# Northline Workshops — October 2026 results (CAD, tax excluded)

Period: 2026-10-01 to 2026-10-31 inclusive (inputs/policy.md, line 2). Sources: inputs/sales.csv, inputs/refunds.csv, inputs/ads.csv. All figures below are produced by `output/reconcile.py` from those files. Revision 2; changes from the first submission are listed in output/CHANGELOG.md and the first submission is kept in output/original-submission/.

## 1. Headline

**Course and combined net revenue, contribution and net revenue per order are PROVISIONAL.** Refund RF05 (refunds.csv R06) settled on 2026-10-25 against paid order N1011 (sales.csv S13, Course, 360.00) but its amount is blank. Per policy.md it is treated as unknown, not zero, and is left out of the calculable refund total. Workshop figures are not affected.

| Measure | Workshop | Course (provisional) | Combined (provisional) |
|---|---:|---:|---:|
| Paid orders | 5 | 4 | 9 |
| Gross sales | 1920.00 | 900.00 | 2820.00 |
| Applied refunds (settled in October, calculable) | 240.00 | 60.00 | 300.00 |
| Net revenue | 1680.00 | 840.00 | 2520.00 |
| Advertising spend | 300.00 | 150.00 | 450.00 |
| Contribution after advertising | 1380.00 | 690.00 | 2070.00 |
| Net revenue per paid order | 336.00 | 210.00 | 280.00 |

Calculations: net revenue = gross sales − applied refunds; contribution after advertising = net revenue − ad spend (policy.md); net revenue per paid order = net revenue ÷ paid orders.

- Workshop: 1920.00 − 240.00 = 1680.00; − 300.00 = 1380.00; 1680.00 ÷ 5 = 336.00.
- Course: 900.00 − 60.00 = 840.00; − 150.00 = 690.00; 840.00 ÷ 4 = 210.00.
- Combined: 2820.00 − 300.00 = 2520.00; − 450.00 = 2070.00; 2520.00 ÷ 9 = 280.00.

**Bound on the provisional figures (not an estimate).** Policy says refunds cannot exceed paid gross, so RF05 can be at most 360.00. If it were the full 360.00: Course net revenue 480.00, Course contribution 330.00, Course net revenue per order 120.00; combined net revenue 2160.00, combined contribution 1710.00, combined net revenue per order 240.00. The table above is the upper limit of that range; the true figures are lower by whatever RF05 turns out to be.

## 2. What went into each figure

**Paid orders / gross sales** (output/cleaned-sales.csv):

| Product | Order (source row) | Date | Gross |
|---|---|---|---:|
| Workshop | N1001 (S01) | 2026-10-02 | 240.00 |
| Workshop | N1003 (S03) (replaces/collapses S04) | 2026-10-05 | 480.00 |
| Workshop | N1005 (S06) | 2026-10-10 | 240.00 |
| Workshop | N1010 (S12) | 2026-10-20 | 720.00 |
| Workshop | N1012 (S14) | 2026-10-31 | 240.00 |
| Course | N1002 (S02) | 2026-10-03 | 120.00 |
| Course | N1006 (S07) | 2026-10-11 | 240.00 |
| Course | N1008 (S10) (replaces/collapses S09) | 2026-10-15 | 180.00 |
| Course | N1011 (S13) | 2026-10-22 | 360.00 |

**Applied refunds** (output/cleaned-refunds.csv):

| Product | Refund (source row) | Order | Settled | Amount |
|---|---|---|---|---:|
| Workshop | RF01 (R01) | N1003 | 2026-10-09 | 240.00 |
| Course | RF02 (R02) (collapses R03) | N1006 | 2026-10-14 | 60.00 |

Neither applied refund exceeds the paid gross of its order (240.00 vs 480.00 on N1003; 60.00 vs 240.00 on N1006).

**Advertising spend** (output/cleaned-ads.csv):

| Product | Ad (source row) | Date | Channel | Spend |
|---|---|---|---|---:|
| Workshop | AD01 (A01) | 2026-10-03 | search | 150.00 |
| Workshop | AD03 (A03) (collapses A04) | 2026-10-10 | social | 100.00 |
| Workshop | AD05 (A06) | 2026-10-28 | search | 50.00 |
| Course | AD02 (A02) | 2026-10-04 | social | 90.00 |
| Course | AD04 (A05) | 2026-10-17 | search | 60.00 |

Spend by channel: search 260.00; social 190.00. sales.csv has no channel field, so revenue cannot be tied to a channel.

## 3. Reconciliation of included and excluded records

Every source row appears exactly once: either in a cleaned file or in output/exceptions.csv (checked by the script).

| Source file | Source rows | Included | Excluded | Sum of all stated amounts | Included amount | Excluded amount | Rows with unknown amount |
|---|---:|---:|---:|---:|---:|---:|---:|
| sales.csv | 15 | 9 | 6 | 4140.00 | 2820.00 | 1320.00 | 0 |
| refunds.csv | 7 | 2 | 5 | 960.00 | 300.00 | 660.00 | 1 |
| ads.csv | 7 | 5 | 2 | 630.00 | 450.00 | 180.00 | 0 |

Excluded rows (full explanations and policy basis in output/exceptions.csv):

| ID | Source row | Record | Reason | Amount |
|---|---|---|---|---:|
| E01 | sales.csv S04 | N1003 | exact_duplicate | 480.00 |
| E02 | sales.csv S05 | N1004 | status_pending | 120.00 |
| E03 | sales.csv S08 | N1007 | status_cancelled | 240.00 |
| E04 | sales.csv S09 | N1008 | superseded_conflicting_version | 120.00 |
| E05 | sales.csv S11 | N1009 | outside_reporting_period | 240.00 |
| E06 | sales.csv S15 | N1013 | outside_reporting_period | 120.00 |
| E07 | refunds.csv R03 | RF02 | exact_duplicate | 60.00 |
| E08 | refunds.csv R04 | RF03 | refund_pending | 240.00 |
| E09 | refunds.csv R05 | RF04 | no_valid_paid_order | 120.00 |
| E10 | refunds.csv R06 | RF05 | missing_refund_amount | unknown (blank) |
| E11 | refunds.csv R07 | RF06 | settled_outside_reporting_period | 240.00 |
| E12 | ads.csv A04 | AD03 | exact_duplicate | 100.00 |
| E13 | ads.csv A07 | AD06 | outside_reporting_period | 80.00 |

- sales.csv: 2820.00 included + [480.00 (S04) + 120.00 (S05) + 240.00 (S08) + 120.00 (S09) + 240.00 (S11) + 120.00 (S15)] = 2820.00 + 1320.00 = 4140.00.
- refunds.csv: 300.00 included + [60.00 (R03) + 240.00 (R04) + 120.00 (R05) + 240.00 (R07)] = 300.00 + 660.00 = 960.00; plus R06 with no stated amount.
- ads.csv: 450.00 included + [100.00 (A04) + 80.00 (A07)] = 450.00 + 180.00 = 630.00.

## 4. Open items that could change the numbers

1. **RF05 (refunds.csv R06) — amount missing.** Makes Course and combined net revenue provisional (section 1). Needs the settled amount from the payment record.
2. **RF04 (refunds.csv R05) — 120.00 settled 2026-10-19 against an order ID that is not in sales.csv.** Flagged and excluded as policy requires. If it turns out to belong to an October paid order, net revenue falls by 120.00; the inputs do not say which product. Source conflict: policy.md says 'every refund here relates to an October order', but this refund cannot be matched to any order in sales.csv. The conflict is not resolved here; the explicit flag-and-exclude rule in policy.md was followed.
3. **RF03 (refunds.csv R04) — 240.00 pending** on a Workshop order. Correctly not subtracted in October; it will reduce revenue in the month it settles.
4. **RF06 (refunds.csv R07) — 240.00 settled 2026-11-02** on a Workshop order placed in October. Falls into November under the settlement-date rule.

Items 3–4 do not change October's reported result, but they mean 480.00 of October Workshop gross is already refunded or awaiting refund outside the period (the pending one only if it settles). Together with the 240.00 applied, that is 720.00 of 1920.00 Workshop gross (37.5%).

## 5. Recommendation

1. **Do not move advertising budget between products on this month's data.** On the calculable figures both products return the same net revenue per ad dollar: Workshop 1680.00 ÷ 300.00 = 5.60, Course 840.00 ÷ 150.00 = 5.60. The Course figure is the upper limit; it can only fall once RF05 is known (to 3.20 if RF05 is a full refund). RF04 (120.00, order not found) could also lower either product's figure.
2. **Close the two refund gaps before circulating October as final:** get the RF05 amount and trace order N9999 for RF04. Until then, label Course and combined results provisional.
3. **Ask why refunds are happening, on both products.** Workshop has the larger known refund amount: 720.00 of 1920.00 gross (37.5%) once pending RF03 and November-settled RF06 are counted with the 240.00 applied, on three of its five October orders (N1003, N1005, N1012). Course shows 60.00 known plus RF05, which is unknown; at the 360.00 ceiling Course would be 420.00 of 900.00 (46.7%). So which product has the higher refund rate cannot be determined until RF05 is known. The inputs give no refund reasons.
4. **Fix the exports.** Each file contained a repeated row for an ID that should be unique (S04, R03, A04), and order N1008 was exported in two versions (S09, S10). Uncorrected, these would have overstated gross sales by 600.00, refunds by 60.00 and ad spend by 100.00.

**This is not profit.** Contribution after advertising is net revenue less ad spend only. No cost of delivery (COGS), payment fees, salaries or overhead were supplied (policy.md), so the 2070.00 combined figure says how much is left to cover those costs, not what the business earned. Ad spend is assigned to the product named on each ad row, as policy.md instructs; the inputs do not show which sales, if any, the ads produced.

## 6. Method and assumptions

- Cleaned files contain only the records counted in the totals, one row per order/refund/ad, keyed by the source `row_id` that was kept; `collapsed_row_ids` lists duplicate or superseded source rows folded into it. All other source rows are in exceptions.csv with raw date and raw amount preserved.
- Dates: ISO, DD/MM/YYYY and named-month forms normalised to YYYY-MM-DD (e.g. S02 `03/10/2026` → 2026-10-03, S03 `Oct 05 2026` → 2026-10-05, R02 `14/10/2026` → 2026-10-14). Money: `CAD`/`$` stripped, two decimals (S01, S02, A06). Product and channel: trimmed, case-insensitive (S01, S02, A01, A03). No unmapped products were found.
- Order of tests for sales: collapse exact duplicates → keep latest `updated_at` among conflicting rows → status must be paid → order date in October. An order's month is taken from the `date` column, not `updated_at` (assumption; they agree on month for every included order).
- Assumption: when exact duplicates are collapsed, the first row ID is kept (S03, R02, A03). The choice does not affect any total.
- Assumption: a refund's product is the product of its order in cleaned-sales.csv; refunds.csv has no product column.
- R06 (blank amount) is kept blank in exceptions.csv rather than placed in cleaned-refunds.csv, because policy.md says to exclude it from the calculable total.
- Limitation: results depend only on the three CSVs and policy.md; nothing was verified against a bank, payment processor or ad platform.
