Data Contracts — Week 2 CheckWeek 2 of Data Quality, Contracts & Observability: contracts in production at GoCardless, Whatnot, and Adidas — and the shift-left argument. Every question traces to a case study you read this week.

GoCardless got software engineers to ADOPT data contracts voluntarily — 300 contracts across 28 teams. What made adoption work?

In the GoCardless design, what does the transactional outbox guarantee — and what did teams start MISUSING it for?

What problem was Whatnot's Actor-Action-Object taxonomy solving?

Whatnot checks event quality at THREE lifecycle stages. Which set is correct?

A schema registry with compatibility checking is already enforced. What does a data CONTRACT add that the registry cannot?

GoCardless's honest scar: which adoption gap persisted after six months?

The 'shift-left' argument against medallion architecture says the bronze layer largely exists because…

Adidas encoded data-product sharing as Git-versioned JSON definitions that auto-provision access grants. What's the governance win of Git as the control plane?