Multi-Source Ingestion Readiness Check (Onboarding)

Multi-Source Ingestion Readiness Check (Onboarding)

1. 01

   Inventory every source system feeding the client warehouse before writing a single connector

   List each origin (CRM, ad platform, billing, product events), its owner, refresh cadence, and whether it exposes an API, a nightly export, or only a manual CSV drop.

2. 02

   Confirm row-level grain and primary key for each source with the client's data owner

   Agencies routinely inherit two systems that both claim to be the customer table. Document which one wins and how duplicates get resolved before load.

3. 03

   Map timezone and date conventions across sources

   Ad platforms report in account timezone, billing systems often use UTC, and client CRMs may store local time. Pick one canonical zone and record the conversion rule.

4. 04

   Decide the storage tier per dataset: hot query layer versus cold archive

   Active reporting tables belong in a query-optimized layer, while raw landing files can sit in flat-rate object storage such as Wasabi or Backblaze B2 where egress fees do not accumulate.

5. 05

   Test one full refresh cycle end to end before scheduling production runs

   Run the pipeline manually, compare row counts against the source, and verify that late-arriving records do not silently overwrite earlier loads.

6. 06

   Write the schema contract into the client statement of work

   Name the fields, grain, and refresh SLA you will support, plus the change-request process when the client alters a source system mid-retainer.

7. 07

   Set a rollback point and a named owner for the first 30 days

   If a load corrupts the reporting layer, the agency needs a documented restore path and one person accountable for executing it.