The write-up is in `output/accuracy-note.md`, and the script's figures are in `output/accuracy.json`. The script ran cleanly. Closed months tie to the P&L, and the Budget version matches the P&L budget.

**Revenue**
- The 9+3 reforecast was the most accurate. It ended the year $3K, or 0.02%, from actual, with a MAPE of 0.63%.
- Budget forecast revenue too high by $362K, or 3.2%. Its MAPE was 3.7%.
- The 3+9 and 6+6 reforecasts were still too high at full year, by 1.6% and 1.4%. Their MAPEs were 3.3% and 2.9%, so the monthly misses stayed large.
- 6+6 missed 2025-09 by $151K, or 15.0%. Budget missed 2025-07 by $132K, or 15.0%. The data doesn't show why, so the note flags both months for a look at the assumptions behind them.

**Operating expenses**
- Budget landed within ($649) for the year, but its MAPE was 4.1%. Large monthly misses cancelled out, so the full-year figure says little about monthly control.
- 6+6 had a MAPE of 0.77%, and 9+3 had 0.30%.

**Caveat on 9+3:** it only had three months left to forecast. Its low MAPE reflects the short horizon, so it doesn't show that the forecasting improved. It is the version to anchor on for the year-end landing, not proof of method.

The note also records the input files, row counts and hashes.