Machine learning systems can reproduce or amplify patterns embedded in their data, and their decisions can change the conditions they later measure. A practical approach to fairness examines the full lifecycle—from defining targets and collecting data to deployment, feedback and the wider consequences of automation.