The jam study is often reduced to a design rule: offer fewer choices and more people will buy. The research is more useful when you retain the original comparison and the later uncertainty. A study can motivate a question without answering what should happen on your pricing page.
For an analyst, designer, or product marketer, the immediate job is to distinguish three things: what the original experiment measured, what subsequent reviews concluded, and what your own proposed change still needs to establish. None of those steps requires memorizing a universal limit on the number of options.
What happened in the original jam study
Sheena Iyengar and Mark Lepper’s 2000 paper reported three studies. The best-known was a field experiment at a California grocery store, comparing tasting displays with 6 or 24 jams. The larger display attracted more shoppers, but purchasing among those who stopped was higher for the smaller display: about 30% versus about 3%.
Those percentages describe particular groups in that field setting. They are not a tenfold sales forecast for another business. The paper also studied choices involving essays and chocolates; its design deliberately sought contexts where participants were unlikely to arrive with strong preferences for a specific option.
The intervention changed the tasting display, not the full range available to buy. Shoppers in both conditions could purchase from the entire shelf assortment. Removing subscription plans changes availability as well as presentation, so it is not the same intervention as the jam display.
The result challenges the assumption that a larger assortment must always be more motivating. It does not establish that a smaller assortment must always be better. Attention to a display and the decision to purchase are also different outcomes, which is a useful distinction to retain when describing this study.
Why later reviews changed the simple story
Benjamin Scheibehenne, Rainer Greifeneder, and Peter Todd’s 2010 meta-analysis examined 63 conditions from 50 published and unpublished experiments, involving 5,036 participants. Its average choice-overload effect was close to zero, with substantial variation. The authors did not identify sufficient conditions that consistently explained when the effect would occur.
An average near zero does not mean every individual study found zero. It means the pooled result does not support treating a broadly beneficial effect of reducing choices as the default prediction for every setting. A dramatic individual example is an incomplete summary of a varied literature.
The review’s title is Can There Ever Be Too Many Options? A Meta-Analytic Review of Choice Overload. Keeping it separate from the original jam paper helps avoid mixing the original observation with the later synthesis.
What the 2015 moderator analysis adds
Alexander Chernev, Ulf Böckenholt, and Joseph Goodman reviewed 99 observations involving 7,202 participants. Their 2015 analysis identified four relevant moderators: the complexity of the choice set, difficulty of the decision task, uncertainty about preferences, and whether the decision goal emphasized minimizing effort.
These categories help formulate more specific questions about an interface. They do not constitute a validated four-question diagnostic for a particular website. Four “yes” answers do not produce a conversion forecast, and four “no” answers do not establish that adding options is harmless.
The review also does not provide an optimal number of SaaS pricing tiers. Its contribution is a more conditional account of choice overload, rather than a replacement universal rule. Use the categories to structure an investigation and keep the proposed application separate from the review’s empirical findings.
A worked example: choosing what to investigate on a pricing page
Imagine a fictional software team with five plans. Some users ask which plan fits their needs; others cannot find the feature they want. A stakeholder proposes reducing the page to three plans because of the jam study. This is an illustrative planning example, not a result from a client project or an experiment I ran.
At a design review, the team can first separate the observed problems. Confusion about plan differences, difficulty finding a feature, and difficulty choosing an affordable plan may need different responses. The number of plans is one possible factor, but the observations do not yet identify it as the cause.
The team writes a task: “Help a prospective customer determine which plan supports their required workflow and understand the main limitation before proceeding.” That is more specific than “make the page simpler.” It makes room for an appropriate decision not to buy when no plan fits.
Next, they inspect the information needed for the task. Do the plans use consistent feature names? Are important exclusions visible? Does a comparison require jumping between several sections? Are charges and limits understandable? Can someone explain why a plan fits, rather than merely select a visually prominent button?
A useful first investigation might involve observing people attempt this task and asking them to explain their reasoning. If the trouble concerns unfamiliar terminology, removing plans may not address it. If participants repeatedly compare irrelevant distinctions, a clearer comparison or filtering approach is another hypothesis. Neither response should be declared the winner before evidence exists.
The value of the research is the narrower question it creates. Instead of asking whether “choice overload is real,” the team asks whether a particular change helps the intended users make an informed decision in this interface.
Compare design options without assuming the answer
There are several plausible responses to a difficult choice. Reducing the number of options is one. Explaining distinctions, grouping related options, revealing relevant information at the right time, or helping users state their needs are others.
Each proposal has a possible cost. Removing an option may eliminate the best fit for a smaller group. A recommendation can hide assumptions. A comparison table can become hard to read on a phone. A default can save effort while making an alternative harder to discover. A filter can help only if people understand what it means and can change it.
Write a reason for the specific proposal and the concern that could make it fail. For example: “We propose a comparison organized around three common workflows because participants struggled to map feature names to their intended work. We need to check whether the grouping hides a requirement that does not fit those workflows.” That is an inspectable design hypothesis.
Keep the source claim modest. The choice-overload literature supplies reasons to examine effort and uncertainty; the team’s own observations supply the case for this particular change. The literature cannot substitute for understanding the page or the audience.
Decide what would count as a useful result
A button click is an observable action, but it may be an incomplete measure of a better choice. Someone can click confidently after misunderstanding a restriction. Another person can benefit from a comparison by discovering that none of the options fits.
For a research session, define task success in terms that match the reader’s job: identifying a suitable option, accurately stating a relevant limitation, and explaining the comparison. Record confusion and contradictory cases, not just successful selections. These observations describe the participants and sessions you studied; avoid turning a small qualitative exercise into a population percentage.
For an experiment with sufficient traffic and a suitable design, specify the primary outcome and potential harms before interpreting the result. Depending on the decision, relevant checks might include errors, unsuitable selections, support questions, or later cancellations. Those are proposed measurement choices, not outcomes established by the jam study.
If the available traffic cannot support a useful quantitative comparison, do not force a confident A/B-test conclusion. Improve obvious clarity problems, document what changed, and use appropriately scoped research to reduce uncertainty. A before-and-after increase alone cannot separate the effect of a design change from other changes occurring over the same period.
Keep a source note that survives the next presentation
A useful reading note should contain the study or review, its actual finding, and the boundary of the conclusion. Avoid copying only the memorable percentage into a slide. The number without its denominator and setting is easy to transform into an unsupported business promise.
For the jam study, keep the tasting-display setting and the distinction between attention and purchase. For the 2010 review, keep both the near-zero average and the variation. For the 2015 analysis, keep the moderators and the fact that they do not supply a site-specific forecast.
When presenting a design proposal, label the source of each claim. “The paper reports” belongs beside research evidence. “We observed” belongs beside documented observations. “We propose” belongs beside an untested change. This small distinction helps colleagues challenge the proposal without having to dispute an entire field of research.
The source note at the top of this page includes caveats and research links. You can copy it into reading notes and then add the question you are actually trying to answer. It is a synthesis to inspect, not an endorsement of a particular interface treatment.
Practice with a real choice
The free choice-architecture exercise asks you to examine comparison effort, defaults, and user control. Pick one decision in an interface you know and describe the user’s task before proposing a change.
Use the experiment brief to record the evidence, uncertainty, proposed comparison, and outcome. The WEIRD critique provides a further check: how closely do the populations and tasks in the research resemble the people you intend to help?
Sources
- Iyengar, S. S., and Lepper, M. R. (2000). When choice is demotivating: Can one desire too much of a good thing? Journal of Personality and Social Psychology, 79(6), 995–1006. Original paper hosted by the University of Washington (PDF). DOI: 10.1037/0022-3514.79.6.995.
- Scheibehenne, B., Greifeneder, R., and Todd, P. M. (2010). Can There Ever Be Too Many Options? A Meta-Analytic Review of Choice Overload. Journal of Consumer Research, 37(3), 409–425. Institutional publication record. DOI: 10.1086/651235.
- Chernev, A., Böckenholt, U., and Goodman, J. (2015). Choice overload: A conceptual review and meta-analysis. Journal of Consumer Psychology, 25(2), 333–358. Author-hosted research paper (PDF). DOI: 10.1016/j.jcps.2014.08.002.
FAQ
Is choice overload real?
Some studies find that larger choice sets make decisions harder or reduce willingness to choose. The reviews show why that result should not be generalized into “fewer choices are always better.”
Did the 2010 review prove that the jam study was false?
No. A pooled average near zero does not show that a particular original observation did not occur. It challenges a simple generalization across the literature examined.
What is the best number of pricing plans?
These papers do not establish a universal number. Begin with the decisions your users face, the distinctions they need to understand, and the evidence for a proposed change.
Can the four moderators predict my conversion lift?
No. They help organize questions about a decision context. They are not a validated scoring system or a substitute for evidence from your application.
Should I remove options or add filters?
Neither is automatically better. Identify the observed problem, propose a change that addresses it, and examine the consequences for people whose needs may differ.
How should I cite the jam study in a presentation?
Identify the original setting and measured outcome, then acknowledge that later reviews found a more conditional picture. Separate the research finding from any proposed design recommendation.