Semantic Layer

This solves the most common and most damaging failure in enterprise analytics: three dashboards showing three different revenue figures, all technically correct, each embedding a slightly different definition of the same word.

The definitions diverge because they live inside individual reports. One excludes internal test accounts, another includes cancelled orders until refund, a third uses order date rather than shipment date. Nobody is wrong, and the organisation cannot answer its own question. Executive time then goes into reconciling numbers rather than deciding anything.

A semantic layer defines metrics once — as code, in version control, reviewed — and every tool queries through it. Revenue means one thing, and changing what it means is a reviewed change with a history.

The architectural question is where it sits. Embedded in a BI tool, it governs only that tool, which fails as soon as a second tool or a notebook appears. As a standalone layer with an API, it can serve BI, notebooks, embedded analytics and applications alike, which is the shape that actually delivers consistency.

The obstacle is rarely technical. It is that agreeing what a metric means requires a decision from someone with authority, and the definitional arguments that surface are ones the organisation has been avoiding for years.