Context
Experimentation already existed, but ad hoc - different teams tracked results differently, and there was no shared bar for what counted as a valid test.
My role
Defined governance, tooling standards, and prioritization criteria, then owned adoption across business lines.
Approach
Started with the two teams already testing well and turned their practice into the default standard, rather than designing a framework in the abstract and asking everyone to adopt something unfamiliar.
Outcome
Scaled from ad hoc testing to 12+ concurrent experiments, embedding experimentation into standard quarterly planning.
What I'd do differently
Nothing structurally - borrowing from teams that already worked got faster buy-in than a top-down framework would have. Worth remembering before defaulting to designing from scratch next time.