I've written the accuracy review to `output/accuracy-note.md` and the script's figures to `output/accuracy.json`. The script confirmed that every closed month ties to the P&L and that Budget equals the P&L budget.

- **Revenue:** The 9+3 reforecast was the most accurate. Its full-year variance was $3K, or 0.02%, and its MAPE was 0.63%.
- **Early versions:** Budget, 3+9 and 6+6 all over-forecast revenue. Their full-year variances were $362K, $177K and $158K, and their biases were 3.2%, 2.1% and 2.8%. The first two reforecasts did not correct the over-forecast.
- **Single large misses:** Budget's worst revenue month was 2025-07, off by 15.0%. 6+6's worst was 2025-09, also off by 15.0%, against a MAPE of 2.9%. The note asks what each forecast assumed for those months.
- **Operating expenses:** 9+3 again had the lowest MAPE, at 0.30%. Budget landed within ($649) for the year, but its MAPE was 4.1%, so its monthly misses cancelled out.
- **Why 9+3 looks best:** It had only three months left to forecast. The data can't separate a shorter horizon from better forecasting, and the note says so.

The note's figures are copied from the script output, and the data notes list both input files with their row counts and hashes.