{"id":"cs-usage-drop-discovery-questions","audience":"customer-success","category":"support-work","technique":"verification","track":"support","level":"intermediate","title":"Ask the right questions about a usage drop","free":false,"task":"Saltmarsh Media's usage report looks bad and you have a call with their admin. Write exactly 5 numbered, open-ended questions (no yes/no), each under 20 words, tied to a real change in the data, and each ending with a tag in square brackets: [Reports], [Editors] or [Mobile]. Use each tag at least once. Don't ask about the part of the drop caused by your own tracking change, and don't blame the customer. Output only the questions.","materialLabel":"Account notes (attached to your prompt automatically)","material":"Saltmarsh Media - usage report, Sept vs Oct\n\n- Logins: down 35%\n- Reports generated: down 50%, almost all in the marketing team (from 400 to 190)\n- API calls: flat\n- Active editors: 46 -> 39\n- Mobile app sessions: up 12%\n\nNote from our data team (Oct 2): \"On Oct 1 we changed how we count logins: SSO sessions now count once per day instead of once per browser tab. Every account shows roughly a 20% login drop from this alone. Not real usage.\"\n\nCSM notes:\n- New marketing director, Petra Lowe, started Sept 15.\n- Saltmarsh admin: Owen Fraser.\n- Contract renews in June.","rules":[{"type":"numberedItems","count":5}],"mistakes":[{"title":"Treating the tracking change as real","why":"The AI asks the customer why logins fell 35%, which is mostly our own counting change.","fix":"Say: \"About 20% of the login drop is our tracking change. Don't ask about it.\"","checks":["r0"]},{"title":"Generic questions","why":"The AI writes questions any account could get.","fix":"Ask: \"Tie each question to a specific change in the data.\"","checks":["r1"]},{"title":"Yes/no questions","why":"The AI asks \"Did something change?\" which ends the conversation.","fix":"Say: \"Open-ended only, starting with What, How or Which.\"","checks":["r2"]},{"title":"No fixed count","why":"The AI writes a long list grouped under headings.","fix":"Ask for \"exactly 5 numbered questions, each under 20 words.\"","checks":["numberedItems"]}],"promptHints":[{"label":"Sets the count","patterns":["\\b(5|five)\\b.{0,10}questions?"]},{"label":"Accounts for the tracking change","patterns":["tracking","counting","20\\s*%","not real"]},{"label":"Asks for open questions","patterns":["open.?ended","no yes.?/?no","what.{0,5}how"]},{"label":"Ties to the data","patterns":["data","specific","tied?"]},{"label":"Avoids blame","patterns":["blam","neutral","accus"]}],"hint":"Part of the drop isn't real. Tell the AI what the data team found, and what makes a question useful on a call.","rubricLabels":["No question asks the customer to explain the overall 35% login drop or the login tracking change (about 20% of it is our counting change)","Every question ends with a tag [Reports], [Editors] or [Mobile], all three tags appear at least once, and the questions refer to real changes (the marketing team's report drop or the new marketing director, fewer active editors, more mobile use)","Every question is open-ended (starts with words like What, How or Which, not Do/Did/Is/Are), and none blames the customer"]}