The idea
Sales, shipments, support cases, and returns are different business processes, so each deserves a fact table at its own grain. Integration does not require forcing them into one giant table. It requires shared descriptive meaning at their boundaries.
A conformed dimension supplies that shared meaning. When two models use compatible definitions and values for attributes such as date, product, customer, or location, their separately calculated results can align on those attributes. “Product category” must mean the same thing on both sides—not merely have the same column name.
Make it concrete
A retailer stores one row per order line in a sales fact and one row per returned item in a returns fact. Analysts calculate revenue from the first and return count from the second, then align both result sets by the same conformed product and calendar dimensions. They can compare return rate by category without joining raw facts together and accidentally multiplying rows.
Conform deliberately
Agree with business owners on identifiers, attribute definitions, allowed values, and rollups. Reuse that agreement across process models. A smaller rollup can still conform if every value maps cleanly to the shared higher level. When definitions genuinely differ, name the difference instead of hiding it behind a familiar label.
Try it
Choose two analytical processes you want to compare. List the dimensions they appear to share. For one dimension, write a precise definition and one edge case that could cause the teams to classify the same record differently.