The best feature I ever shipped was a question
In 2008 I wrote a bowling app called BowlSheet. Buried in it was a feature called What-if. I never told anyone what it did. No tooltip, no tutorial, nothing in the release notes. You tapped it, and it took the game you just bowled and rewrote it with one change: every open frame became a spare.
Then you could toggle. Actual score, what-if score. Back and forth.
If you don't bowl: an open frame is when your two throws leave at least one pin standing. In league play it's usually a near miss. A good first ball, one pin left, and then you miss it.
Here's what the toggle showed people. A game with a few opens, most of them single-pin leaves, sat 30 or 40 pins below its what-if twin. Not because of bad bowling. Because of two or three boring misses on pins that were standing alone, begging to be hit.
Bowlers chase strikes. Strikes are loud, they feel like skill, and the whole house turns around when you string four of them. Spares are homework. But the math doesn't care about noise. A game of nothing but nine counts and spares, without a single strike all night, is a 190. Most league bowlers would sign for a 190 right now.
I never told a user to practice spares. The app didn't either. It just put the two numbers next to each other and let you flip between them. And people changed. They started caring about the ten-pin. So did I. My league average climbed, and I watched the same thing happen for other people. Not because anyone got a lecture. Because the gap was visible, every game, in their own numbers. I'll be honest about the evidence: this was league bowling in 2008, not a controlled study. What I have is what I watched. But I trust what I watched.
It took me seventeen years and a book to understand what I'd built.
I recently finished Warren Berger's A More Beautiful Question. Berger's central move is a three-stage sequence: Why, What If, How. Why is it this way? What if it were different? How do we get there? Somewhere in the middle of the audiobook it clicked that I had shipped his middle stage as a button. What-if wasn't an analytics feature. It was a question rendered in software: what if you'd picked up the easy ones? And it worked precisely because it never answered. It just asked, every game, in numbers you couldn't argue with.
That's the part that transfers to the day job, leading engineering teams.
Most roadmap conversations are strike conversations. The big bet, the flashy launch, the thing that gets applause at the all-hands. The questions that actually move the score are usually the spare pair, and they cost almost nothing to ask.
What if we build this? Not "how long will it take." What does it open up? What becomes possible for the customer, or for the next milestone, that isn't possible today?
What if we don't? What does standing still cost, and who pays it? The cost of not shipping is a what-if score nobody renders. It never shows up on a burndown chart. Someone has to make it visible, the way the toggle did, or the team keeps optimizing whatever is loudest.
Notice what the feature never did. It never assigned homework. Telling a bowler to practice spares is advice, and advice mostly bounces. Showing them their own 40-pin gap is a question they answer themselves, and that's when behavior actually changes. I think about this constantly as a leader. The instinct is to hand people answers. The leverage is in making the gap visible and letting smart people draw their own conclusion, because a conclusion you reached yourself is one you'll act on.
It matters more now than it did in 2008. AI is collapsing the cost of implementation. Building the thing is less and less the constraint. So the score no longer hinges on how hard you can throw strikes. It hinges on what you ask before you build: what if we do, and what if we don't?
The best feature I ever shipped didn't add a capability. It asked a question and got out of the way.
What's the ten-pin your team keeps walking past?