I wrote the review to `output/accuracy-note.md`, with the underlying figures in `output/accuracy.json`. All versions tied to the P&L: every closed month carries the actual, and the Budget version matches the P&L budget.

**Answer:** The reforecasts got closer to the actuals over the year, but revenue was forecast too high in every version until 9+3.

- **Revenue:** 9+3 was the most accurate version, missing the full year by $3K, or 0.02%, with a MAPE (average monthly miss) of 0.63%. But it only had three months left to forecast, so its accuracy says little about the forecasting itself.
- **Revenue ran high:** Budget, 3+9 and 6+6 all came in high, with bias of 3.2%, 2.1% and 2.8%. The two mid-year reforecasts didn't correct this.
- **Large single misses:** Budget and 3+9 both had their biggest miss in July 2025, by $132K and $80K. 6+6 overshot September 2025 by $151K, or 15.0%, against a MAPE of 2.9%. That's one large miss rather than a trend. The forecast assumptions for July and September would explain these, but they aren't in this data.
- **Opex:** Budget opex landed within ($649) for the year, yet its MAPE was 4.1%. Monthly misses in both directions cancelled out, so the full-year figure makes it look more accurate than it was. 6+6 opex was consistently accurate month to month, with a MAPE of 0.77%.

**Recommendation:** Don't rely on the mid-year revenue reforecasts until someone explains why they kept coming in high. The 6+6 opex forecast is the most dependable model from this year.