Done. The FY2025 reforecasts showed strong accuracy progression:

**Bottom line:** The 9+3 reforecast (made after 9 months actual) was nearly perfect with only $3K variance on revenue and a 0.63% MAPE. This makes sense—with 75% of the year locked, only the final quarter needed forecasting.

**Revenue story:** Each reforecast improved dramatically. Budget missed by $362K (3.2%), but by 9+3, the miss shrunk to $3K (0.02%). The biggest consistently missed month was July, which fell 15% below plan in both Budget and 6+6 versions, though 9+3 had already closed that month.

**Operating expenses:** Even tighter. All versions stayed within 1% of actuals, with 9+3 nearly perfect at $631 variance (0.02%).

**Key takeaway:** The 9+3 reforecast's excellence reflects timing, not forecasting skill—when you've actually closed 9 months, you only predict 3. Budget's 3.7% revenue MAPE against a full-year horizon is a more honest measure of forecasting quality. The progression also shows there wasn't a consistent pattern that could have improved earlier versions; if July had simply tracked Budget, the misses would have been much smaller.

Full analysis is in `/output/accuracy-note.md`.