{"id":"fpa-forecast-bias-by-owner","audience":"fpa","category":"finance-work","track":"finance","technique":"format","level":"intermediate","title":"Score each owner's forecast error and flag the ones over 10%","free":false,"task":"You're reviewing forecast accuracy by owner. Write a prompt for a markdown table with columns Owner | Forecast | Actual | Error % | Flag, one row per owner. Error % = (forecast - actual) / actual to one decimal, with a sign, so positive means over-forecast. Flag is \"Review\" if the absolute error is more than 10%, otherwise blank. Don't explain any owner's miss. No commentary. Under 80 words.","materialLabel":"Forecast vs actual by owner, in $ thousands (attached to your prompt automatically)","material":"FORECAST ACCURACY BY OWNER - Q3 2026 new business ($ thousands)\n\nOwner | Forecast | Actual\nPriti | 500 | 450\nOmar | 300 | 330\nDev | 200 | 200\n\nMethod: error % = (forecast - actual) / actual. A positive error means the forecast was too high. Review threshold: absolute error above 10%.\n\nComments: Priti says two deals slipped because the customers' legal teams were slow; Omar says his number was 'conservative on purpose'. These explanations are not part of the accuracy table.","rules":[{"type":"maxWords","max":80},{"type":"maxBullets","max":0}],"promptHints":[{"label":"Names the columns","patterns":["owner.*forecast.*actual.*error.*flag","columns"]},{"label":"Gives the formula","patterns":["forecast - actual","/ actual"]},{"label":"Defines the flag","patterns":["review","above 10%"]},{"label":"Bans explanations","patterns":["(do not|don'?t).*explain"]},{"label":"Limits output","patterns":["no commentary","only the table"]}],"mistakes":[{"title":"Dividing by forecast","why":"The AI divides by the forecast and gets +10.0%.","fix":"Say: \"Error % = (forecast - actual) / actual.\"","checks":["r1","r2"]},{"title":"Flagging by sign","why":"The AI flags Omar because his error is negative.","fix":"Add: \"Flag Review only if the absolute error is above 10%.\"","checks":["r2"]},{"title":"Adding explanations","why":"The AI includes Priti's slipped deals and Omar's 'conservative' comment.","fix":"Add: \"Do not explain any owner's miss.\"","checks":["r2"]},{"title":"Wrong columns","why":"The AI drops the Flag column or adds notes.","fix":"Say: \"Columns exactly: Owner | Forecast | Actual | Error % | Flag.\"","checks":["r0","maxBullets"]},{"title":"Trusting the AI to count words","why":"AI is bad at counting and runs past word limits.","fix":"Ask for fewer words than the real cap and keep each line short.","checks":["maxWords"]}],"hint":"Signed error and a threshold on the absolute value are easy to mix up. Give the formula, flag rule and ban on explanations.","rubricLabels":["Table has the columns Owner | Forecast | Actual | Error % | Flag with rows Priti, Omar and Dev","Error % values are +11.1% (Priti), -9.1% (Omar) and 0.0% (Dev)","Only Priti is flagged Review, and the output has no commentary or owner explanations"]}