Skip to Content Skip to Footer
Want to Convert Shoppers into Buyers? Internal Categorizations May Be the Key

Want to Convert Shoppers into Buyers? Internal Categorizations May Be the Key

Serwaa Karikari and Maha Alshammari

woman in grocery store

Journal of Marketing Research Scholarly Insights are produced in partnership with the AMA Doctoral Students SIG – a shared interest network for Marketing PhD students across the world.

One of the aims of organizing related products into categories is to convert shoppers into buyers. To achieve this conversion, retailers must consider both a category’s shelf layout (product arrangement) and its assortment (size and composition). Thus, retailers face the uphill challenge of determining an optimal shelf layout taking into account both these factors. A recent article by Robert Rooderkerk and Donald Lehmann, published in the Journal of Marketing Research, investigates the best approach to optimizing external categorization using both field and laboratory experiments. The article focuses on categories in which products are presented together on multiple shelves in offline contexts, specifically examining consumer packaged goods and consumer electronics categories.

Past research shows that when congruency exists between consumers’ internal categorization and a shelf layout’s implied external categorization, consumers think that there is greater variety, and their satisfaction with their choice(s) increases as a result. Drawing on this, the authors propose the following formula for effective shelf layout design: (1) having a valid measurement of consumers’ internal categorizations, (2) making accurate inferences about how these internal organizations interact with external categorizations to influence purchase behavior, and (3) using a methodology that optimizes the external categorizations using points 1 and 2 as input.


Through the use of experiments, the authors find that when congruency exists between shelf layout and consumers’ product categorizations, these consumers indeed perceive increased variety but lower complexity. Existing research has measured consumers’ internal categorizations using “decision trees,” which has not been completely effective. In contrast, this study builds on past research by using improved measures of internal categorizations, allowing for the generalization of the effect to physical assortments.

We corresponded with the authors to gain some additional perspective on the motivations underlying their study as well as other key insights.

Q: In your article, you mention that though previous studies have examined the size and composition of assortments, you examined shelf layout. What inspired this?

A: Shelf planning is part of the broader field of assortment optimization, Robert’s area of expertise. Don has conducted research on the effect of choice set size and composition as well as stocking level, which is quite related.

Assortment optimization is a fascinating field because it combines insights from marketing and psychology to understand how consumers perceive assortments and make purchase decisions using econometric techniques and optimization algorithms to model these decisions and arrive at optimal decisions. It embodies a nice combination of theory and advanced analytics at the intersection of marketing and operations.

The problem was introduced to us by a large biscuit manufacturer in the Netherlands. They were asked by a national grocery retailer to advise them on how to increase category-level conversion (% of shoppers that buy from the category) without changing the assortment composition. That implied changing the assortment layout, which seemed like a very interesting research problem. How could we change the assortment layout—the arrangement of all products for sale in the category—so that consumers would perceive more variety and/or experience less complexity, two likely drivers of conversion? We believed that coming up with a method to answer this question would greatly benefit retailers, manufacturers, and consumers.

Q: You employed both field and laboratory experimental designs. How did you arrive at this decision?

A: Each has its own strengths in our setting and their combination provides convergent evidence for the main findings.

The field experiment involved intercepting shoppers in a supermarket and asking them to take part in a shopping exercise using an imbalanced assortment that reflected reality very well. We aimed for high external validity. But the field experiment did not use incentive alignment and used posters, not real shelves. Bringing things to the lab allowed us to test whether the effect would replicate with a different stimulus and sample, and focus more on internal validity. Here, we used incentive alignment and balanced assortments. Moreover, the digital experiments allowed us to use highly realistic virtual shelves with pack shots of actual products to heighten the realism.

Across the three studies (two in the lab, one in the field) we were able to offer variation in (1) sample, (2) stimulus, (3) balance of the product design, (4) measurement of the internal categorizations (sorting vs. planogram task), (5) format of the shopping task (free choice vs. forced choice), (6) variation in the orientation of the grouping on shelf (horizontal vs. vertical), and (7) incentive alignment.

Q: Your paper mentions that you experienced methodological challenges as a result of your combination of between- and within-subjects experimental design, among others. Could you elaborate on this?

A: In the field study we use a within-subjects design that fully counterbalanced the order of layouts that participants were shown. This is a good way to increase sample size (obtaining three observations per consumer) if you control for individual differences in the statistical analyses (something you hope random allocation takes care of in a between-subjects design). Still, within-subjects designs could lead to carryover effects between conditions, which we statistically correct for in one of our robustness checks. We used larger samples in combination with between-subjects designs for the lab studies in which there are no carryover effects by design.

[Advice for upcoming scholars]: Talk to practitioners and spend time in the problem context…. Define a problem that is relevant to retailers and then maximize rigor when tackling it, not the other way around.


Q: Were there any particularly surprising or unexpected findings?

A: The finding that congruency between a layout and a consumer’s product categorization could increase choice satisfaction had been shown in the context of product lists using a simple measure of internal product categorization based on revealed attribute importance. We showed that internal categorizations correspond less well with these so-called consumer decision trees and found good ways to measure them. With these better measurements that were applicable to two-dimensional planograms rather than product lists, we were able to generalize the effect to physical assortment settings. We also provided a richer depiction of the underlying process, which included complexity perceptions in addition to perceived variety. Instead of seeing variety and complexity as two different constructs, traditional research has simply equated too much variety with complexity. We show they are distinct but related constructs. Finally, we showed how this knowledge can be used to arrive at normative implications about how to best arrange the shelf layout.

What we had not expected was that the effect would be much stronger in the field than in the lab. But then again, in the lab we used more balanced and stylized assortment layouts, whereas in the field, the assortments better represented the messy reality in which different brands, flavors, pack sizes, etc. occur at very different frequencies. We had also not anticipated the differential effect of familiarity in the field (it makes the effect stronger and shifts focus on variety perceptions) and the lab (no effect). This may have been due to differences in the stimuli we used (more complex category of biscuits in the field vs. less complex yogurt snacks in the lab). Future research should dive into this. Finally, we were pleasantly surprised by the added value of optimizing the shelf layout based on our insights compared to the traditional approach based on consumer decision trees.

Q: How should practitioners implement your findings?

A: Shelf layouts that are better tailored to customers allow them to better appreciate the variety of offerings without feeling the limitations of complexity. This increases the chances consumers will buy (benefiting retailers and manufacturers) and increases the satisfaction with the chosen item (benefiting customers). Our paper provides a method for doing exactly that. It describes several ways to measure internal categorization that are straightforward to implement. Part of this is the online planogram tool (build your own shelf) that we have built for this purpose. Second, our research shows how a relatively straightforward and tractable optimization problem can be solved to maximize the average congruency between layout and a consumer’s internal categorization across consumers. Hence, our method is able to deal with heterogeneity in internal categorizations across consumers and still arrive at one single layout that works best across all. It may be even more effective when employed at the segment (or individual store) or individual level (e.g., online).

Q: Could you share any advice or tips you may have for upcoming scholars in this research area?

A: Talk to practitioners and spend time in the problem context (in our case, the store). Define a problem that is relevant to retailers and then maximize rigor when tackling it, not the other way around.

Read the full article:

Rooderkerk, Robert P. and Donald R. Lehmann (2021), “Incorporating Consumer Product Categorizations into Shelf Layout Design,” Journal of Marketing Research, 58 (1), 50–73.

Serwaa Karikari is a doctoral student at Morgan State University.

Maha Alshammari is a doctoral sudent at Morgan State University.