> No, we don't need olap cubes. But we do need some type of rigor around analytic data. Otherwise, why go to all the trouble to collect it, count it, and predict from it if it may be wrong with no measure of that uncertainty?

Modern data warehouses don't need to build cubes for performance reasons. Low latency data warehouses like ClickHouse or Druid can aggregate directly off source data. The biggest driver for modeling is allowing non-coders to access data and perform their own analyses. I don't see that problem ever going away. Cube modeling with dimensions and measures solves it well.