The cheap cardboard stand that won

Two stores. Same retailer, same eleven days, same item, same price cut down to the cent. Matched on every measurable thing we could match them on, which is the only way to run a test anyone will believe.
Store A kept the display the way the brand had always done it. End of aisle, every variant lined up, a sign shouting the percentage off.
Store B changed three things and spent nothing. Turned the stand so it faced people coming down the aisle instead of sitting parallel to them. Cut the range from seven variants to two. Replaced the percentage sign with five plain words naming the use: everything for dinner tonight.
Store B sold over forty percent more on the line. Basket value rose with it, which matters more than the line does, because a display that quietly steals from the shelf two aisles over has not made anybody any money.
When I show those numbers somebody always pushes back that the deal was identical, so how. Exactly. The deal was identical. The architecture around it was not.
What decision architecture means
The frame that holds a choice. Where the options sit, how many appear, what gets said about them, and what she has to do next.
It isn’t merchandising and it isn’t design. It’s the structure a decision happens inside, and it changes outcomes without changing the thing being decided. Price was identical in both stores. So were the product, the packaging, the staff, the days of the week. What moved was the cost of saying yes.
Each of Store B’s changes lowered that cost somewhere specific.
The angle was the cheapest and did the most. A stand sitting parallel to the aisle presents its edge to someone walking past, which is to say it presents almost nothing. Turned, it presents its face. Same stand, same stock, roughly double the visible area at the moment she’s moving at walking speed with her eyes at shelf height.
The five words did something the percentage sign couldn’t. “Everything for dinner tonight” names a job. The sign is no longer announcing that the price is lower, which she half-expects and half-distrusts, having seen that sign every week of her adult life. It’s telling her this solves the problem she walked in with. She didn’t come to the store to save eighteen percent. She came because people are going to want feeding at seven.
Then there’s cutting seven variants to two, which is the move people argue with, and the one where I have to walk something back.
The study I quoted for a decade
For years the slide was the jam experiment. Sheena Iyengar and Mark Lepper set up a tasting booth at a grocery store near Stanford on two weekends, once with twenty-four jams on the table and once with six. The big display pulled more people in. Sixty percent of passers-by stopped at it, against forty percent at the small one. Then the numbers turn over. Of the shoppers who sampled from the six-jam table, thirty percent bought a jar. From the twenty-four, three percent did (Iyengar and Lepper, Journal of Personality and Social Psychology, 2000).
Clean story. Great slide. I used it in rooms across four countries.
It doesn’t hold up the way I was telling it. In 2010, Benjamin Scheibehenne, Rainer Greifeneder and Peter Todd pooled sixty-three conditions from fifty experiments, about five thousand people, and found the average effect of offering more options was indistinguishable from zero. Not small. Zero, with wild variation between studies and a trace of publication bias, the unpublished experiments being disproportionately the ones that found nothing. Jason Collins has made the blunter case that the jam study ought to be retired or heavily qualified, and he’s right.
So why did cutting seven variants to two still work in Store B?
Because choice overload was never an accurate description of what happens at the end of an aisle. Overload is a laboratory idea about mental strain while comparing alternatives. A woman with a trolley, a list and a child is not comparing alternatives. She is spending time she doesn’t have. Seven options aren’t confusing to her. They’re slow. Two she can resolve without breaking stride. Seven is a small project, and small projects get postponed, which on a shop floor means walking on.
That distinction decides where you apply this. On a primary shelf she came to on purpose, range helps, because she arrived to find her thing and wants it in stock. On a secondary display she passed by accident, range is a tax. Half the reason the literature looks so contradictory is that it keeps averaging those two situations into one number.
The time it didn’t work
Same three moves, a beverage client, convenience format. We turned the stand, cut the range, rewrote the sign to name an occasion. Flat. Fractionally down, well inside the noise.
What we’d missed is that nobody in that format walks an aisle. They come through the door, scan one wall, pay, leave, in under a minute. There was no “coming down the aisle” to face the stand toward, so the best of the three moves was answering a question the store never asked. The occasion we named, something about an afternoon break, was one she had settled before she touched the door handle.
Decision architecture isn’t a kit you apply. It’s a question about where in her trip the decision actually gets made, and that is a different place in a hypermarket than in a corner store.
The part that’s uncomfortable
The beautiful display lost. It had better photography, a proper structure, and it had been signed off by people with taste, including me on several occasions.
Taste is not the variable. I’ve stopped being able to predict these tests by looking at the two displays, and I’ve been doing this since 1989. The cheap cardboard stand wins often enough that I now assume nothing until the stores report.
Which is the argument for testing rather than arguing. The forty percent isn’t the interesting number. The interesting number is how certain the room was beforehand, how wrong that certainty turned out to be, and the fact that being wrong cost nobody in that room anything at all.
Try it on something you’re running now
Pick a display that’s underperforming.
Walk the aisle in the direction she walks it, at her speed, and look at the stand. If you’re seeing its edge, so is she. This takes a morning, costs nothing, and I have never once done it without finding at least one stand presenting itself to a wall.
Count the variants, then ask who put each one there and write the reason down. Range on a secondary display is almost always there for internal reasons, to keep a brand manager calm or to move stock that didn’t sell. If the reason you wrote down is an internal one, it’s costing you conversions.
Read the sign and decide whether it names a price or a job. If it names a price, write a version that names a job and run the two against each other instead of settling it in a meeting. You’ll be wrong about which one wins. So will I, and I’ve had thirty-odd years of practice.
None of that needs budget. It needs somebody to go to the store and look, which is the whole method and the hardest part to get funded.
Decision architecture is chapter fourteen of Shopper Marketing in the Era of AI. It follows Seen, safe, easy. Next, where the insight behind a test like this comes from: Ask the data, ask the shopkeeper, ask her.