I scored all four FY2025 versions (Budget, 3+9, 6+6, 9+3) against the actual P&L. The 9+3 outlook was the most accurate on both revenue and operating expenses. Every closed month ties to the P&L, and the Budget version matches the P&L budget.

**Revenue**

| Version | Full-year miss | Typical monthly miss (MAPE) |
|---|---|---|
| Budget | $362K high, 3.2% | 3.7% |
| 3+9 | $177K high, 1.6% | 3.3% |
| 6+6 | $158K high, 1.4% | 2.9% |
| 9+3 | $3K high, 0.02% | 0.63% |

- **Early outlooks ran high:** Budget, 3+9 and 6+6 all overstated revenue. The 6+6 version still had a bias of 2.8%, and it missed September by $151K, or 15.0%.
- **Budget's worst month:** July was forecast at $1,009K against an actual of $878K.
- **Operating expenses:** the 9+3 version again had the lowest MAPE, at 0.30%.
- **Budget on operating expenses:** it landed within ($1K) for the year, but its monthly MAPE was 4.1%. Misses in different months cancelled out, so that full-year figure flatters the budget.

**Caveat for the CFO:** 9+3 had nine closed months and only three left to forecast. Its accuracy reflects a short horizon as much as better forecasting, and the data can't separate the two. The data also doesn't say why the early outlooks ran high. What the 6+6 version assumed for September would answer that.

The full tables, findings and data notes are in `output/accuracy-note.md`, and the underlying figures are in `output/accuracy.json`.