Data analysts and scientists are increasingly adopting software engineering best practices into their workflows. In this post, Sarah Krasnik discusses how analytics practitioners might adopt the framework of developing locally, testing in staging, and deploying to production. She recommends multi-environment development where SQL jobs are executed on replicas of production data and data quality tests are run in staging when a new feature or data model is created. Although this setup may be costly, she argues that the cost of incorrect data is much higher.