# Northline Workshops — October 2026 results (CAD)

Reporting period: 2026-10-01 to 2026-10-31 inclusive. All amounts CAD, tax excluded.
Rules applied from `inputs/policy.md` (accounting note, October 2026). Sources:
`inputs/sales.csv`, `inputs/refunds.csv`, `inputs/ads.csv`. Generated by
`reconcile.py` (standard-library Python); every figure below is computed from the
cleaned files, which preserve source row IDs.

**Status: PROVISIONAL.** Refund RF05 (row R06, order N1011) is settled in October
but its amount is missing. It is excluded from the calculable total and never
treated as zero, so Course and combined net revenue, contribution and per-order
figures below are upper bounds pending that amount (Workshop figures are
unaffected by RF05) (`inputs/policy.md`: "Missing refund amounts
mean the resulting net revenue is provisional"). Refund RF04 (row R05) also
references order N9999, which does not exist in the sales file — flagged, not guessed.

## Results

| Metric basis | Paid orders | Gross sales | Applied refunds | Net revenue | Ad spend | Contribution after ads | Net revenue / paid order |
|---|---|---|---|---|---|---|---|
| Workshop | 5 | 1920.00 | 240.00 | 1680.00 | 300.00 | 1380.00 | 336.00 |
| Course | 4 | 900.00 | 60.00 | 840.00 | 150.00 | 690.00 | 210.00 |
| Combined | 9 | 2820.00 | 300.00 | 2520.00 | 450.00 | 2070.00 | 280.00 |

Definitions (per `inputs/policy.md`): revenue = paid order gross; applied refunds =
settled refunds with October settlement dates against valid paid October orders;
contribution after advertising = net revenue − ad spend. **This is not net profit**:
no COGS, payment fees, salaries or overhead were supplied.

Included records:
- Sales (paid, October): S01, S02, S03, S06, S07, S10, S12, S13, S14.
- Refunds (settled, October, valid paid order, amount present): R01, R02
  (RF01 → N1003 Workshop 240.00; RF02 → N1006 Course 60.00).
- Ads (October, mapped product): A01, A02, A03, A05, A06.

Checks performed on applied refunds: RF01 240.00 ≤ N1003 paid gross 480.00;
RF02 60.00 ≤ N1006 paid gross 240.00. No refund exceeds paid gross. RF05's
ceiling cannot be checked because its amount is missing (N1011 gross is 360.00).

## Reconciliation of source rows

| File | Source rows | Included | Exact duplicates collapsed | Superseded (conflict) | Excluded after dedup |
|---|---|---|---|---|---|
| sales.csv | 15 | 9 | 1 | 1 | 4 |
| refunds.csv | 7 | 2 | 1 | 0 | 4 |
| ads.csv | 7 | 5 | 1 | 0 | 1 |

Every source row is accounted for: included + collapsed duplicate + superseded
+ excluded = source rows (asserted in `reconcile.py`). Row-level reasons are in
`output/exceptions.csv`. Key items:

- **S04** duplicates S03 (order N1003) exactly — collapsed, counted once (480.00).
- **S09 vs S10** (order N1008): conflicting gross 120.00 vs 180.00; kept S10
  (updated_at 2026-10-16T09:00:00, latest) per policy; S09 excluded.
- **S05** pending and **S08** cancelled — not revenue per policy.
- **S11** (2026-09-30) and **S15** (2026-11-01) — outside October.
- **R03** duplicates R02 (refund RF02) exactly — collapsed, deducted once (60.00).
- **R04** pending — not deducted per policy.
- **R05** settled 120.00 but order N9999 is not a valid paid order — flagged and excluded.
- **R06** settled, amount missing — excluded from calculable total; makes results provisional.
- **R07** settled 2026-11-02 — November settlement, outside the October window.
- **A04** duplicates A03 (ad AD03) exactly — collapsed, spent counted once (100.00).
- **A07** dated 2026-11-01 — outside October.

## Normalization applied

- Dates: slash dates read as DD/MM/YYYY (S02 03/10/2026 → 2026-10-03; S07
  11/10/2026 → 2026-10-11; R02 14/10/2026 → 2026-10-14); named month S03/S04
  "Oct 05 2026" → 2026-10-05; ISO kept. All valid dates written YYYY-MM-DD.
- Money: "CAD 240.00", "$120.00", "$50.00" parsed to plain two-decimal values.
  The missing amount in R06 is preserved as empty, not zero.
- Product/channel: trimmed and case-folded (" Workshop " → Workshop, "course" →
  Course, " Search "/"SOCIAL" → search/social). No unmapped products occurred.

## Recommendation

Workshop produced 1380.00 of the 2070.00
combined contribution (66.7%)
on 300.00 of ad spend, and its net revenue per paid order
(336.00) is well above Course (210.00).
Notably, contribution per ad dollar is identical for both lines
(Workshop 4.60, Course
4.60), so October data alone does not
justify shifting budget from Course to Workshop on efficiency grounds. Practical steps:

1. **Resolve the two refund exceptions before booking the month**: obtain RF05's
   amount (if the refund cannot exceed N1011's 360.00 paid gross, Course net
   revenue is bounded 480.00–840.00,
   combined net revenue 2160.00–2520.00,
   and combined contribution 1710.00–2070.00)
   and investigate RF04's orphan order N9999 — it may indicate a missing sales record.
2. **Prioritize Workshop capacity/fill rate**: it carries most contribution and the
   higher per-order value; modest spend increases there are reasonable to test, but
   with equal ad efficiency, run a small controlled test rather than a large reallocation.
3. **Do not treat 2070.00 as profit.** Cost of delivery,
   payment fees, salaries and overhead are absent from the inputs; true profitability
   could be materially lower, especially for delivery-heavy workshops. Gather those
   costs before pricing or budget decisions.

## Assumptions and limitations

- "Exact duplicate" = identical on all fields except row_id (policy allows collapsing).
- Refund RF05 (R06) is attributed to Course via order N1011 for the provisional-range
  note only; it is excluded from the calculable totals and its unknown amount is
  not treated as zero.
- Ads are attributed to the product named in each row (per policy); no allocation
  across products was needed.
- No refunds exceeded paid gross among calculable records; the missing-amount refund
  could not be ceiling-checked.
- Figures are provisional solely due to R06's missing amount; all other exclusions
  are policy-driven and final.

Estimated human finish time (not a measured result): ~20–30 minutes to obtain the
RF05 amount, investigate order N9999, and re-run `reconcile.py`.
