ECS 2027

31/05/2027, Congress Center Düsseldorf
Düsseldorf, Germany

The Clean Factory: CI/CD and Quality Gates for Semantic Models

A semantic model gets built, reviewed by someone reading it on screen, published and then edited by three different people. Nobody notices until a number moves, and nobody can say when it started. Most teams answer this with a list of best practices. A list in a review checklist is advice; the same list running in a pipeline is enforcement, and the distance between the two is where an analytics platform quietly loses its integrity. This session walks the clean factory you build around a semantic model, level by level. What is worth checking while the model is still being authored, and what makes a model reviewable at all. What runs before anyone merges. How environments really stay separated. What you verify after deployment — including whether row-level security actually works, and how you find out rather than assume. And what can honestly be certified, as against what cannot. Gates have a reputation: slow, bureaucratic, resented, and switched off the first time they block a release. Built well they do the opposite. They are comfortable to work under, they make visible exactly what needs to be visible to the people who need to see it, and they stop being a tax on delivery. That is a question of method far more than of tooling — which is why CI/CD for analytics fails on method long before it fails on technology.