Migrating from the legacy Data Warehouse
The legacy warehouse (client_analytics) is deprecated. It keeps working, untouched, during the migration period — you migrate at your own pace by updating two things in your queries: the database name and, where they changed, table and column names. Your credentials and connection host may also change; your Account Manager will confirm.
The legacy client_analytics database will be decommissioned after the migration period. All new integrations should be built against treble_client_analytics only.
Table mapping
Column renames to watch for
Behavior differences you should know
- Full history. The legacy warehouse trimmed most tables to a rolling 3 months. The new warehouse serves everything. If your queries relied on the window as an implicit filter, add explicit date filters (also good for performance).
- Fresher “today”, stated consolidation. Data lands in minutes instead of hours, and figures for a day consolidate fully within 48 hours (see the data contract). During a migration comparison you may see the new warehouse slightly ahead of the legacy on the most recent day — that is legacy refresh lag, not a discrepancy.
- Messages of deleted conversations are excluded from
fact_agent_conversation_messages by design (the conversation they belong to no longer exists).
- Contact enrichment on old conversations freezes.
helpdesk_contact_id and contact_wa_id on agent conversations older than 90 days keep the value they had — CRM changes no longer propagate backwards to them.
- Two documented formula refinements in
fact_agent_daily: first-response time now comes from the platform’s own computation, and CSAT averages consider only conversations with a rating (rating != 0). Both are corrections; historical parity was verified before release.
- No aliases. Legacy names are not mirrored in the new database — the mapping above is applied once, in your queries, and both systems coexist while you do it.
Suggested migration path
- Point a copy of your dashboard/report at
treble_client_analytics using the mapping above.
- Run both versions side by side for a few days; expect exact matches on consolidated days.
- Switch over, keeping explicit date filters.
- Tell your Account Manager when you no longer use
client_analytics — it helps us schedule the decommission.
