Your Data Debt Isn't Technical

Most of the data debt I see at 40-person companies is three definitions of ‘customer’ written in SQL.

Sales counts anyone with a signed contract. Finance looks at who paid this month, while product goes by logins. Each team built its own model, each one is right for that team, and the dashboards disagree.

Then the CTO opens the 600-line query that tries to reconcile them, and it’s ugly enough to call technical debt. Rewrite it in cleaner SQL, or move it to a nicer tool, and all three definitions survive. Six months later the new version is just as ugly.

Paying it down starts with a meeting and a shared doc. Write down what “customer”, “active” and “revenue” mean and who owns each one. Pick the version that goes in the board deck, then change the code to match.

That doc is boring to write. In my experience it deletes more SQL than any refactor.

How many definitions of ‘customer’ does your warehouse have right now?