Consider the following situation:
Inside an organization, each Business Unit has requested specific, separate models. After some time, some high level members over Finance and Sales have requested a model containing multiple data sets. This is fine but now we have to maintain three models and duplicate both sets of measures.
- Model Finance can only be seen by Finance
- Model Sales can only be seen by Sales
- Model C can only be seen by Execs over both units and has all the logic and measures from Finance and Sales.
Assuming this was an issue across the whole organization, please can you clarify which of the following is best practice:
- Many, many, models containing the same data, measures and relationships. Just mashed up in different ways
- A single, massive model with complex RLS.
From my experience, a hybrid approach is preferred where 60% of the org can be fed by a single Model, and the other 40% is fed by purpose build models. However, this seems to go against my perceived BI Ideology, ONE VERSION OF TRUTH. Will we ever be able to, (or can we already?), combine multiple simple models, in to purpose built, derived models?
Is there another technology?
Is the answer to use flattened, imported SSAS Tabular models?
It seems like an easy answer.
Semi-Potential solutions:
- Cross report Drill through
- No good because you can't do cross model comparisons.
- One big model
- Sluggish, difficult to measure change impact across all BUs.
- Many, many permutations of all BUs
- Just no.