Measure delivery from two sides: the pipeline and the people. DORA covers the pipeline with two groups of metrics, and the pairing is the point. Throughput covers change lead time, deployment frequency and failed deployment recovery time. Instability covers change fail rate and deployment rework rate.
Read the pair
A team that deploys daily while a third of its deployments need an unplanned follow-up fix has strong throughput and a process that exports problems to production. Its next investment is test coverage and smaller changes. Measure per service, watch the trend rather than a league table, and ask the team why the numbers moved.
Ask the people
The DevEx framework from Abi Noda, Margaret-Anne Storey, Nicole Forsgren and Michaela Greiler covers what the pipeline can’t see: feedback loops, cognitive load and flow state. When a build takes twenty minutes, running a service locally takes three wiki pages, and meetings fill the mornings, each is its own lever. Survey developers on each dimension, pair their answers with build and review times, and show them what changed because of it.
What both leave out
Neither says whether the work was valuable, and the two sides can split: Annie Vella’s research found perceived productivity rising while flow eroded with AI assistants. Pair both with outcome measures, and remember Goodhart: once they become targets, they get gamed.
Related: Measuring AI impact · Value stream mapping · Goodhart’s Law