Ask HN: How do large and complex organization make “data-backed” decisions?

-- Im looking for more of a general answer, references, or even tooling.

A naive example is Amazon Web Services and their offerings. How do they decide what to build, offer, improve, remove, etc in their AWS catalog? Givens would be things like their employment capabilities & constraints, work in progress, expected outcomes ($) , time, and Im sure much much more. Maybe it would be many decision trees that ultimately form a singular tree where they could feed it data from the past and run simulations to produce forecasting? I'm just left wondering how do they standardize this across the organization and make it simple for everyone to provide their necessary inputs (ie an engineering team providing WIP and velocity).