Not an information supermarket

The most useful thing we did to one of our products this week was delete four features nobody had built yet. Four breadth plays — each reasonable, each sitting in the backlog with quiet momentum — all closed as won’t do in a single pass, and a whole round of tickets went to the graveyard with them. The principle we wrote on the headstone: this product is not an information supermarket.

The pull toward breadth is the strongest and least examined force in an engineering-led team. Every one of those features is easy to justify one at a time — a user might want to search, a user might want another data tab, a user might want a radar — and each is a clean, shippable, satisfying build. Breadth is where engineers are comfortable: the requirement is legible, the scope is bounded, the demo is obvious. So the backlog fills with things that are individually defensible and collectively fatal, because the thing they add up to is “a slightly worse version of every tool that already exists,” and nobody ever decided to build that. It accretes.

What made the deletion possible was having an actual bet to measure against. Breadth features don’t lose an argument on their own merits — they lose it against a stated thesis. Ours is narrow: show the judgment, keep the receipts, own the misses. A generic search box doesn’t serve that; it serves the generic job of “look something up,” which a dozen incumbents already own. Once the bet is written down, the question for each backlog item stops being “would someone use this?” (the answer is always yes) and becomes “does this make our one differentiated thing deeper, or just wider?” The four features made it wider. So they went.

The uncomfortable part is that saying no to breadth means saying no to obviously-usable things. A half-built feature is evidence of a decision already drifting — two of the four were partially scaffolded. Killing them means writing off real work and disappointing the version of yourself that wanted the tidy checkbox. But a product is defined at least as much by what it refuses to do as by what it ships, and a team that only ever adds has no mechanism for staying pointed at anything.

The rule we’re keeping: breadth needs a prosecutor, not just a sponsor. New features arrive with a built-in advocate — the person who wants them — and no built-in opponent. So the backlog review’s job is to be the opponent: for each item, name the thesis it deepens, and if the honest answer is “none, it just widens us,” close it and write down why. The closed tickets aren’t waste; they’re the record of a company that knows what it isn’t.