Calibrate support responses using fictional cases

Support replies can differ because writers assume different facts, not because one writer is more helpful. A calibration session makes those assumptions visible. Work with invented cases and independent drafts before comparing responses. Keep the case sheet and response notes in an existing document; this exercise does not require live customer messages.

Build a case with deliberate missing facts

Use this fictional practice case: a person says a document they used yesterday is missing. The case confirms that they can sign in and that they remember the document title. It does not identify the workspace, the route used to open it, or whether another person changed its location.

Write those facts and unknowns explicitly. Add the response constraints: no request for passwords, no promise of recovery, and no assumption that the document was deleted. The missing workspace is intentional; participants should ask for useful context rather than invent it to make their answer sound complete.

Draft the first reply independently

Ask each writer for a short first response before showing any other draft. The reply should acknowledge the reported inconvenience, request the next useful detail, and offer a safe action only when the case supports it. Keep the original drafts so the discussion can examine what changed.

One fictional draft might say: that sounds frustrating; which workspace were you using, and how did you open the document yesterday? Another might immediately say the file was deleted. Compare the assumptions behind those sentences rather than voting for the friendliest tone.

Apply the same response rubric

Review four dimensions: does the reply acknowledge the actual report, ask a question that narrows uncertainty, propose a proportionate next step, and avoid unsupported promises? Attach a reason to each observation. For the second draft, the deletion claim fails because the case contains no evidence about deletion.

Distinguish an essential omission from a stylistic preference. A different greeting may be acceptable; asking for an authentication secret is not. If reviewers disagree about a question, explain what its answer would change. A long list of questions is not necessarily better than one question that determines the next investigation.

Revise one answer and test a variant

Rewrite the selected reply around the missing workspace and original access route. Keep it brief enough to answer. For example, ask which workspace they used and whether they opened the document from a saved link or a document list. Explain why those details are useful without suggesting either route guarantees recovery.

Then change one case fact: the person already supplied the workspace. Review the revised reply again. Remove the question that is now redundant and choose the next relevant unknown. This variation checks whether the group learned a reasoning pattern or merely copied a sentence.

Keep the pattern and its limits

Record the reusable pattern: acknowledge the reported event, separate known facts from assumptions, ask for the context that changes the next step, and make only supported commitments. Include the original and revised case conditions so another writer knows when the pattern applies.

Group agreement does not establish customer satisfaction or actual resolution. No real missing document was recovered in this fictional exercise. A useful failure check is whether a reviewer can identify the evidence behind every diagnosis and promise. If a claim has no support, revise it before treating the response as a reference example.