Duplicates, Typos, and Stale Records

Most data quality problems aren't exotic. They're boring. Duplicates. Typos. Records nobody updated since 2019.

The same customer listed three times under three spellings. Addresses from two moves ago. A "current" spreadsheet that nobody current maintains. None of this looks dramatic, and all of it quietly poisons whatever AI system learns from it. Small errors compound. A model trained on messy records makes messy decisions, confidently.

The fix is equally boring: regular cleanup. Merge the duplicates. Fix the spellings. Archive what's stale. Put one person in charge of each dataset and give them time to tend it.

Auditors love clean records, by the way. So do models. Boring maintenance beats heroic rescue projects every time.

I explain frameworks; I don't give legal counsel.