While thinking through a data engine product at Minerva, I keep coming back to one product trap: treating verification as a green checkmark. Real verification is rarely that clean. It is evidence, source quality, confidence, review history, and the cost of being wrong.
A system that says only verified or not verified hides the part operators actually need to reason about. What matched? Which source said so? How fresh is it? Was the decision automatic, reviewed, disputed, or only good enough for the next step?
The data model has to leave room for uncertainty because the real world does. The product can still make decisions, but it should not pretend the decision erased the ambiguity underneath it.