A cohort-based reporting system had grown one table at a time until it held roughly 150 near-duplicate tables and close to 190 measures, each a slight variation on the last. Dozens of queries had quietly broken along the way, and there was no single model anyone could point to as the source of truth.
Kingsley rebuilt the Power Query (M) layer so every cohort fed one shared model instead of its own copy, then rationalized the DAX measure library down to a core set that could serve every report. Broken queries were fixed as part of the same pass rather than left for later.
~150 tables became 1 scalable model, ~190 duplicated measures collapsed to about 9, and 158 broken queries were fixed, leaving one model behind every cohort instead of 150.