The locations (target these, not mailboxes)
Purview retention for Teams uses DEDICATED locations — Teams chats, Teams channel messages, Teams private channel messages (post-2025/26 migration, its own line) — not the Exchange location, even though the compliance copies live in mailboxes. Meeting artifacts split by type: recordings/transcripts are FILES (OneDrive/SPO retention governs them, plus the recording expiry timer from the meetings module — two clocks, the stricter of deletion wins/holds override).
Retain-vs-delete mechanics are standard Purview: retain-only (keep at least N), delete-only (purge at N), or both; static scopes or adaptive scopes (query-based populations — the modern default for anything dynamic).
The timing folklore, corrected
- Policy application to new locations: hours-to-days, not instant.
- Deletion after expiry: the pipeline sweeps on multi-day cycles — a 30-day delete policy means "gone in ~30-45 days", never "day 30 at midnight". Users may see messages vanish from clients days after the compliance purge or before — the two copies (messaging module) purge on separate schedules.
- Shortest-wins conflicts resolve by Purview's principles of retention: retention wins over deletion; longest retention wins; explicit wins over broad. Write the org's expected outcomes down against these rules ONCE and most disputes end.
Design that survives audits
Separate policies per conversation class (chats aggressive, channels project-lifetime, private channels explicit post-migration) + adaptive scopes for populations (leavers, regulated roles) + a WRITTEN mapping of policy -> regulation clause. The mapping document is what the auditor actually wants.
What to watch (proofs)
- Coverage truth: Purview > Data lifecycle > policies — each Teams location's policies and scope; the serv365-adjacent habit: export and diff monthly (policy edits are audit-logged, reviewable).
- The pipeline in action: a short-retention TEST policy on a pilot scope — measure your tenant's real send-to-purge latency; keep the number for litigation-hold conversations.
- Distribution success: policy distribution status per location (Purview shows errors) — a policy stuck in error is coverage you believe in but don't have.
- The mailbox reality: eDiscovery hits before/after purge windows (messaging module's proof pattern) — the end-to-end verification legal will accept.