{"id":"data2-metric-drop-investigation-order-numbered","audience":"managers","category":"data","track":"data","technique":"constraints","level":"intermediate","title":"Order four hypotheses for a conversion drop by cheapest check","free":false,"task":"Conversion fell on Tuesday and your team has four suspects. Write exactly 4 numbered lines in the form '1. Hypothesis - check - time', ordered from the quickest check to the slowest, plain text, under 100 words. Each check must say what data to look at, using only what the notes mention.","materialLabel":"Notes from the incident channel (attached to your prompt automatically)","material":"Conversion drop - Tuesday Nov 4 (Maple Marketplace)\nConversion went from 4.8% (typical) to 4.1% on Tuesday.\n\nFour possible causes people raised, with how long each check would take:\n\nA) Tracking broke (events not firing). Check: compare checkout events against orders in the payments system, and look at the SDK release log. Time: 10 minutes.\n\nB) Traffic mix changed (more low-converting visitors). Check: break Tuesday's visits by source and device and compare shares with last Tuesday. Time: 30 minutes.\n\nC) A site release introduced a bug. Check: review the release notes and error rates by page, and have engineering confirm. Time: 2 hours.\n\nD) Pricing or promotion change. Check: the marketing calendar plus a cohort analysis of discount usage. Time: 1 day.\n\nOther facts:\n- The payments system shows orders down 12% versus a typical Tuesday.\n- Nobody has looked at any of these yet.\n- The VP wants to know where to start; she said \"start with whatever we can rule out fastest\".","rules":[{"type":"numberedItems","count":4},{"type":"plainText"},{"type":"maxWords","max":100}],"mistakes":[{"title":"Ordering by likelihood","why":"The AI puts the 'most likely' cause first, such as a release bug.","fix":"Add: \"Order by time to check, quickest first.\"","checks":["r0"]},{"title":"Vague checks","why":"The AI says 'look into traffic' with no data named.","fix":"Add: \"Each line names the data to look at, using only the notes.\"","checks":["r1"]},{"title":"Declaring a cause","why":"The AI concludes the tracking is broken or the release is to blame.","fix":"Add: \"Don't say which one is the real cause.\"","checks":["r2"]},{"title":"Leaving the format to chance","why":"The AI picks its own length and layout unless told the exact shape.","fix":"Add: \"Exactly 4 numbered lines. Plain text, no bold or headings.\"","checks":["numberedItems","plainText"]},{"title":"Trusting the AI to count words","why":"AI is bad at counting. Ask for a total word limit and it often lands over it.","fix":"Cap each part instead, for example \"each line under 15 words.\"","checks":["maxWords"]}],"promptHints":[{"label":"Asks for four lines","patterns":["4 numbered","exactly 4","start .{0,15}1\\."]},{"label":"Sorts by time","patterns":["quickest","fastest","shortest","time"]},{"label":"Names the data","patterns":["use only","what to look at","data"]},{"label":"Includes the times","patterns":["10 minutes","1 day","2 hours"]},{"label":"Rules out a verdict","patterns":["don.t say","real cause"]}],"hint":"Triage order is a sort. Give the AI the sort key (time) and the line shape.","mission":"Where should we start with the conversion drop?","musts":["Four numbered lines","Quickest check first","Data named per check"],"rubricLabels":["The lines are ordered by check time: tracking (10 minutes) first, traffic mix (30 minutes) second, site release bug (2 hours) third, and pricing or promotion (1 day) fourth","Each line names the data to look at from the notes (for example checkout events versus payments orders and the SDK log; visits by source and device; release notes and error rates; marketing calendar and discount usage)","Each line ends with its time (10 minutes, 30 minutes, 2 hours, 1 day), and there are exactly 4 numbered lines with no intro or recommendation of which is the true cause"]}