A reconstruction, not evidence of thoughts
DOM-based tools such as Clarity reconstruct sessions from HTML and events. This is not an exhaustive video: masking, unsupported components and missing events limit the conclusions available.
| Check point | Preventive measure | Evidence to seek |
|---|---|---|
| Forms and dynamic content | Exclude or mask before capture | Test using fictional values. |
| URLs and messages | Remove identifiers and secrets | Inspect the data actually transmitted. |
| Access and issue tickets | Limit people and excerpts | Check permissions, deletion and shared content. |
Connect a signal with a reproducible fault
- Define a precise question and inspect successful journeys too. A selection of failed sessions does not represent all visitors.
- Connect the observation with a step, error and technical context, then reproduce using fictional personal data.
- After a fix, check the same scenario and task outcome. Repeated clicks do not establish emotion or the reason for abandonment.
Check rules and the data lifecycle
Masking does not automatically make the entire processing anonymous or compliant. Check information, choices and consent where required, access and retention. The French exemption for certain audience-measurement tools is conditional; it does not automatically extend to individual replay.
Primary references
- Microsoft Learn — Recordings overview [EN]
- Microsoft Learn — Masking content [EN]
- OWASP — Logging Cheat Sheet [EN]
- CNIL — Cookies et mesure d’audience [FR]
Session replay needs review of captured data
Replay can expose fields, URLs, messages or personal content when masking is incomplete. Check actual recording in dynamic states, not just initial configuration. Define access and retention; extracts explaining defects should not unnecessarily reveal individuals.
- Test masking across forms, errors and dynamic components.
- Check applicable choices before recording.
- Limit replay access and data copied into tickets.
Does replay represent every user?
No. Collection can be limited by choices, devices or sampling. Use observations to understand a case, then assess its extent with suitable measurements.
Primary reference: CNIL — Audience measurement tools. Related method: read the practical guide.