The FY2025 reforecasts got the year's revenue total roughly right, but every version forecast revenue too high month by month. Operating expense forecasts were tighter.

The script ran cleanly. Its checks passed: each version's closed months match the P&L actuals, and the Budget version matches the P&L budget.

**Revenue**

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

MAPE is the size of the typical monthly miss. Bias is the net over- or under-forecast as a share of actuals; positive means the forecast was too high.

- **9+3 was the most accurate on both lines**, with a MAPE of 0.63% on revenue and 0.30% on opex. It only had 3 months left to forecast, so its accuracy comes mostly from the short horizon rather than from better forecasting.
- **Revenue forecasts were consistently too high.** Every version has a positive bias, and Budget's is 3.2%. This is a pattern across the year, not one bad version.
- **6+6 had one large miss in September 2025:** it forecast $1.2M against an actual of $1.0M, a miss of $151K, or 15.0%. That compares with its MAPE of 2.9%. Someone should find out what pipeline or deal assumption sat behind that month.
- **Budget hit full-year opex almost exactly, at ($649), but its monthly phasing was off.** Its opex MAPE was 4.1%, and it forecast May 2025 too high by $32K, or 15.7%. Monthly misses in opposite directions cancelled out over the year.
- **Which version to trust:** for opex, anchor on 6+6, which had a MAPE of 0.77% over 6 forecast months. For revenue, no version has earned that trust until the upward bias is fixed.

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