Define a comparable cohort
Start with cases closed during a specific period and allow the same follow-up window for every case. Count a repeat only when it concerns the same underlying problem under your documented rule. A thank-you message or a new order question should not automatically count. Keep an unmatched category when you cannot confidently connect contacts across channels.
Investigate the reason for return
Separate incorrect answers, incomplete action, unclear instructions and new facts. Review by intent, language and channel before choosing a remedy. A higher observed rate may reflect better contact matching rather than worse service, so document changes in the matching method. Use de-identified examples in reporting and avoid exporting raw customer transcripts into a public dataset.
Fictional worked example
Fictional cohort: 100 closed cases all receive a seven-day follow-up window. Twelve customers return, but three have unrelated questions and two only say thanks. Seven same-issue returns give a 7% observed repeat-contact rate under this definition. This is an arithmetic illustration, not an industry benchmark or a result from Anti Sandbox.
Action checklist
- Fix the cohort, follow-up window and same-issue rule.
- Exclude unrelated questions and acknowledgments consistently.
- Retain unmatched contacts as a stated limitation.
- Review return reasons before changing the content or workflow.