Cleanup feels like tidying. The hand reaches for delete. Last year’s conversion series is attached to the name you are about to retire. Finance notices before product does.
Wei Liang Tan treats cleanup as a mapping problem. Each duplicate cluster gets a surviving name, a retired name, a date from which the surviving name is canonical, and a sentence about how the old series should be read. The catalogue becomes shorter. The history remains addressable.
A cluster is not a synonym list
purchase and order_placed may be the same conversion, or they may be pay-in-app versus pay-on-delivery. We do not collapse them because they look similar in a spreadsheet. Screen walkthroughs and a product owner still matter. Reconstruction of the tracking plan sometimes has to precede cleanup for that reason, which is why the two engagements are priced apart.
What “retired” means in our documents
Retired means the name should not be instrumented in new work and should not appear in new saved reports. It does not mean the rows are gone. If a vendor console offers a hidden or archived state, we use it. If it only offers delete, we leave the name and mark it in the catalogue with a do-not-instrument line, then argue with the vendor later rather than with finance this week.
Read the taxonomy cleanup scope before you ask us to “just remove the extras.”