240.00; S15: November order, 120.00.

### Excluded refunds
Source: `inputs/refunds.csv`.
- R03: duplicate of R02, 60.00.
- R04: pending, 240.00; not an October deduction despite its populated settlement-date field.
- R05: settled but no matching paid order for N9999, 120.00; investigate, do not assign a product.
- R06: settled against paid N1011, amount missing; retain blank and flag provisional results.
- R07: 240.00 settled November 2, so not an October deduction even though N1012 was paid in October.

Source inconsistency handled explicitly: the policy says every refund relates to an October order, but R05 names N9999, which does not appear in `inputs/sales.csv`, S01–S15. That general statement does not supply a valid paid order or product. The same policy explicitly requires a valid paid order before applying a settled refund, so R05 remains excluded and unassigned rather than being forced into either product (`inputs/policy.md`, **Northline Workshops accounting note — October 2026**).

### Excluded ads
Source: `inputs/ads.csv`.
- A04: duplicate of A03, 100.00.
- A07: November ad, 80.00.

## Cleaning decisions and assumptions

Rules follow `inputs/policy.md`, **Northline Workshops accounting note — October 2026**:
- Treat row_id as provenance, not a business field when detecting otherwise identical duplicates. Retain the first source occurrence for exact duplicates (S03, R02, A03); retain the latest update for conflicting order N1008 (S10). The first-occurrence tie-break is an explicit implementation choice; identical business values mean it does not change totals.
- Normalize date fields to YYYY-MM-DD, including DD/MM/YYYY and named English month dates. Preserve `updated_at` as an ISO timestamp to retain the conflict-resolution evidence; it is not an order-date field.
- Normalize money to two decimal places and strip currency markers. Keep R06's unresolved amount empty. Normalize products to Workshop/Course and ad channels to lowercase without surrounding whitespace. All supplied sales/ad products map successfully.
- Refund product and `linked_sales_row_id` are derived from the included paid order, not inferred from refund amount. R05's linkage and product remain blank because its order is absent.
- A blank exclusion reason means included; `retained_row_id` is populated only for duplicates/superseded records. Original business statuses are retained, so pending/cancelled remain visible.
- Known unique settled refund amounts linked to paid orders do not exceed their respective paid gross, including November R07. R06's amount cannot be cap-checked, and R05 cannot be validated against order gross. No missing amount or missing order was reconstructed.

## Recommended action

First obtain RF05's settlement amount and resolve RF04's missing order linkage before closing October. Then rerun the calculations, checking aggregate refunds against each order's gross. Keep the November RF06 settlement out of October under the stated policy. Evidence: `inputs/refunds.csv`, R05–R07; `inputs/policy.md`, **Northline Workshops accounting note — October 2026**.

Both products have positive calculable contribution after advertising, but do not increase ad budgets solely on these totals. Workshop currently contributes 1,380.00 versus provisional Course contribution of 690.00 using the records cited above; that difference alone does not establish incremental ad effectiveness. Gather delivery costs, payment fees, salaries and overhead before judging profitability or scaling. **Contribution after advertising is not business profit**: none of those costs is supplied (same policy heading).

## Checks actually performed and handoff

A local Python standard-library reconciliation used decimal arithmetic and successfully checked:
- Product and combined totals against the table above.
- All 29 source row IDs preserved; 9/2/5 included sales/refund/ad rows and 6/5/2 excluded rows.
- One exception entry per excluded row (13 total).
- Output CSV date and money formatting, and preservation of R06's empty amount.
- Known linked, unique settled refunds against paid gross; unresolved R05/R06 cannot be fully checked.
- Input-file SHA-256 hashes unchanged across processing.

Deliverables: `output/cleaned-sales.csv`, `output/cleaned-refunds.csv`, `output/cleaned-ads.csv`, `output/exceptions.csv`, `output/summary.md`.

### Review of the initial submission

The second-pass review found no numerical or CSV correction necessary. The four CSVs remain byte-identical to the initial submission. The revised narrative makes the R05 policy/source inconsistency explicit. An independent validator derives duplicate, superseded and eligible records from the raw CSVs rather than using the initial script's hard-coded exclusion lists, checks every original field and derived inclusion/linkage field, matches exceptions one-to-one, and compares all 21 summary results to computed totals. See `output/review-checks.md` for executed checks and precise scope. Re-run with `python output/validate.py` from the workspace root; it reads inputs and deliverables and writes only the review-checks report. This is a validator for the supplied fixture, not a production accounting application.

Original deliverables are preserved separately in `output/original-submission/` with `sha256.json`. Review additions are `output/validate.py`, `output/review-checks.md` and `output/review-change-log.md`. No missing settlement information has been resolved by this review.

Remaining limitations: no evidence supplies R06's amount, R05's missing order, or the costs needed for profit. No external settlement or order verification was performed. Estimated human completion effort: **15–30 minutes once the missing records are available**, to resolve the two exceptions, rerun and approve; obtaining those records may take additional unknown time. This is an estimate, not a measured duration.
