Forum Discussion
Struggling with data integrity
The solution would be to have a Data Governance body that actually has teeth, money, and C suite support. You can have endorsed and (especially) certified semantic models, and you can have policies in place that anyone not using these will get inferior support etc. I would love to see that becoming reality some day.
For any sufficiently large company "single source of truth" is an unattainable myth. The raw data maybe (via a data lake with medallion architecture). But the semantic models need to be tailored to the individual business question for each report. Trying to stuff everything into a single data model with stumble at the 50K cardinality hurdle and come to a screeching halt at the 1M row limit for Direct Query.
Thanks Ibendlin
So what's the solution to having some complex DAX measure in multiple different semantic models, it needing to change, and keeping track of everywhere that complex DAX measure exists so all the copies of it can be updated simultaneously to avoid inconsistency between models? Our different schools want to track and report in different ways and it would be counter-productive for us to force them into alignment just because it will make the data management easier, so we can't easily have standardised reports, but we also need to report at the high level on metrics between different schools without the reports giving out different numbers.
I can think of ways to do it but none of them are great, either we have large inconsistency risks or an overbearing admin load for the data team...
The solution would be to have a Data Governance body that actually has teeth, money, and C suite support. You can have endorsed and (especially) certified semantic models, and you can have policies in place that anyone not using these will get inferior support etc. I would love to see that becoming reality some day.
We can dream... 🙈🙉🙊