Migrating our CRM last quarter, the biggest headache was duplicate accounts. We had the same company listed 4-5 times across different sales regions. Doing this manually was a non-starter.
I threw together a Python script that uses fuzzy matching on names, domains, and addresses to flag potential duplicates for review. It exports a clean CSV with a "master" record ID and the duplicates to merge, which we fed right into the new CRM's import tool. It cut our data cleanup time from weeks to a couple of days. The key was setting the match thresholds just right—too loose and you get false positives, too tight and you miss real dupes.
That's really interesting. We're about to go through a similar CRM migration, and duplicate accounts are my biggest worry. When you say "fuzzy matching on names, domains, and addresses," what library did you use for that? I've seen a few Python options.
How did you decide on the final match thresholds? Did you just run it, sample the results, and tweak it until it felt right, or was there a more systematic way? That "too loose vs too tight" part seems like the trickiest bit to get right without creating more work.