
← ResearchResearch note12 min read
Bundles that respond
A bundle that treats every shopper the same is answering a question nobody asked. The research on how choices are presented explains why.
Abstract
Does it matter how a bundle lays out its choices, or only which products are in it? Four decades of work on choice architecture say the layout matters: the number of options, their order, what is preselected and how they are grouped all change what people pick, and a field experiment with 750 car buyers showed that the order of steps in a configurator changed both the car and its price. The same research is honest about its limits, and the famous claim that more choice always hurts has not held up. For a bundle, the practical reading is that the next step should depend on who is building and what they have already chosen.
Key findings
- 01There is no neutral way to present a bundle: every layout has an order, a number of options and a starting state, and each one moves choices.
- 02More options do not reliably reduce sales; a meta-analysis of 50 experiments found an average effect of virtually zero, and later work shows overload appears under specific conditions.
- 03In a field experiment at German car dealerships, changing only the order of eight configurator steps changed how often buyers accepted the default and moved the average price by about 1,480 euros.
- 04People value what they have put together themselves, but only when the build goes well and feels like their own work.
- 05A bundle that shows, hides, requires or limits products according to earlier choices is a practical way to keep each step easy for the person in front of it.
Watch someone build a bundle and you will see that the second choice depends on the first. The person who starts with decaf is not building the same box as the person who starts with espresso beans. A page that shows both of them the same next step is ignoring the only thing it knows about them.
Most bundle pages do exactly that. Every shopper sees every product, in the same order, with the same things ticked, whatever they have already put in the box. The usual defence is that this is the neutral option: show everything and let people decide. The research we review here says there is no neutral option. A merchant who has not thought about how the choices are presented has still presented them, and the presentation is doing some of the choosing.
There is no neutral layout
The idea has a name. Richard Thaler and Cass Sunstein, in their 2008 book Nudge, called it choice architecture: the design of the setting in which people choose. Four years later a group of twelve decision researchers led by Eric Johnson set out the working parts in a paper titled “Beyond nudges: Tools of a choice architecture”. Its first sentence is the whole argument: the way a choice is presented influences what a decision-maker chooses.1
Johnson and his co-authors sort the tools into two groups. Some structure the task: how many alternatives to show, whether one is preselected, and what help is offered in narrowing down. Others describe the options: how they are grouped and which of their attributes are put in front of the chooser.1 A bundle builder uses every one of these tools whether its designer meant to or not. It has a number of products per step, an order of steps, a state it starts in and a way of grouping products. The useful question is what each of those does.
How many options: the jam study, told honestly
The best known finding in this field is also the most overstated. In 2000 Sheena Iyengar and Mark Lepper set up a tasting table for Wilkin and Sons jams in an upmarket Californian grocery. On some hours it held 24 flavours, on others 6. The large display drew more people: 60% of the 242 shoppers who passed it stopped, against 40% of the 260 who passed the small one. But of those who stopped, about 30% of the small-display group went on to buy a jar, against 3% of the large-display group.2
That result became a rule of thumb: offer less and sell more. It deserved more caution than it got. In 2010 Benjamin Scheibehenne, Rainer Greifeneder and Peter Todd gathered every comparable experiment they could find, published or not: 50 experiments with 5,036 participants in total. The average effect of offering more options was, in their words, virtually zero, with a great deal of variation from one study to the next.3 Some studies found that large assortments hurt. Others found that they helped.
A zero average with wide variation does not mean nothing happens. It means something else decides which way the effect goes. In 2015 Alexander Chernev, Ulf Böckenholt and Joseph Goodman analysed 99 observations covering 7,202 participants and named four conditions under which a large assortment does lead to overload: the set is complex (options are hard to compare, or none stands out), the task is difficult (time pressure, many attributes to weigh), the shopper is unsure what they want, and the shopper’s aim is to get the decision over with quickly. With those four accounted for, the overall effect of assortment size was significant.4
For a merchant this is a more useful finding than the original. The number of products in a step is not the problem in itself. A regular customer who knows the range will pick from thirty teas without difficulty. A first-time gift buyer facing the same thirty is in exactly the situation Chernev and colleagues describe: unsure of their preferences, short of time and trying to finish. The same page is easy for one of them and hard for the other.
What is preselected
A default is the option a person ends up with if they do nothing. The classic demonstration is Eric Johnson and Daniel Goldstein’s 2003 paper “Do Defaults Save Lives?”. In an online experiment, 161 people were asked whether they would be organ donors. When the form assumed they were not donors and they had to opt in, 42% consented. When it assumed they were and they had to opt out, 82% did. With no default at all the figure was 79%.5 These were stated choices in an experiment, and the gap between them came from nothing more than which box started ticked.
Defaults are among the most replicated effects in the field, though they are not uniform. A 2019 meta-analysis by Jon Jachimowicz and colleagues pooled 58 studies with 73,675 participants and found a sizeable average effect (a standardised difference of 0.68) alongside wide variation. Defaults worked better in consumer settings than in environmental ones, and better when people read the default as a recommendation from whoever set it, or came to feel it was already theirs.6 So a preselected option carries a message as well as a convenience: the shop suggests this.
Order and position
Position within a list matters too. Eran Dayan and Maya Bar-Hillel rearranged restaurant menus in a laboratory study and in a real café in Tel Aviv. Items placed at the beginning or the end of their category were up to twice as popular as the same items placed in the middle.7 The products, the prices and the customers were unchanged. Only the position differed.
The configurator experiment
The study closest to a multi-step bundle is by Jonathan Levav, Mark Heitmann, Andreas Herrmann and Sheena Iyengar, published in the Journal of Political Economy in 2010.8 It is worth describing in detail, because it tested the order of steps in a real product configurator with real purchases.
The first study was a framed experiment. Seventy-three students designed a tailored suit by working through booklets of swatches: 100 suit fabrics, 50 shirt fabrics, 42 ties, and so on down to 5 linings. In each booklet one option was marked as the tailor’s standard recommendation, which served as the default. Half the participants went from the largest booklet to the smallest, half the other way. Those who began with the 100 fabrics accepted the recommendation more often overall, and more often as they went on. Those who began with the 5 linings accepted it at a roughly steady rate throughout.8
The next two studies moved to the dealerships of a large European car maker in Germany, with 750 customers who had come in to order an entry-level saloon. They configured their car on the manufacturer’s own software and then signed a purchase order for what they had configured. The researchers changed one thing: the order of eight steps, out of 67 configurable attributes. One group began with interior colour (56 options) and ended with the style of gearshift knob (4 options). The other group went in reverse. Every step had a preselected option, which for all but one attribute was the cheapest.8
The suit result repeated. Buyers who met the big decisions first became steadily more likely to accept whatever was preselected as they moved through the sequence. Buyers who started with the small decisions did not. The authors read this as a kind of fatigue: weighing 56 colours uses up some of the capacity needed to weigh what comes next. Total configuration time was the same in every group, just under 30 minutes, so the buyers who took the defaults were not simply rushing.8
The effect on price is the detail most summaries get wrong. In the first car study the many-options-first group paid more, not less: 37,290 euros on average against 35,808, a difference of about 1,480 euros. The reason was the engine. It had 25 options and a cheap default, so that group chose it early, while they were still weighing alternatives, and more of them traded up. In the second car study the engine was moved out of the eight reordered steps and replaced with a radio. The order effect on defaults was weaker and the price difference disappeared. The authors’ conclusion was that price depends on which step sits where, not on one ordering being better. In both car studies, buyers who began with the small decisions reported being more satisfied.8
The step a shopper reaches third is not the same decision it would have been had they reached it first.
Grouping, and the value of building it yourself
How options are grouped has its own effect. Cassie Mogilner, Tamar Rudnick and Sheena Iyengar found that simply dividing an assortment into categories made shoppers who were new to the product perceive more variety and feel more satisfied with what they chose, whatever the category labels said. Shoppers who already knew the product well did not need the help.9 It is another case of the same page working differently for the newcomer and the regular.
There is also a reason to let people build at all, when a prebuilt box would be simpler. Nikolaus Franke, Martin Schreier and Ulrike Kaiser ran five experiments on products that customers designed with an online toolkit. In one, people who had reproduced a T-shirt design themselves bid 6.85 euros for it on average, against 4.75 euros from people offered the identical shirt off the shelf: over 40% more for the same object. The premium ran through a feeling of accomplishment, and it depended on the result turning out well and on the person feeling the work was their own.10
Michael Norton, Daniel Mochon and Dan Ariely found the same pattern with physical assembly and named it the IKEA effect. Participants who assembled a plain storage box bid 78 cents for it against 48 cents from those handed one ready-made, a 63% premium. When people were stopped before finishing, or built something and then took it apart, the premium went away.11 A build-your-own bundle is worth more to the person who built it, provided they get to the end and the box feels like theirs.
| Lever | Study and setting | What was found |
|---|---|---|
| Number of options | Jam tasting table in a grocery, 6 or 24 flavours2 | About 30% of those who stopped at the small display bought, against 3% at the large one |
| Number of options | Meta-analysis of 50 experiments, 5,036 participants3 | Average effect of more options virtually zero, with wide variation between studies |
| Number of options | Meta-analysis of 99 observations, 7,202 participants4 | Overload appears with complex sets, difficult tasks, uncertain preferences and a wish to minimise effort |
| Default | Online organ donation experiment, 161 respondents5 | 42% consented when they had to opt in, 82% when they had to opt out |
| Default | Meta-analysis of 58 studies, 73,675 participants6 | Sizeable average effect (d = 0.68), stronger in consumer settings |
| Position in a list | Restaurant menus, laboratory and café7 | Items first or last in their category were up to twice as popular as in the middle |
| Order of steps | Car configurator at German dealerships, 750 buyers8 | Many-option steps first raised default acceptance later; average price moved by about 1,480 euros in the first study |
| Grouping | Field and laboratory experiments on categorised assortments9 | Categories raised perceived variety and satisfaction for unfamiliar shoppers |
| Self-design | T-shirt design toolkit, bids with real money at stake10 | 6.85 euros bid for a self-designed shirt against 4.75 for the identical shirt off the shelf |
A worked example: the tea box
Take a tea merchant with a four-step box: choose a size, choose teas, add an infuser, add a card. The catalogue has thirty teas. Two people arrive on the same afternoon.
The first is a regular. She knows the range, wants her usual six and resents anything that slows her down. Thirty teas on one screen suit her well. On the published evidence she is at little risk of overload, because she is certain of what she wants.4 Categories add little for her.9
The second is buying a present for a colleague and does not drink much tea. For him the same screen meets three of the four overload conditions at once.4 He would be helped by teas grouped into a few plain categories, by a shorter list, and by meeting the small decisions before the big one. If the first thing he does is say the box is a gift, that is the most useful fact the page will ever have about him.
A static page has to pick one of these people and fail the other. A responsive one does not. With rules, the merchant can say: when the gift option is chosen, show the gift selection and the card step, and hide the bulk pouches. When someone picks loose-leaf tea, require an infuser or show the step that offers one. When a caffeine-free tea goes in first, bring the other caffeine-free teas forward and hide the breakfast blends. Limit the strongest teas to two in a box of six. Each rule turns a hard screen into an easier one for a particular shopper, and leaves the regular alone.
What is not known
The evidence has gaps, and some of them are large.
- •The centrepiece study is about cars and suits. They are expensive, rarely bought and configured over half an hour with a salesperson nearby. Nobody has published the same experiment on a six-item box that takes three minutes. The direction of the effect may carry over. The size may not.
- •Levav and colleagues explain their result as mental fatigue. Their data establish the behaviour (order changes default-taking) more firmly than the mechanism, and their second car study found a weaker effect than the first.8
- •Defaults can backfire. In a 2020 study of sequential choices, Bas Donkers and colleagues found that accepting a default is a passive act that weakens the carry-over from one choice to the next, so setting expensive defaults to push people upmarket can leave a firm worse off.12
- •The jam study is a single field experiment from one shop, and the meta-analyses show how far a single result can sit from the average.3 The same caution applies to any one study in this article, including the ones we find most persuasive.
- •Most of these studies measure one purchase. For a subscription box, what matters is whether the customer is still happy with their choices at the third delivery. The satisfaction findings point the right way, but they were measured at the point of sale.8
There is a question of intent as well. The car study’s authors note that a firm could use ordering to exploit tired customers.8 The same tools can make a choice easier or steer it somewhere the shopper would not have gone. We think the test is whether the shopper would thank you for the rule if they could see it. Hiding teas that clash with what is already in the box passes. Preselecting the dearest option at the moment people are least likely to question it does not, and on the evidence above it may not even pay.12
What it means for a bundle
Three things follow from the research. First, the layout of a bundle is a set of decisions with measurable effects, so it is worth making them on purpose. Second, the right layout differs by shopper, because the conditions that make a choice hard (uncertainty, haste, unfamiliarity) vary from one person to the next. Third, the best information about which shopper is in front of you is what they have just chosen.
In Kitenzo this is the job of the Conditions Engine, which is live. A merchant describes the behaviour once, in the admin: show, hide, require or limit products depending on what has been chosen, and apply rule-based pricing and discounts. Every storefront that renders the bundle follows those rules, including pages a developer has coded by hand, where support is in beta.
It is the most hands-on of the things we work on, because the merchant writes every rule. The research gives good grounds for what to write: fewer and better-grouped options for the uncertain, small decisions before large ones, and nothing on screen that conflicts with what is already in the box. Whether a given rule helps a given shop is an empirical matter, and a rule can be tested like any other change by running two versions of a bundle side by side. What interests us next is the other direction: noticing which responses a bundle is missing and suggesting them. That work is in the lab and is not something merchants have today.
References
- 1Johnson, E. J., Shu, S. B., Dellaert, B. G. C., Fox, C., Goldstein, D. G., Häubl, G., Larrick, R. P., Payne, J. W., Peters, E., Schkade, D., Wansink, B. and Weber, E. U. (2012). Beyond nudges: Tools of a choice architecture. Marketing Letters, 23(2), 487-504.
- 2Iyengar, 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.
- 3Scheibehenne, 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.
- 4Chernev, A., Böckenholt, U. and Goodman, J. (2015). Choice overload: A conceptual review and meta-analysis. Journal of Consumer Psychology, 25(2), 333-358.
- 5Johnson, E. J. and Goldstein, D. (2003). Do Defaults Save Lives?. Science, 302(5649), 1338-1339.
- 6Jachimowicz, J. M., Duncan, S., Weber, E. U. and Johnson, E. J. (2019). When and why defaults influence decisions: a meta-analysis of default effects. Behavioural Public Policy, 3(2), 159-186.
- 7Dayan, E. and Bar-Hillel, M. (2011). Nudge to nobesity II: Menu positions influence food orders. Judgment and Decision Making, 6(4), 333-342.
- 8Levav, J., Heitmann, M., Herrmann, A. and Iyengar, S. S. (2010). Order in Product Customization Decisions: Evidence from Field Experiments. Journal of Political Economy, 118(2), 274-299.
- 9Mogilner, C., Rudnick, T. and Iyengar, S. S. (2008). The Mere Categorization Effect: How the Presence of Categories Increases Choosers’ Perceptions of Assortment Variety and Outcome Satisfaction. Journal of Consumer Research, 35(2), 202-215.
- 10Franke, N., Schreier, M. and Kaiser, U. (2010). The “I Designed It Myself” Effect in Mass Customization. Management Science, 56(1), 125-140.
- 11Norton, M. I., Mochon, D. and Ariely, D. (2012). The IKEA effect: When labor leads to love. Journal of Consumer Psychology, 22(3), 453-460.
- 12Donkers, B., Dellaert, B. G. C., Waisman, R. M. and Häubl, G. (2020). Preference Dynamics in Sequential Consumer Choice with Defaults. Journal of Marketing Research, 57(6), 1096-1112.


