Skip to main content


The purpose of this article is to introduce a new tool—the Open Science Online Grocery—for studying the effects of the choice context on purchasing decisions. We first review the features of the tool: a mock online grocery store containing over 11,000 products wherein researchers can modify the choice context (e.g., positioning, labeling, suggestions) and observe resulting choice. Then, we present three studies illustrating how the tool can help assess how changes to labeling, ordering, and positioning affect choice. We find that both ordering and positioning have significant effects on choice while labeling does not. These findings largely align with existing research in field and laboratory settings. We hope this tool proves useful to researchers wanting to test choice context modifications in a relatively affordable and efficient manner.

A healthy diet rich in fruits and vegetables is critical for the prevention of many chronic diseases (Hee Lee-Kwan et al. 2017). However, less than 10% of Americans consume the recommended amount of fruits and vegetables each day (Hee Lee-Kwan et al. 2017). Policy makers have shown increasing interest in nudging consumers toward healthier food choices, including posting nutrition information in restaurants (Shah et al. 2014), mandated front-of-package labels (Hawley et al. 2013), and limiting soda portion sizes (John, Donnelly, and Roberto 2017).

Given the importance of and interest in nudging consumers toward healthier choices, it falls upon researchers to determine which nudges are most effective at changing behavior. However, conducting this research has proven difficult. A randomized control field experiment requires the cooperation of a major retailer. Retailers understandably may be concerned that randomized control trials (RCTs) will affect profit margins, confuse consumers, or require too much of their employees. Often, retailer hesitance prevents large scale randomized field testing; rather researchers end up testing ideas on single product categories, not having an appropriate control condition, or not running the trial at all. Absent retailer cooperation, researchers run choice architecture studies in artificial lab settings or using hypothetical scenarios with a limited range of products. Because of these limitations, it can be difficult to make well-controlled ecologically valid assessments of how the choice context affects consumer food choice.

The possibilities for choice context modifications are also expanding as an increasing number of consumers shop for groceries online (Inmar Intelligence 2020). As consumers move from brick-and-mortar to online retailers, certain interventions become possible or much easier to implement (e.g., product recommendations, following consumers over time, store reorganization). However, it is cost prohibitive for individual researchers to build their own mock online stores, and the issues inherent in retailer collaboration are largely similar in online versus brick-and-mortar retailers.

We developed the Open Science Online Grocery (OSOG) platform to provide a cost-effective and relatively ecologically valid option to researchers studying choice context’s effect on food choice. The OSOG platform mimics a large online grocery store, like Instacart or Amazon Fresh. OSOG allows researchers to easily change the choice context of the store and observe effects on consumers’ behavior. Because all purchases in OSOG are hypothetical, participants are neither charged for nor sent their groceries. However, researchers could fulfill participants’ orders using a local grocer if they so desired. OSOG integrates easily with existing survey software like Qualtrics and can be easily run with online panel participants.

In this article, we first provide a descriptive overview of the store and its functionality. The rest of the article is divided into three illustrations: labeling, sorting, and categories. Within each illustration, we briefly review the existing literature on the nudge before presenting empirical evidence from OSOG. Across three studies, we show that findings from the OSOG largely mirror those from the existing literature suggesting that novel innovations tested in OSOG would be likely to predict real world response to such interventions.

The Open Science Online Grocery

The Open Science Online Grocery (OSOG) platform is a free research tool designed to study the effect of choice context changes in a grocery setting. It is located at OSOG is designed to be accessible to researchers with no web coding experience. However, researchers who are interested in making their own modifications to the store can access the source code at: or contact the lead author to discuss options. Academic and government researchers are free to use, change, and distribute the software including components like photographs, item descriptions, and nutrition information. Industry researchers should reach out to the first author to discuss options.

OSOG consists of a researcher interface—where the researcher can easily make changes to the choice context—and a participant interface—where participants complete a mock grocery shopping trip that integrates the researchers’ choice context changes. We discuss each of these interfaces separately.

The Participant Interface

The participant interface of OSOG looks similar to any other online grocery store that participants may be familiar with (e.g., Instacart, Amazon Fresh; see fig. 1). Currently, the grocery store is populated with 11,096 products sourced from the website of a large grocery retailer in the Southeastern United States in 2018. The store includes unbranded products (e.g., produce, meat, fish), extant branded products (e.g., Cheerios, Skippy), and a fictional store brand called Howe’s. Product offerings will likely expand as more researchers use the store.

Figure 1. 
Figure 1. 

The participant interface of the Open Science Online Grocery (OSOG).

Researchers are also welcomed to expand and update the product offerings in OSOG. For example, researchers with some coding experience could use the original OSOG code to create a version of OSOG using product information sourced from a variety of different sources. For example, OpenFoodFacts ( contains information (i.e., photograph, nutritional information, ingredients) from 660,000 products from around the world. Similarly, researchers could pull products from FoodDB (a database of all products available in UK supermarkets; Harrington et al. 2019) or Ciqual (a database containing nutrition information for unprocessed and prepared foods;

For each product in the store, we sourced price, product size, serving size, nutrition information, ingredients, a photograph of the product, and manufacturer’s product description (where applicable). Please note that the FDA does not require manufacturers to list certain micro- and macro-nutrients on product labels (e.g., vitamin C, monounsaturated fat, calcium), so the database does not contain complete information on these nutrients.

On the main page of the grocery store, participants see an overview of the available products: product photographs, names, sizes, prices, and any front-of-package labels the researcher specifies. If participants click on a product, they see more information including the nutrition facts label, the ingredients, and the manufacturer’s description. Unless the researcher specifies otherwise, participants can add products to their cart from either the product overview page or the detailed product information page.

Participants can browse products by using a navigational menu at the top of the screen. The navigational menu includes nine built-in categories (i.e., produce; meat, dairy, and eggs; bakery, pasta and grains; dry goods, breakfast and spices; pantry; canned; snacks; beverages; frozen foods), each with related subcategories. Participants can also directly search for products by name.

Participants view their cart and check out by clicking a cart icon. On the cart summary screen, participants view all the products in their cart, the price of each item, the before-tax cart total, the tax amount (7.5%), the total price, and any other information the researcher chooses to include.

Data Privacy

OSOG does not collect any identifiable information that is linked with participants’ responses. When participants enter the store, they are prompted to enter their session ID. Session IDs are provided by the researcher and may be unique to each participant (like a participant ID). All the participant’s actions in the store are recorded under their session ID. Researchers who need to link participants’ actions in the store to survey or demographic responses should encourage participants to use the same ID in the store as in a Qualtrics survey.

All participant actions are recorded and kept indefinitely. The researcher could choose to delete the data from the server by deleting the entire experiment.

As is the case on the vast majority of websites, the IP address of all web visitors to OSOG are automatically logged in the web servers that run the store. This information is not made available to researchers. IP addresses are not linked to participants’ actions in the store and are automatically deleted every 30 days. In other words, although it is necessary to collect IP addresses, this information is never made available to researchers. In the unlikely case of a data breach, hackers would have access to IP addresses from the past 30 days but would not be able to link them to any given experimental procedure or any participant actions.

The Researcher Interface

The researcher interface is a point-and-click interface that allows researchers to easily modify the choice context of the store. Figure 2 lists all the currently modifiable components of the store. These components can be modified separately or in conjunction with one another. It is our hope that the capabilities of OSOG expand over time as more researchers use the platform and develop code to meet their specific needs.

Figure 2. 
Figure 2. 

The modifiable choice context components of Open Science Online Grocery (OSOG).

Please examine for more specific instructions about how to use each feature and details on how to integrate OSOG with existing survey software like Qualtrics.

Downloadable Data

Researchers can download two .csv files from OSOG, depending on their research questions. One data set shows all the products in each participants’ cart when they checked out. This output includes nutrition information and price for each product, and summary statistics for the entire basket (e.g., total cost, number of items, total calories). The “All Participant Actions” file shows the researcher everything the participant did in the store: which items participants viewed detailed information on the order in which items were presented on the screen, the order in which items were added to the cart, and whether items were deleted from the cart.

Comparing OSOG to Existing Tools

OSOG is not the first online grocery store to be used in academic research. Other tools (i.e., NUSmart (Finkelstein 2020), Dutch SN Virtumart (Hoenink et al. 2020), and The Virtual Supermarket (Waterlander et al. 2011)) have been used to conduct similar research. Table 1 compares the capabilities of OSOG to the other online grocery stores available to researchers. Of note, OSOG has several advantages compared to the other tools currently on the market. First, OSOG contains more products than any other online grocery tool. Second, OSOG is the only grocery tool containing products from American supermarkets, which may be familiar to North American consumers. Third, OSOG easily allows for the collection of data from online samples. Fourth, OSOG has more flexibility in terms of researcher interventions in terms of both (a) having built-in capabilities that the other tools do not (e.g., suggested products, the ability to alter nutrition labels, cart-level feedback) and (b) being open-source such that researchers can modify the store to add additional capabilities.

Table 1. 

Comparing Open Science Online Grocery (OSOG) to Other Grocery Research Tools

 Grocery store
OSOGNUSmartDutch SN VirtumartThe Virtual Supermarket
StyleOnline Grocery StoreOnline Grocery StoreVRVR
Cost to researchersFreeFreeFreeFree
Number of products11,0004,0001,200700
Country where products are soldUSASingaporeNetherlandsNetherlands
Compatible with an online sampleYesYesNoNo
Add custom labels to productsYesYesYesYes
Change the way products are ordered within a pageYesN/AYesNo
Alter the categories that appear within a store and/or create custom categoriesYesN/AYesNo
Recommend products based on participant actionsYesN/ANoNo
Change the way the Nutrition Facts Label is displayedYesN/ANutrition facts not displayedNo
Set a budget for participantsYesYesYesYes
Provide participants with feedback on the contents of their cartsYesYesNoNo
Change the prices of products in the storeNoYesYesYes
Does the store include brand name products?YesYesN/ANo

In the following sections, we provide three illustrations of the store. We test three choice context interventions that are backed by empirical work: front-of-package labels, within-page sorting, and product categorical organization. By testing well-known interventions, we hoped to replicate the findings of prior field studies thus demonstrating that novel innovations tested in OSOG would be likely to predict real world response to such interventions. Recall that our purpose in running these studies was not to make a robust theoretical contribution, rather to introduce some of the functionalities of the tool and examine whether consumer behavior in the store is concordant with the existing literature.

Illustration I: Front-of-Package Health Labels

Front-of-package (FOP) labels appear on the outward facing side of product packaging in brick-and-mortar stores. In online stores, FOP labels appear directly on or below the product photograph when consumers browse the product overview page. The purpose of these labels is to provide easily understandable information to the consumer. FOP labels can indicate a variety of characteristics (e.g., organic, allergen-free). Research has largely focused on the effects of FOP health labels (e.g., traffic lights, guiding stars, heart check).

The effects of FOP health labels on food choice is highly variable. A Cochrane review (Crockett et al. 2018), a review of field studies (Cadario and Chandon 2020), and a large-scale RCT (Dubois et al. 2021) suggest a small effect. To our knowledge, only three studies have examined the effect of FOP labeling in an online grocer (Sacks et al. 2011; Finkelstein, Ang, and Doble 2020; Shin, van Dam, and Finkelstein 2020). Two of these studies found null effects, and one found an effect when FOP labels were combined with cart feedback (Shin et al. 2020).

The effects of FOP labels also differ depending on the label format, with healthy stars (i.e., the intervention used in study 1) not having a significant effect on nutrient content of foods purchased and consumed (Croker et al. 2020). In this study, we sought to replicate the findings of existing field studies on FOP health labels. We expected to find a small or null effect of health labels in the OSOG.


We recruited 394 participants from Prolific (187 men, 202 women, 5 nonbinary; Mage=32.56, SD=11.00; 252 White, 41 Black, 45 Asian, 7 Southeast Asian, 24 Hispanic or Latinx, 2 Indigenous, 2 Pacific Islander, 19 bi- or multiracial, 2 other; MBMI=26.01, SD=7.94).

Participants were asked to use the OSOG to shop for groceries for the upcoming week. They were asked to buy what they would usually purchase in a normal shopping trip and to shop for other members of their household if that was something they would normally do. To improve ecological validity, all participants were informed that two participants would win the contents of their grocery cart. We could not logistically send the actual contents of grocery carts to the winning participants, but we did provide these participants with a monetary bonus in the amount of their cart total (a fact that all participants were informed of in an end-of-study debriefing).

Participants were randomly assigned to one of two versions of the OSOG. In the control condition, grocery items were unlabeled. In the experimental condition, some grocery items were labeled using the “Healthy Stars” algorithm which rates a food’s nutritional value from 0 to 3 stars. The star point algorithm mimics the Guiding Stars algorithm that is used to label nutritious foods in Canadian grocery stores (Guiding Stars Licensing Company 2012). The details of the algorithm can be found in the appendix. Across both conditions, participants saw grocery items presented in the same order and were prompted to spend $35 before being able to check out.

After completing their hypothetical shopping, participants were asked how many people they imagined shopping for, how much money they spend on groceries in a typical week, and how similar the groceries they “purchased” in the experiment were to their typical purchases from 1 (not at all similar) to 7 (very similar). Then, all participants completed demographic questions and were debriefed regarding the true nature of the bonus.


Shopper Profile

The majority of participants imagined shopping for themselves (27.2%) or themselves and another person (28.9%). On average, participants reported purchasing similar items to their typical grocery shopping trip (M=5.92/7.00, SD=1.10); purchase typicality did not differ across conditions, t(392)=.35, p=.73.

Primary Results

We did not observe any significant differences in any of the measured health variables across conditions. Full results can be found in table 2.

Table 2. 

The Effect of Three Interventions on Cart Price, Number of Items in Cart, and Item Healthiness

OutcomeIllustration I: The effect of FOP labelsIllustration II: The effect of page sortingIllustration III: The effect of categorization
Test group
M (SD)
Control group
M (SD)
t(392)Test group
M (SD)
Control group
M (SD)
t(362)Test group
M (SD)
Control group
M (SD)
Total price (USD)96.05110.881.44109.2595.421.3994.26106.121.06
(89.42)(113.58) (103.85)(85.25) (78.41)(106.21) 
Number of items purchased24.7230.442.10*29.0226.081.1028.4833.221.31
(21.01)(31.94) (28.24)(22.81) (23.59)(36.19) 
Calories per serving123.55122.76.16115.36130.693.10**116.67120.35.78
(53.35)(44.01) (35.70)(55.95) (44.96)(34.14) 
Total fat (g) per serving3.
(2.09)(2.18) (1.78)(1.94) (2.01)(1.80) 
Saturated fat (g) per serving1.
(.92)(.89) (.89)(.85) (.85)(.89) 
Trans fat (g) per serving.*.01.022.24*
(.06)(.03) (.017)(.061) (.02)(.05) 
Cholesterol (mg) per serving12.7312.70.2512.1715.361.7915.6913.421.49
(11.83)(12.33) (11.33)(21.22) (14.74)(10.30) 
Sodium (mg) per serving133.34139.75.60124.50144.351.85144.15164.501.69
(104.90)(108.57) (93.15)(110.31) (94.10)(107.23) 
Carbohydrates (g) per serving20.4720.07.3318.6221.882.58*17.4518.05.54
(12.67)(11.12) (7.77)(15.08) (11.09)(6.48) 
Fiber (g) per serving2.
(1.29)(1.15) (1.04)(1.30) (1.15)(.99) 
Sugar (g) per serving10.7810.18.588.5811.432.81*7.757.59.19
(10.91)(9.65) (6.37)(12.04) (8.91)(4.78) 
Protein (g) per serving4.144.29.614.264.601.464.774.73.14
(2.37)(2.46) (2.22)(2.28) (2.33)(1.95) 
Star points per serving.56.77.411.97.712.51*−.48.902.42*
(5.19)(4.92) (4.53)(5.05) (4.97)(4.66) 

p < .10.

*p < .05.

**p < .01.

View Table Image


In this study, we demonstrated that there was no effect of a healthy stars FOP labeling system on the healthiness of purchased items in OSOG. This finding replicates Croker et al.’s (2020) study of star FOP labels in an online grocery setting and accords with a large body of literature finding small or null effects of FOP labels on food choice. In the next study, we sought to replicate another choice context intervention that is backed by a large body of literature: page sorting.

Illustration II: Page Sorting

In the OSOG, it is possible to change the default way foods are displayed on the page. We specifically studied whether bringing healthy foods to the top of the page would affect behavior. Conceptually, this intervention is similar to visibility enhancements (e.g., placing healthy objects at eye level, placing healthy entrees early in a menu) or positioning nudges (e.g., putting healthier items closer to the participant) in brick-and-mortar stores, both of which have consistent effects on purchasing (Bucher et al. 2016; Cadario and Chandon 2020). Studies of online shopping behavior also point to an effect of ordering. An observational study of online grocery shoppers suggests that consumers are likely to select items appearing closer to the top of each category (Anesbury et al. 2016). For nongrocery online retailers, there also seems to be a consistent effect of position such that items positioned closer to the top of a list are more likely to enter the consideration set (Cai and Xu 2008; Xu and Kim 2008), more likely to be clicked (Murphy, Hofacker, and Mizerski 2006), and ultimately more likely to be chosen (Tam and Ho 2005). In this study, we sought to replicate the effects of sorting using the OSOG. We expected to observe an effect of product sorting on the healthfulness of the items in participants’ carts at checkout.


We recruited 364 participants from Prolific (171 men, 186 women, 7 nonbinary; Mage=31.30, SD=10.62; 254 White, 22 Black, 39 Asian, 7 Southeast Asian, 17 Hispanic or Latinx, 1 Indigenous, 1 Pacific Islander, 20 bi- or multiracial, 3 other; MBMI=25.35, SD=6.79, excluding one participant who declined to answer).

Participants were asked to use the OSOG to shop for groceries for the upcoming week. Instructions were the same as in prior experiments, including two participants winning the contents of their cart.

Participants were randomly assigned to one of two versions of the OSOG. In the control condition, grocery items were organized randomly within each category. In the experimental condition, grocery items were organized by nutritional value with the most nutritious items being displayed at the top of each category. Specifically, items were ordered by the same star points algorithm used in study 1. Across both conditions, participants were not shown FOP labels, were able to use the built-in search function, and were prompted to spend $35 before being able to check out. A glitch in the study allowed 13 participants to check out without meeting this minimum threshold. After completing their hypothetical shopping, participants completed the same shopper profile, demographics, and debriefing as in study 1.


Shopper Profile

The majority of participants imagined shopping for themselves (30.8%) or themselves and another person (30.5%). On average, participants reported purchasing similar items to their typical grocery shopping trip (M=5.88/7.00, SD=1.13); purchase typicality did not differ across conditions, t(362)=.52, p=.60.

Primary Results

Participants in the experimental reordering condition purchased items with fewer calories, less trans fat, fewer carbohydrates, less sugar, and more star points than participants in the control condition, all p<.03. Several other micro- and macronutrients (saturated fat, cholesterol, sodium) trended in the expected direction, all p<.08. Full results can be found in table 2.


In this study, we find that changing the default sorting to present healthier items first results in overall healthier purchases (compared to a randomly ordered store). This effect is particularly notable given that OSOG has a search function. If participants knew exactly what they were looking for, they would have searched the items and the within-page sorting would not have affected their purchases. The observed effect of within-page sorting on purchase behavior replicates visibility findings from brick-and-mortar retailers (Bucher et al. 2016; Cadario and Chandon 2020) and replicates page sorting findings from nongrocery retailers (Tam and Ho 2005).

In our final illustration, we investigate whether creating curated product categories of healthy foods (similar to having a “healthy” section of a brick-and-mortar store) would affect consumer behavior.

Illustration III: Categorical Organization within the Store

In the OSOG, researchers can curate a custom category. In this experiment, we created a category that contained only the healthiest products in the store. By definition, curating a category allows consumers to opt-in to a limited consideration set. For example, rather than choosing among all the available soups, consumers are limiting themselves to only the healthiest soups.

We expect that a curated category would improve consumer choice for a variety of reasons. First, consumers experience less choice overload, and are more motivated to actually purchase from small versus large assortments (Iyengar and Lepper 2000; Boatwright and Nunes 2001). For example, a consumer who is curious about veggie burgers might have been overwhelmed by the large selection in the standard store. However, she may find it easier to make a purchase decision if choosing from a limited set of only the healthiest veggie burgers. Second, creating a curated category reduces consumers’ search costs. When searching among large assortments, consumers often eliminate options that do not meet minimum criteria (Bettman, Luce, and Payne 1998). By curating a category of healthy foods, we are enabling elimination. By allowing consumers to easily act on their desire to choose healthier products, we should facilitate healthier choices.

There is limited empirical data on category curation and choice. A study examining consumers’ response to restaurant menu format found that, when calorie information is posted, consumers avoid a menu section containing only low calorie items, leading them to choose an food item with more calories overall (Parker and Lehmann 2014). However, this finding is within a hedonic context where consumption is immediate; in the context consumers’ taste goals may be more salient than their health goals (VanEpps, Downs, and Loewenstein 2016). Furthermore, data from non-food-related consumption contexts suggests that products appearing in curated categories are consumed at much higher rates than products that do not appear in curated categories (Aguiar and Waldfogel 2021). Based on this research and the literature on consumer consideration sets, we suspect that consumers will purchase healthier foods when a “healthy section” is made available.


We recruited 283 participants from Prolific (123 men, 157 women, 3 nonbinary or prefer to self-describe; Mage=32.75, SD=11.80; 198 White, 22 Black, 29 Asian, 1 Southeast Asian, 17 Hispanic or Latinx, 1 Pacific Islander, 13 bi- or multiracial, 2 other or prefer not to answer; MBMI=25.68, SD=6.19).

Participants were asked to use the OSOG to shop for groceries for the upcoming week. Instructions were the same as in prior experiments, including two participants winning the contents of their cart.

Participants were randomly assigned to one of two versions of the OSOG. In the control condition, participants saw nine product categories (i.e., produce; meat, dairy and eggs; bakery, pasta and grains; dry goods, breakfast and spices; pantry; canned; snacks; beverages, and; frozen foods) as headings to help navigate the store. In the experimental condition, participants could opt in to seeing an additional heading “Healthy Section.” The healthy section included subcategories representing each of the other store categories (e.g., produce, dry goods). Items appearing in the Healthy Section were in the seventieth percentile or higher of health as rated by the star point algorithm outlined in study 1. In both conditions, participants were not shown FOP labels and were prompted to spend $35 before being able to check out. After completing their hypothetical shopping, participants completed the same shopper profile, demographics, and debriefing as in study 1.


Opting In

Recall that we allowed participants to opt in to seeing the healthy items category. We did this to reduce reactance responses to the feature. 15 participants chose not to see the Healthy Section. Statistics are presented below using an intent-to-treat analysis. However, results are similar if the participants who opted out are assigned to the control condition.

Shopper Profile

The majority of participants imagined shopping for themselves (31.8%) or themselves and another person (33.9%). On average, participants reported purchasing similar items to their typical grocery shopping trip (M=6.05/7.00, SD=1.10); purchase typicality did not differ across conditions, t(280)=.47, p=.64.

Primary Results

Participants in the experimental reordering condition purchased items with less trans-fat and more star points than participants in the control condition, all p<.03. Sodium trended in the expected direction, p=.09. Full results can be found in table 2.


Across three studies, we demonstrate an effect of sorting and categorization on consumer choice in the OSOG store. We fail to demonstrate an effect of labeling. The effects for ordering and labeling increase our confidence in the relative ecological validity of the store because they mirror the effects generally found in large scale field studies.

The effects of curated categories is also important in that it implies areas where future research is needed. The observed effect of curated categories is encouraging, suggesting that categorical organization may be helpful in a grocery setting, when consumers may be more actively pursuing a health goal. However, many questions remain. For example, what is the optimal assortment size within these categories? Do curated categories outperform filters that may perform a similar exclusionary function?

Our goal in running these studies was to introduce researchers to a new tool to examine the effects of choice context on consumer food choice. In this article, we present three illustrations of the store’s capabilities. However, we do not see these illustrations as the ultimate empirical demonstration of the store’s features. For example, although we found a null effect of star labels, we encourage researchers to make labeling changes within the store and observe how they affect purchase behavior. As we did not test the effect of traffic light labels, organic labels, or any other of the myriad of FOP labels currently being used in the market, more research could contribute valuable knowledge to understanding the effects of labeling.

It is important to note that we do not envision the store as a substitute for field studies. Although the store can be made incentive compatible by providing bonuses, it is not a substitute for an experimental or quasi-experimental field study. Rather, we envision the store as a first step in a research project: a way to determine if there is any support for a hypothesis before beginning a cost- and labor-intensive field study. Furthermore, researchers can use results from the store to address any concerns retailers may have about profits or consumer satisfaction, perhaps making retailer collaboration more likely.

Beyond Food Choice: Using OSOG to Study Consumer Behavior More Broadly

The studies included in this package do not represent the full capabilities of the store. There are many store features that we did not test as part of this package (see fig. 2). Furthermore, there are many other ways of using the OSOG that do not necessarily rely on changes to the choice context. For example, researchers have used the store as a pre- post-intervention measure of behavioral flexibility in food choice in adolescents with anorexia nervosa and their parents (Timko et al. 2021).

Furthermore, consumer behavior researchers can use the store to study research questions unrelated to health behavior. For example, one group of researchers are using the store to mimic a consumer-brand interaction and assess downstream consequences for brand connection and brand community (Rifkin, Valsesia, and Cutright, personal communication). A second group of researchers is using the store to study a variety of recommendation algorithms (Xu, Deng, and Mela, personal communication). Because researchers can be certain where items appeared on the screen, the store also lends itself to eye-tracking studies. When combined with an eye-tracker, OSOG may provide valuable insights into how consumers seek and engage with product information.

As the capabilities of the store improve over time, it may become possible to use the tool to study consumer reactions to price changes or promotions, packaging changes, or stock-outs. We encourage readers to be creative with the store and to contact the first author if they have questions about store capabilities.



Holly S. Howe (corresponding author: ) is an assistant professor of marketing at HEC Montréal, Montréal QC, Canada. Peter A. Ubel () is the Madge and Dennis T. McLawhorn University Professor, Marketing Department, Fuqua School of Business, Duke University, Durham, NC, USA. Gavan J Fitzsimons () is the Edward S. and Rose K. Donnell Professor of Marketing and Psychology, Marketing Department, Fuqua School of Business, Duke University, Durham, NC, USA. The authors would like to thank Melanie Gotz and Sydney Palumbo for their contribution to the programming of the store. They would also like to thank Abby Huang, Carmen Leslie, and Sarah Thomas for their research assistance throughout.