Slowly Changing Measures

In data warehousing, measures such as net sales, customer reliability scores, churn likelihood, or sentiment indices are transactional data scored from the business events by measurement functions. Dimensions model subject-oriented data used as analysis perspectives when interpreting the measures. While measures and measurement functions are traditionally regarded as stable within the Data Warehouse (DW) schema, the well-known design concept of slowly changing dimensions (SCDs) supports evolving dimension data. SCDs preserve a history of evolving dimension instances, and thus allow tracing and reconstructing the correct dimensional context of all measures in the cube over time.

Measures are also subject to change if DW designers (i) update the underlying measurement function as a whole, or (ii) fine-tune the function parameters. In both scenarios, the changes must be obvious to the business analysts. Otherwise the changed semantics leads to incomparable measure values, and thus unsound and worthless analysis results.

To handle measure evolution properly, this paper proposes Slowly Changing Measures (SCMs) as an additional DW design concept that prevents incomparable measures. Its core idea is to avoid excessive schema updates despite regular changes to measure semantics by a precautious design, handling the changes mostly at the instance level. The paper introduces four SCM types, each with different strengths regarding various practical requirements, including an optional historical track of measure definitions to enable cross-version queries. The approach considers stable business events under normal loading delays of measurements, and the standard temporality model based on the inherent occurrence time of facts. Furthermore, the SCMs concept universally applies to both, flow and stock measure semantics.