We tend to assume more options mean better decisions, but researchers who set up a grocery-store tasting table found shoppers offered 24 jams were a tenth as likely to buy as those offered 6 — and less satisfied when they did

You went in for olive oil. You’re now four minutes into the olive oil aisle, holding two bottles, comparing them on criteria you invented ninety seconds ago, and you’re going to leave with neither.

Nobody taught you to do this. It happens anyway, to most people, most weeks.

We walk into stores carrying a background assumption: more options means a better chance of finding the right one. Twelve olive oils beats four. Forty streaming shows beats six. A bigger menu makes a better restaurant.

It’s a reasonable belief, and across plenty of decisions it holds up.

But there’s a famous experiment that punched a hole in it, and the hole has never quite been patched. It ran at a grocery store, which is a stranger psychological environment than we give it credit for, and it involved jam.

The table with 24 jams pulled a bigger crowd and sold a tenth as much

In the late 1990s, Sheena Iyengar and Mark Lepper set up a tasting booth at Draeger’s, an upscale grocery in Menlo Park, California. The store was chosen deliberately, because it already carried hundreds of jams and shoppers there were used to in-store tastings. Nothing about the booth would have looked out of place.

The display rotated. Some hours it held six jams. Other hours it held 24. Everyone who came near got a coupon for a dollar off any jar of that brand.

The first result matched the more-is-better assumption exactly. The 24-jam table drew the bigger crowd. Roughly 60 percent of passing shoppers stopped at it, against about 40 percent for the small one. Variety is a magnet, and it worked as a magnet.

Then the second result arrived.

Of the shoppers who stopped at the six-jam table, about 30 percent later redeemed the coupon and bought a jar. Of the shoppers who stopped at the 24-jam table, about 3 percent did.

Ten times the conversion, from a display that attracted fewer people to begin with. The finding became the anchor study of a 2000 paper in the Journal of Personality and Social Psychology arguing that a limited array of six options can be more motivating than an extensive one, and it has been quoted in business books, TED talks and marketing decks ever since.

The authors themselves flagged the obvious weakness. This was a field study. Nobody was randomly assigned to anything, and shoppers chose for themselves which table to approach.

Quite possibly the people who wandered over to gawk at 24 jams were browsers, while the ones who stopped at six had walked in wanting jam. Which would mean the display sorted the crowd instead of changing it.

The satisfaction result came from a lab study with chocolate

This part gets flattened in almost every retelling, including probably the version you’ve heard.

The jam study measured one thing: whether a coupon got used. It had no way to ask anyone how they felt about the jar they took home. The claim that extensive choice leaves people less satisfied comes from a different experiment in the same paper.

In that one, Iyengar and Lepper brought 67 participants into a lab and offered them either six exotic chocolates or 30. Everyone picked one and ate it.

People choosing from 30 said they enjoyed the process more. They also found it harder and more frustrating. And when asked to rate the chocolate they had actually chosen, the large-assortment group came in lower: about 5.5 on a seven-point scale, against 6.3 for the small-assortment group.

Then came the part that’s hard to argue with, because it involved a real trade. At the end, participants could take a box of chocolates or cash as compensation. In the six-chocolate condition, 16 out of 33 took the chocolates. In the 30-chocolate condition, 4 out of 34 did.

Same product, same experimenters, wildly different appetite for it depending on how many versions of it people had just looked at. A third study, using optional essay assignments, found students were more likely to turn one in when offered six topics than 30, and that the essays came back better.

So the satisfaction finding is real. It belongs to a lab room with 67 undergraduates in it, and that’s where it should stay when people quote it.

The same pattern showed up in 800,000 people’s retirement accounts

Chocolate is low stakes. If the effect were only about hedonic trinkets it wouldn’t matter much.

A few years later, Iyengar worked with Gur Huberman and Wei Jiang on something that mattered a great deal. They obtained Vanguard records covering nearly 800,000 employees across 647 retirement plans in 69 industries, and looked at whether the size of the fund menu tracked with whether people enrolled at all.

It did, in the direction you’d now expect. Plans offering a handful of funds had higher participation than plans offering ten or more. The estimated cost of each additional fund was small, somewhere around a fifth of a percentage point of participation, but it accumulated across menus that sometimes ran to dozens of options.

Declining to enroll usually means walking away from an employer match worth thousands of dollars a year. That’s a different order of consequence than a jar of jam.

The caveat is significant and worth explaining. Nobody assigned these companies their fund menus, so the data is observational. The Department of Labor’s evidence clearinghouse reviewed the study and rated the quality of its causal evidence as low, on the grounds that companies offering huge fund menus may differ from other companies in a dozen ways nobody measured.

Suggestive, at a scale that’s hard to ignore. Not proof.

A meta-analysis of 50 experiments put the average effect at almost zero

Here’s where the story stops being a clean one.

By the late 2000s, researchers had been running versions of the jam experiment for a decade, and the results were all over the place. Benjamin Scheibehenne tried a direct replication in an upscale German supermarket with 504 shoppers and found nothing. A jelly bean version failed. A chocolate replication by Rainer Greifeneder failed.

So Scheibehenne, Greifeneder and Peter Todd gathered everything they could find, published and unpublished, and ran the numbers. Fifty experiments, 63 conditions, 5,036 participants.

The mean effect size was d equals 0.02, which is effectively nothing.

Two other findings from that analysis deserve to sit next to it. Published studies were somewhat more likely to report the effect than unpublished ones, which is what a mild publication bias looks like. And more recent experiments were less likely to find it than early ones, a pattern that shows up often after a striking result gets famous.

The authors were careful about what this did and didn’t mean. The variance between studies was high, high enough that something real is probably moving underneath the average. They simply couldn’t identify any condition that reliably produced the effect.

The effect shows up when you don’t already know what you want

Alexander Chernev, Ulf Böckenholt and Joseph Goodman took the same puzzle and asked a different question. Forget whether choice overload happens on average. What has to be true for it to happen at all?

Their 2015 meta-analysis covered 99 observations across 7,202 participants, and it identified four factors that reliably moderate the effect: how complex the choice set is, how difficult the decision task is, how uncertain your preferences are, and whether you’re actually trying to buy something or just browsing.

Hold those four steady and the effect appears consistently. Average across them, as the earlier meta-analysis did, and the positive cases and the negative cases cancel each other out to something near zero.

The two teams have disagreed about this in print, and the disagreement is genuine. But the practical upshot survives either way, and it’s the least glamorous sentence in this literature.

People who know what they like do better with more options. People who don’t know what they like do worse.

Iyengar and Lepper’s studies were built, deliberately, to catch the second group. They used exotic jams and unfamiliar chocolates precisely so that nobody would walk in with a favorite. That was good experimental design. It also narrows considerably who the finding was ever about.

Sorting the options into categories does most of the work that removing them would do

If cognitive effort is what’s doing the damage, then anything that lowers the effort should help, and the raw count drops to a secondary detail.

That’s roughly what the follow-up work found. Cassie Mogilner, Tamar Rudnick and Iyengar found that a larger assortment reduced satisfaction only when the options weren’t pre-sorted into categories. Add the categories back and the penalty went away, even though the number of options hadn’t changed. Other work found the effect appeared under time pressure and faded without it, or appeared when options were described on many attributes and faded when they were described on one.

Which is why the retail lesson people took from the jam study, cut your product line, misfires most of the time. Grocery chains that stripped back assortments have often seen sales fall.

For a person standing in an aisle, the equivalent move is to arrive with a rule already in hand. Decide before you look what you’re optimizing for, take the first thing that clears the bar, and stop. The research on maximizing versus satisficing has been saying this for two decades, and it’s the same instinct that lets some people be perfectly content with good enough while others keep auditing the road they didn’t take.

Across six countries, having too little choice was the more common complaint

One more result reframes the whole conversation, and it comes from Iyengar herself, alongside Elena Reutskaja, Nathan Cheek and Barry Schwartz.

They surveyed 7,436 people across Brazil, China, India, Japan, Russia and the United States, asking about six different domains of everyday choice, and measured two things: how often people felt they had more options than they wanted, and how often they felt they had fewer.

In most domains in most of those countries, having too little choice was the normal experience. The United States was the lone place where overload was commonly reported. And deprivation showed a stronger link to dissatisfaction than overload did.

The paradox of choice, in other words, is substantially an American paradox, built on American samples, describing a condition most of the world would happily trade for.

The number of options is rarely what’s actually costing you

So what’s left of the jam study, twenty-five years on?

The field result stands as reported and has never been directly replicated. The satisfaction finding is real and belongs to a lab study with 67 people. The retirement plan data is large, suggestive and correlational. The average effect across the whole literature sits near zero, and the effect appears reliably only under conditions the original studies were built to create.

That’s a lot of hedging for one of the most-cited findings in consumer psychology. It’s also the honest version.

What survives all of it is smaller than the slogan and more useful. Choosing costs something. The cost goes up when the options are hard to compare, when you’re rushed, when you have no idea what you want, and when you feel you ought to be able to find the best one rather than a fine one.

Notice what’s missing from that list. The count doesn’t appear on it. What the count does is make everything else on the list more expensive.

Which puts the fix in the olive oil aisle somewhere other than the aisle. You decide before you get there that you want something Italian under fifteen dollars, you take the first one that qualifies, and you walk away without checking whether the one behind it was better.

It will be, sometimes. That was always going to be true. The checking is the part that costs you.