What was the problem?
Nine cron jobs produced the numbers in the board deck. Any schema change upstream broke them silently, and diligence would have found it.
How did Tekvo build it?
- 01
Introduced data contracts at every producer boundary, enforced in CI so a breaking change fails the producer’s build.
- 02
Replaced batch cron with streaming ingestion and incremental models.
- 03
Rewrote the warehouse layer for cost: partitioning, materialization strategy, and query budgets per team.
What changed?
Reporting lag went from a day to under a minute and warehouse spend fell 54% — with lineage a diligence team could actually follow.