Forecast Error Metrics

You cannot improve forecast accuracy without measuring it precisely. Forecast error has multiple dimensions: magnitude (how far off was the forecast), direction (was it too high or too low), and timing (was the error consistent across the quarter or concentrated at the end). Each dimension requires a different metric, and together they produce a complete picture of forecast quality.

Forecast Accuracy Metrics

Metric                     Formula                           Target
─────────────────────────  ──────────────────────────────    ──────
Absolute Error             |Forecast - Actual|               < 10%
Percentage Error           (Forecast - Actual) / Actual      +/- 8%
MAPE (trailing 4 qtrs)    Avg of |% Error| across quarters  < 12%
Directional Bias           Avg of signed % Error             < +/- 3%
Week-over-Week Stability   StdDev of weekly forecast changes < 5%

Example:
  Q4 Forecast: $1,100K | Actual: $980K
  Absolute Error: $120K
  Percentage Error: +12.2% (over-forecasted)

  Trailing MAPE: Q1: 8%, Q2: 6%, Q3: 15%, Q4: 12% → MAPE = 10.3%
  Directional Bias: Q1: +8%, Q2: +6%, Q3: -15%, Q4: +12% → Bias = +2.8%
  (Positive bias = systematic over-forecasting)