Fixing the data, then passing the re-audit.

An external audit had flagged the retailer's own-brand food data (allergens, ingredients, supplier records) as inconsistent and unreliable. The re-audit was already scheduled. The data sat across legacy spreadsheets, ageing internal tools, and an on-prem Oracle warehouse that nobody wanted to touch. Patching the audit findings line by line would have shifted the same problem to the following year.

We argued for fixing the layer underneath, not the reports on top of it. The sponsor backed it.

The work split three ways. We migrated the warehouse from on-prem Oracle to AWS and Snowflake, so the team could query data without queueing for batch jobs. We introduced data contracts, so every upstream feed had clear ownership and quality expectations. And we built a metadata catalogue, so any team could find the data they needed without asking an engineer.

We coached the squad through the migration in parallel. Several were new to Snowflake, and a few of the data contracts ran across team boundaries that had been protected for a long time. We worked through both.

Outcome: the re-audit passed. Audit scoring went from around 60% to over 90%. Manual reporting time fell. Data accessibility went up across the business. The contract ended on its natural expiry.

What this shows: governance and migration work where the deadline is real and the problem is plumbing, not personality. This is the kind of engagement Engine Room exists to take on.