Sixty-three deleted lines

The pull request was titled "remove saved views" and it deleted sixty-three lines. It took two weeks, and eleven of those days were spent proving that it was safe rather than making it so.

Saved views had shipped fourteen months earlier. In that time, forty-one people had created one. Nine had opened one more than once. Two had opened one in the last ninety days, and both of them were on our team.

The cost was never the code

Sixty-three lines is nothing. But saved views had a settings row, an empty state, a keyboard shortcut, a migration, a paragraph in the docs, three tests, and — the expensive one — a place in everybody's mental model of the app. Every new feature had to answer the question how does this interact with saved views? That question had been asked in design review eight times that year. Nobody ever had a good answer, because the honest answer was that it did not matter.

That is the real bill. Not the maintenance. The way an unused feature quietly taxes every decision downstream of it.

How we did it without a fight

1. Found the two active users. Both were internal. Asked them directly.
2. Shipped a banner in the feature itself, for thirty days, saying it was going away and offering an export.
3. Removed the entry points but left the data, for one release.
4. Then deleted the code and the row and the shortcut and the paragraph.

Nobody wrote in. Nobody has mentioned it since. The settings screen is one row shorter and the design reviews are shorter than that.

The part I keep thinking about

We knew within about six weeks of launching it that it had not landed. It stayed for fourteen months because removing it required someone to say out loud that a thing we had built was not working, and there is no ticket type for that.

I would like a ritual for it. Not a metric or a dashboard — a recurring hour where the only agenda item is what we should take out. We have started doing it monthly. So far it has been the most productive hour on the calendar.