Retention settings should be reviewed for every artifact and service in the call path. A single dashboard value does not establish how long recordings, transcripts, traces, backups, and exported notes remain available.
Buyers need observable retention behaviour
A retention review checks approved storage periods, deletion mechanisms, access during that period, and exceptions. It should distinguish product settings from contractual commitments and actual observed deletion behaviour.
For a proposed buyer checklist, ask the supplier to identify each artifact, storage owner, configurable period, default, deletion trigger, backup handling, and subservice dependency. Record where the answer is documented and which configuration will apply to the pilot.
No universal retention period is prescribed here. Applicable legal, contractual, and operational requirements must be determined for the actual deployment. A vendor’s default is a product choice, not proof that it meets those requirements.
Review a fictional pilot configuration
A vendor offers short recording retention, but its support exporter stores transcripts separately. The buyer needs to know whether changing the recording setting affects those transcripts, historical exports, and diagnostic traces. If not, each requires its own control and owner.
Ask for a fictional test interaction containing unique markers. After the agreed expiry or deletion process, inspect the supported retrieval interfaces and exported artifacts. Document any paths that cannot be independently verified.
| Buyer question | Evidence to request |
|---|---|
| What is retained by default? | Artifact-level configuration and documentation |
| Can periods differ by artifact? | Supported settings and limits |
| What does deletion include? | Primary, derived, exported, and backup scope |
| Who can change retention? | Role and audit controls |
| What happens after contract termination? | Documented return and deletion process |
Do not accept “we do not train on your data” as an answer to every retention question. Model training use and operational storage are separate issues requiring their own evidence.
Keep the review proportional to the pilot
A small intake pilot can still create personal data in many systems. Limit unnecessary collection and integrations before relying on retention as the main protection. Review whether full recordings are needed for the specific test objective.
More restrictive retention can reduce troubleshooting evidence. Agree on the diagnostic information legitimately needed and how unresolved operations will be reconciled after source artifacts expire. Make the tradeoff explicit to both operations and engineering.
This checklist is proposed. The voice platform case study provides related integration context. See diagnostic retention design.
Buyer retention FAQ
Does one setting cover every provider?
Not necessarily. Review each service and artifact in the actual deployed path.
Is a deletion button sufficient evidence?
No. Establish its scope and test supported retrieval and export paths afterward.
Should retention be reviewed only before launch?
Revisit it when providers, collection paths, contracts, purposes, or operational requirements materially change.