**9+3 landed closest on revenue.** Its full-year variance was $3K, or 0.02%, and its MAPE (the average monthly miss) was 0.63%. **Revenue ran consistently high.** Budget, 3+9 and 6+6 all over-forecast, with biases of 3.2%, 2.1% and 2.8%. 9+3 nearly removed that, at 0.10%.

| Revenue | Closed months | FY variance | MAPE | Bias |
| --- | --- | --- | --- | --- |
| Budget | 0 | $362K, or 3.2% | 3.7% | 3.2% |
| 3+9 | 3 | $177K, or 1.6% | 3.3% | 2.1% |
| 6+6 | 6 | $158K, or 1.4% | 2.9% | 2.8% |
| 9+3 | 9 | $3K, or 0.02% | 0.63% | 0.10% |

- **9+3 had only 3 months left to forecast.** Its accuracy comes mostly from the short horizon, not from better forecasting. Use it for the year-end landing.
- **The planning lesson comes from the earlier versions.** Every version with 3 or more open months forecast revenue too high, so next year's revenue assumptions should be checked for that.
- **6+6 had one large miss.** It over-forecast September 2025 revenue by $151K, or 15.0%, against a MAPE of 2.9%. That is a single month, not a trend; it's worth asking what 6+6 assumed for September.
- **Opex had no consistent bias.** All versions were close on the full year. 9+3 was best again, with a MAPE of 0.30%. Budget opex looks exact at ($649) for the year, but its MAPE is 4.1%: monthly misses cancelled out, including May 2025 at $32K, or 15.7%, too high.

All figures come from the pack's accuracy script. Its checks passed: each version's closed months match the P&L actuals, and the Budget version matches the P&L budget.

Files are in `output/`:
- accuracy-note.md
- accuracy.json