The retainer · $10,000 per month
Keeping it true is the work.
Building the warehouse is the easy half. APIs change, campaigns get renamed, platforms restate last week, and a report that was right in March quietly stops being right in June. This is what stops that.
Tests that block a wrong number.
When a deal references a stage the CRM no longer reports, its amount would drop out of won revenue without a trace. That stops the release instead, so a wrong number never reaches the dashboard. Correctness beats a green run.
Anomaly detection on revenue metrics.
Spend, conversions, sessions and calls are checked daily against their recent baseline. A metric has to break out of range by two separate measures before it pages anyone, so every alert stays worth reading.
A missed run raises the alarm.
Each run checks in when it starts and again when it finishes. If a run never starts or a deploy breaks the schedule, the silence itself raises the alarm. It's the one failure the code can't report on its own.
Backups proven by restore drills.
A copy of the warehouse is restored from backup and rebuilt from scratch, then row counts are compared against the original. A backup nobody's restored is not a backup.
Platform changes absorbed.
Renamed fields, changed endpoints and restated days are handled in the models. Facts rebuild over a trailing window, so a platform revising last week corrects itself.
Dashboard iterations and new sources.
The dashboard follows the questions leadership actually asks. New platforms get scoped, connected and modelled like the rest.
Ask it from Claude or ChatGPT.
Plug either straight into your data and ask anything in plain English. Every answer traces to its source rows, and it can only look, never touch, so there's nothing to break.