Done. Here's your forecast accuracy summary for FY2025:

**The 9+3 forecast landed closest**, missing revenue by just $3K on an $11.2M base (0.02%). That version, made after 9 months of actual results, had only Q4 to predict and nailed it with a 0.63% MAPE.

**You ran consistently high on revenue.** All four versions overstated revenue, with Budget the worst at 3.2% above actual. This bias softened as the year progressed — the 3+9 version (made at Q1 end) still overran by 2.1%, but 9+3 was essentially flat. The pattern points to early-year optimism that reality corrected.

**Operating expense forecasts were tight.** Budget missed by less than $1K on $2.5M, and 9+3 overran by only $631. The 6+6 version achieved a 0.77% MAPE on expenses, showing this line was predictable from mid-year forward.

**July revenue was the biggest single miss** — the Budget forecast $1.0M but came in at $878K, a $132K or 15% shortfall. This one-month variance drove much of the year's high bias.

Full tables and data sources are in `output/accuracy-note.md`.