{"id":"acct-variance-lines-over-threshold","audience":"finance","category":"finance-work","track":"finance","technique":"constraints","level":"intermediate","title":"Explain only the variances that cross the threshold","free":false,"task":"Your controller only wants October variances that are over both $5,000 and 10% of budget. Write a prompt for exactly those, as numbered plain-text lines formatted \"1. Line: actual - budget = $X (Y%) - reason\", using the department notes for reasons. If a reason suggests a cost might be booked in the wrong period, flag it for the controller; don't call it an error. Under 120 words.","materialLabel":"October budget vs actual and notes (attached to your prompt automatically)","material":"October 2026 budget vs actual - Saltmarsh Software (department P&L)\n\nLine | Budget | Actual\nRevenue | $420,000 | $451,000\nSalaries | $180,000 | $184,000\nContractors | $20,000 | $31,000\nSoftware | $12,000 | $11,400\nTravel | $8,000 | $2,500\nMarketing | $40,000 | $52,000\nRent | $15,000 | $15,000\n\nDepartment notes:\n- Revenue: two renewals closed early. Sales VP very happy about it.\n- Contractors: two contractors covering open engineering roles since Oct 5.\n- Travel: the Denver trade show was cancelled by the organizer.\n- Marketing: Q1 2027 campaign invoice ($12,000) was paid early in October and coded to October marketing expense.\n- Salaries: one mid-month hire.\n- Software: a seat count dropped.","rules":[{"type":"numberedItems","count":3},{"type":"plainText"},{"type":"maxWords","max":120}],"promptHints":[{"label":"States both thresholds","patterns":["5,?000","10\\s*%","both"]},{"label":"Shows the line format","patterns":["actual - budget","1\\.","numbered"]},{"label":"Asks for math","patterns":["math","percent","%"]},{"label":"Flags timing","patterns":["flag","period","prepaid","controller"]},{"label":"Sets length","patterns":["\\d+\\s*words","plain"]}],"mistakes":[{"title":"Explaining every line","why":"The AI writes about all seven lines, starting with the biggest dollar change.","fix":"Say: \"Only lines over both $5,000 and 10% of budget.\"","checks":["r0","numberedItems"]},{"title":"Letting dollars win","why":"Revenue's $31,000 looks big, so the AI includes it though it's about 7%.","fix":"Add: \"Both tests must pass. Check the percent, not just the dollars.\"","checks":["r0"]},{"title":"Reasons without math","why":"The AI states the variance in words and skips the numbers.","fix":"Show the line: \"1. Line: actual - budget = $X (Y%) - reason.\"","checks":["r1"]},{"title":"Missing the timing issue","why":"The AI reports the marketing reason as is, without noticing it's next year's campaign.","fix":"Ask: \"If a cost may belong to another period, flag it for the controller.\"","checks":["r2"]},{"title":"Trusting the AI to count words","why":"Variance write-ups get headings and summaries that break the limit.","fix":"Say: \"Each line under 30 words, plain text, no intro.\"","checks":["maxWords","plainText"]}],"hint":"A big dollar change isn't always a big percent. Give the AI both tests, the exact line shape, and tell it to flag what looks like the wrong period.","rubricLabels":["Covers exactly Contractors, Travel and Marketing, and leaves out Revenue (+$31,000 is only about 7% of budget) and the other small variances","Each line shows the math: Contractors $31,000 - $20,000 = $11,000 (55%), Travel $2,500 - $8,000 = -$5,500 (about -69%), Marketing $52,000 - $40,000 = $12,000 (30%)","The Marketing line flags that the early-paid Q1 2027 campaign invoice may belong in a later period for the controller to review, without calling it an error"]}