Must-be Quality

Deciding what to build next by starting from what users asked for has a long history. The problem is the set of things nobody ever asks for, which generate support tickets the day they break.

At Pagedeck, autosave stopped working for eight minutes and 40 tickets arrived within the hour. In the preceding six months, not one request had asked for autosave to be improved, and the feature had never appeared in a planning discussion. Ranking by request volume would have left that feature untouched, and in fact it went untouched for those six months. That week made it clear that satisfaction doesn't sit on a single line, and that the things people are delighted to get need a different name from the things they can't work without.

The defining trait is asymmetry: filling this need well doesn't raise satisfaction, and failing to fill it drops satisfaction sharply. Drawn as a curve, the left side dives and the right side flattens out along the neutral line.

1. There's a ceiling. Cutting the autosave interval from three seconds to one earns no praise, and the sprint spent doing it shows up in no survey answer. The same effort spent on search accuracy produces a difference people can see and describe.
2. There's no floor. A single failure cancels out everything else. Nobody discussed the feature that shipped the same week as those eight minutes.
3. It surfaces in outage tickets rather than requests. The need is too obvious to write down, which is why a survey classifies a feature here when respondents answer that having it is expected and not having it is unacceptable.
4. Other categories sink into this one over time. Several people editing one document at once used to be an attractive quality worth showing off; now its absence gets a product dropped from consideration before anyone opens it. A classification from two years ago is therefore worth re-running rather than reusing.

"This is a must-be, so doing it well won't move any metric. Losing it takes everything down with it."

Effort here doesn't buy satisfaction, and the work still has to happen.

"Zero requests doesn't mean it doesn't matter."

Sorting by request count drops this whole category on the floor.

"I'm cutting three indifferent items and putting the time into autosave reliability."

Trading work that changes nothing for work that shores up the floor.

The week after the autosave outage, Mia and Theo went back through the quarterly list. Of twelve candidate features, the five classified as indifferent quality came off, and surfacing save failures on screen went in. It was added knowing no metric would move. Four months later the same fault occurred and support received six tickets.