Data products

A team wants the same data product to run consistently in development, testing, and production. Which practice most directly addresses differences in installed libraries and system dependencies?