BI & Analytics

Where I use it

SSAS models over the Acecook warehouse feeding Power BI, and SAP Analytics Cloud — the subject of my C_SAC certification in April 2026.

The point of a semantic layer

Without one, every report author writes their own version of every measure. Three Power BI reports quote three different figures for sales-out, all defensible, all built from the same warehouse, and leadership loses trust in the entire platform.

A semantic model fixes the definition in one place. Report authors compose from measures that already mean something rather than reinventing them. It is a governance mechanism far more than a technical one.

What I have learned about reports

Ask what decision it supports. "A sales report" is not a requirement. "Which SKUs are underperforming in which regions this month, so we can act on them next week" is — and the two produce very different reports.

Fewer numbers, better chosen. A dashboard with forty tiles gets scanned and forgotten. Six that matter get used.

Performance is a modelling problem. A slow Power BI report is nearly always a model problem — the wrong grain, missing aggregations, or a relationship the engine cannot use efficiently — not a visualisation problem.

Agree the definition before you build. The written definition of "sales-out", signed off before a table was created, prevented more rework at Acecook than any technical decision on that project.

To expand: SAC vs. Power BI in an SAP landscape, and notes on measure design.