Would it be fair to say that in nearly all the examples here, the “raw” data simply wasn’t raw enough? As soon as you’re pulling data into your ETL system where you have to assume that a column is derived from say [(column A minus Column B) divided by 2pi] instead of just pulling in column A and B and doing the math yourself, you’re increasing the chance of inheriting mistakes beyond your control.

@vliet I feel your pain. It’s been my experience that big Health Organisations are often run by teams of highly intelligent, persuasive well educated people with massive egos. They often ascend to positions of seniority through being excellent nurses or paediatricians or [insert whatever clinical specialty here]

Because they’ve been good at what they did to get where they are, they fall into the trap of thinking that they know best about everything (especially BI architecture!). So they’ll come to the BI team not with a business need, but with a request for what they see as the solution to a business need. It’s the same as if I went to the doctor and said “I want you to resect half of my liver.” instead of “I have some right upper quadrant pain, can you help me?”