Read the trends honestly before acting. The unit of interpretation is the trend, and a single data point is a question rather than a finding. See Interpret governance and delivery measurements.

A bad number has three possible causes, and they take different fixes: A miscalibrated policy, a practice problem, or pipeline mechanics. Name which one you are looking at before changing anything. See Identify the constraint: Policy, practice, or pipeline.

Pick the smallest change likely to move the measure, and make one change per cycle. Record what you expect to be different afterwards, because that expectation is what the last step verifies. See Decide what to change.

When the constraint is the policy, follow Adjust policies, assignments, and scope. When it is the teams or the pipeline, follow Adjust development and delivery practice.

Tell the affected teams what changed, why, and when it takes effect, before the first pipeline feels it. See Communicate the change to affected teams.

On the next measurement cycle, check whether the measure moved in the direction you expected and for the reason you expected. See Verify the effect on the next measurement cycle.

Governance accumulates. Along the way, retire the controls that only produce noise, or that protect against a risk the platform no longer has. See Retire controls that no longer earn their cost.