{"id":"mkt2-ab-test-readout-math-table","audience":"marketing","category":"data","track":"marketing","technique":"verification","level":"intermediate","title":"Write an A/B test readout that shows the math and holds the verdict","free":false,"task":"Your landing-page A/B test is only partway done, and your VP wants a readout before the Friday meeting. Write a prompt for a table with columns Variant | Visitors | Signups | Rate, then a line starting \"Lift:\" that shows the subtraction and the relative change in percent, then a line starting \"Verdict:\" under 20 words that begins with \"Keep running\" and says why. Do not call a winner before the planned visitors per variant is reached. No intro, plain text apart from the table, under 70 words.","materialLabel":"Test results (attached to your prompt automatically)","material":"TEST: pricing page headline - results as of Thursday\nPlan: run until each variant has 10,000 visitors, then decide.\n\nVariant A (current headline: \"Plans that fit your team\")\n  Visitors so far: 4,000\n  Signups: 120\n\nVariant B (new headline: \"Start free, upgrade when you grow\")\n  Visitors so far: 4,000\n  Signups: 148\n\nRate = signups / visitors.\n\nNotes from the analyst:\n- Traffic on Wednesday included a one-day newsletter send that landed mostly on B's side by chance, about 600 extra visits.\n- Everyone on the team is excited about B and the VP wants to ship it Monday.\n- No significance test has been run.\n- Mobile and desktop haven't been split out yet.","rules":[{"type":"plainText"},{"type":"maxWords","max":70}],"mistakes":[{"title":"Declaring B the winner","why":"Higher rate plus an excited team pushes the AI to crown B.","fix":"Add: \"Do not call a winner before each variant reaches the planned 10,000 visitors.\"","checks":["r2"]},{"title":"Skipping the working","why":"The AI states a lift with no numbers, so you can't check it.","fix":"Add: \"The Lift line must show the subtraction and the relative change.\"","checks":["r1"]},{"title":"Wrong or rounded rates","why":"The AI may round or recompute rates loosely.","fix":"Add: \"Rate = signups divided by visitors, to one decimal place.\"","checks":["r0"]},{"title":"Letting markdown through","why":"AI adds bold text and headings by default.","fix":"Add: \"Plain text only, no bold, no headings.\"","checks":["plainText"]},{"title":"Trusting the AI to count words","why":"AI is bad at counting and often runs past a word limit.","fix":"Cap each part (for example \"each line under 15 words\") so the total lands well under the limit.","checks":["maxWords"]}],"promptHints":[{"label":"Names the table columns","patterns":["variant.*visitors.*signups.*rate","columns"]},{"label":"Asks for the lift math","patterns":["lift","show the (math|subtraction)","relative"]},{"label":"Holds the verdict","patterns":["do not call","don'?t call","planned","10,000","keep running"]},{"label":"Caps the verdict length","patterns":["under 20 words","verdict.{0,30}\\d+ words"]},{"label":"Asks for no intro","patterns":["no intro","no preamble","only the table"]}],"hint":"A readout is numbers first, then a careful verdict. Tell the AI the plan, make it show the lift math, and forbid declaring a winner early.","mission":"Give the VP an honest readout of a half-finished A/B test.","musts":["Show lift subtraction","Verdict starts Keep running","No winner yet"],"rubricLabels":["The table shows A with 4,000 visitors, 120 signups and 3.0%, and B with 4,000 visitors, 148 signups and 3.7%","A line starting \"Lift:\" shows the subtraction 3.7% minus 3.0% as 0.7 points and a relative change of about 23%","A line starting \"Verdict:\" is under 20 words, starts with \"Keep running\", and does not call B the winner"]}