{"id":"mkt-x-survey-stats-for-sales-deck","track":"marketing","audience":"marketing","category":"data","technique":"verification","title":"Pull deck-ready stats from a survey without overstating them","free":false,"task":"You're pulling stats from Hearthstone Insurance's homeowner survey for a sales deck slide. Write exactly 3 numbered lines, each one stat written as \"X% of [group] (n=Y) ...\", under 20 words. Use only stats with at least 100 respondents, keep each group exactly as the survey defines it, and plain text with nothing else.","materialLabel":"Survey results (attached to your prompt automatically)","material":"HEARTHSTONE - Homeowner survey results (Sept, online panel)\n\nQ1. \"Have you reviewed your home insurance policy in the last 12 months?\"\nAll respondents (n=1,204): 38% yes\n\nQ2. \"Would you switch insurers to save $200/year?\"\nAll respondents (n=1,204): 57% yes\n\nQ3. \"Do you know your policy's deductible?\"\nFirst-time homeowners (n=312): 29% yes\nHomeowners 10+ years (n=455): 61% yes\n\nQ4. \"Would you use an app to file a claim?\"\nHomeowners under 35 (n=88): 81% yes   <- small sample, analyst says don't publish\nAll respondents (n=1,204): 64% yes\n\nAnalyst notes:\n- Any group under n=100 is not reliable enough for external materials.\n- Say \"homeowners surveyed\", not \"Americans\" or \"all homeowners\". It's an online panel, not a national sample.\n- Don't round 57% up to \"most\" or \"nearly 60%\" - use the exact number.\n\nSales asked for the \"81% of young homeowners want an app\" stat. Analyst said no (see Q4).","rules":[{"type":"numberedItems","count":3},{"type":"plainText"}],"mistakes":[{"title":"Grabbing the flashiest stat","why":"81% is the biggest number, so the AI leads with it.","fix":"Add: \"Only stats with n of 100 or more. Skip the 81%.\"","checks":["r1"]},{"title":"Overstating who was asked","why":"The AI writes \"57% of Americans\" because it sounds bigger.","fix":"Add: \"Say homeowners surveyed, never Americans or all homeowners.\"","checks":["r2"]},{"title":"Dropping the sample size","why":"The AI leaves out n=, so readers can't judge the stat.","fix":"Show the shape: \"X% of [group] (n=Y) ...\"","checks":["r0"]},{"title":"Getting a formatted slide","why":"The AI adds a headline, bold numbers and speaker notes.","fix":"Add: \"Exactly 3 numbered lines, plain text.\"","checks":["numberedItems","plainText"]}],"promptHints":[{"label":"Sets the stat shape","patterns":["n=","sample size","% of"]},{"label":"Filters small samples","patterns":["100","small sample","81"]},{"label":"Keeps groups exact","patterns":["homeowners surveyed","americans","group"]},{"label":"Says how many","patterns":["\\b(3|three)\\b.{0,20}(stats|lines)","exactly \\d+"]},{"label":"Asks for only the list","patterns":["only the (numbered )?list","no (intro|notes|explanation)","nothing else","plain text"]}],"hint":"Survey stats get bent on the way to a slide. Tell the AI the sample-size rule, the exact shape, and how to describe who was asked.","rubricLabels":["Each line gives an exact percentage with its group and sample size in the shape 'X% of [group] (n=Y)', and is under 20 words","Uses no stat with n under 100: never mentions 81% or homeowners under 35","Describes groups as the survey does (homeowners surveyed, first-time homeowners, homeowners 10+ years), never as Americans or all homeowners, and doesn't round numbers into words like most or nearly 60%"]}