DATA SCIENCE FORECAST EVALUATION Space Office | https://www.spaceoffice.ai/blog/data-science Decision and human owner: Operating period and time zone: Prediction cutoff: Outcome and source of actual: Horizon: Available inputs at cutoff: Missing-data policy: Fictional pickup orders. Copy the two columns into a spreadsheet with Week in A1 and Orders in B1. Week,Orders 1,38 2,42 3,40 4,44 5,46 6,42 7,50 8,47 In row 6, the target is week 5. Use these formulas and fill through row 9. C6 (last-week estimate) =B5 D6 (trailing mean) =AVERAGE(B3:B5) E6 (last-week absolute error) =ABS(B6-C6) F6 (trailing mean absolute error) =ABS(B6-D6) Checks Average E6:E9 should be 4.25 orders. Average F6:F9 should be about 3.08 orders. Week 9 estimates are 47 and about 46.33. Keep full precision in evaluation. Inspect shortages separately from leftovers before using a preparation rule. Trial record Cutoff | Forecast | Existing plan | Actual | Prepared | Override reason Who can override: Fallback: Human decision and date: