Before we started, fourteen people on the data team answered: how confident are you that our pipeline outputs are correct? Three said very confident. Nine said somewhat. Two said they had no idea.
Six months later: eleven very confident. Three somewhat. Zero unsure.
Months 1 to 2: The Right Starting Point
We started with our most trusted pipeline, not our most broken one: the customer revenue feed going to the CFO dashboard every morning. Writing the first contract for a working pipeline is the right move. You are formalizing what good looks like. Month two hit the first organizational friction: the upstream team did not see why they needed to honor a contract another team wrote. The real adoption challenge is not technical. It is political.
Month 3: The Enforcement Decision
We moved from advisory to enforced contracts. Any violation stopped the pipeline. It broke three pipelines in the first week and caused tense conversations. But it made the contracts real. Advisory contracts are suggestions. Enforced contracts are commitments.
Months 5 to 6: Defaults Shift
New datasets started arriving with draft contracts attached, from teams that had not been part of the rollout. The semantic layer expanded from 12% to 74% of production datasets. Violation response time dropped to under two hours. The CFO dashboard went eleven straight weeks without a question about whether a number was right.
What We Would Do Differently
Start enforced. And include upstream teams in writing the contracts from the beginning. Co-authorship creates ownership. Ratification creates resentment.
Six months is enough to change the default. Talk to us about your timeline.