Ninety out of a hundred beats nine out of ten

There’s an experiment with jelly beans that makes people argue with themselves out loud. You’re offered a chance to win by drawing a red bean blind from a bowl. Bowl A has one red bean and nine white — one in ten. Bowl B has nine red beans and ninety-one white — nine in a hundred, which is worse odds, and everyone can see it’s worse. A striking number of people draw from Bowl B anyway. When Epstein’s group asked them why, they’d say some version of it out loud: I knew the odds were against me, but there were just more red ones, so I felt like I had a better shot. They watched themselves make the wrong choice and made it anyway.

That’s ratio bias, and underneath it is a plainer bug called denominator neglect. When the brain sizes up a chance, it fixes on the numerator — the count of good outcomes, the winners, the red beans — and lets the denominator, the total it’s divided by, go soft and grey. Nine feels like more than one because it is more, as a raw count. The fact that nine is sitting inside a hundred and one is sitting inside ten is exactly the part that fades. Psychologists call the general reflex numerosity: more discrete things reads as more, almost before the ratio has a chance to load.

⁂

Now look at your product through that lens, because it is covered in numerators you get to choose. A price is a count. “$1 a day” and “$365 a year” name the identical cost, but one is a small number attached to a small unit and the other is a big number attached to your customer’s whole year, and they do not land the same — not because anyone failed at arithmetic, but because the small frame is the one the brain files as small. A progress system is a count too: “8,470 points” feels like real distance traveled in a way that “you’re 40% of the way” never will, even when the 40% is the more useful truth. Storage, limits, ratings, streaks — every one of them is a number sitting inside a denominator you can make loud or quiet.

So the build decision is less about lying and more about which true number you put in front of the eye. If you want an amount to feel affordable, show it in the smallest honest unit — per day, per seat, per use. If you want progress to feel earned, let the accumulated count show, not just the fraction remaining. If you’re proud of a success rate, “worked for 9,200 of the last 10,000 runs” carries a weight that “92%” flattens, because the mind can see the nine thousand. None of this invents a fact. It picks, among faithful ways to state the same fact, the framing that matches how the number will actually be felt.

⁂

Here’s the guardrail, and it matters more for a neuroscience-branded pitch than most. This can curdle into a trick fast. The moment a customer senses that the small unit was chosen to hide the annual total — the “$1 a day” that’s quietly a four-hundred-dollar commitment with a cancellation maze — the framing stops being clarity and becomes a tell that you’re managing them. Sophisticated buyers normalize on purpose; they’ll convert your daily price to yearly precisely because they’ve learned that the daily price is where sellers hide. And the effect is a first-pass, feeling-level thing: it’s strongest on the fast read and weaker the moment someone sits down to actually compare. So use it to make a genuinely good deal legible, not to make a bad one survivable. The first is design. The second gets found out.

It’s also worth saying where the science is firm and where it’s soft. That the brain over-weights raw counts and under-weights the denominator is well replicated across a lot of tasks. That any specific pricing frame will move your specific metric is not something a jelly-bean study can promise you — treat the transfer as a hypothesis worth testing on your own numbers, not a law. The reliable part is the direction of the pull. The size of it is yours to measure.

The odds were plainly better in Bowl A. People reached for Bowl B because they could see more of what they wanted in it. Your users are doing the same read on every number you show them — and you get to decide which number that is.