Every product list starts in an order. Some shoppers change it; many begin with what the store puts in front of them. Offering a sort menu does not remove the decision about that first view.
The practical question is which objective the default should serve, and how to tell whether it succeeds. Grocery experiments show why the outcome matters, why other page information must be considered and why a new order can leave choices unchanged.
🔎 What research says
Test the default directly and measure the quality of the chosen products. Separate sorting from other page changes, and retain a comparison that can reveal no improvement. A health-based result in one grocery task does not establish a universal ranking rule.
🗺️ In this guide
1. Test the default, not just the sort menu
📈 Recommendation
Choose an initial order with a clear purpose, then test it. Keep alternative sorts available, but do not assume shoppers will use them.
Compare the proposed default with the current one while keeping products, prices and availability constant.
Track how often shoppers change the order and whether the default improves the outcome you intended.
🎓 Findings
Godden and colleagues (2025) randomly assigned 1,151 Belgian participants to seven virtual-supermarket conditions. A health-first default improved the mean diet score of their baskets compared with alphabetical order.
Optional health-based sorting and filtering were used by fewer than one in ten participants and did not change basket outcomes in that experiment.
🧠 Why it works
The first order affects what a shopper meets without taking another action. An optional control cannot carry the whole decision if few people use it. This is a reason to measure the default directly.
✋ Limitations
The task used a prescribed shopping list in a virtual grocery store. It does not identify the best default for every category or forecast completed paid orders.
2. Measure the quality of the chosen basket
📈 Recommendation
Define the outcome the new order should improve. If the goal is better-suited products, judge the selected basket against that goal as well as clicks.
Specify a customer-relevant measure before testing: compatibility, nutritional composition or another verifiable attribute.
Track completed purchases, basket value and returns alongside the target measure.
🎓 Findings
Valenčič and colleagues (2024; online 2023) analyzed 175 Australian adults in a simulated grocery task. Both higher-fibre categories and higher-fibre products were moved up rather than down.
Selected baskets contained a median 1.62 g of fibre per 100 kcal, versus 1.34 g in the comparison group. This describes basket composition. Participants did not pay for or receive the groceries.
🧠 Why it works
An order can change what people choose without increasing the number of clicks or items. A measure tied to the shopper’s objective is more informative than engagement alone. Extending this to compatibility or returns is our proposed application.
✋ Limitations
The study changed category and product order together, so it cannot isolate one of them. Of 303 completed responses, 128 were excluded. Differences in basket price and energy were not statistically clear; that does not establish equivalence.
3. Test sorting alongside the information shoppers see
📈 Recommendation
Evaluate the default together with the labels and basket feedback already on the page. Change one factor at a time, or use a test that separates their effects.
Compare sorting alone, existing information alone and their combination when traffic allows a meaningful comparison.
Check whether each change helps, and whether the combined result adds anything beyond the separate changes.
🎓 Findings
Shin and colleagues (2022) assigned 756 people to 16 conditions in a hypothetical grocery store. The four factors were labels with basket feedback, tax, nutritional ordering and healthier substitute offers.
Ordering and the combined labels/feedback factor improved basket diet quality. The study found no statistically clear two-way interactions. It does not demonstrate a special boost from combining the interventions.
🧠 Why it works
Sorting and information act on the same choices. Testing them separately helps distinguish a useful default from another change that happened at the same time. A combined improvement need not imply that the two reinforce each other.
✋ Limitations
Labels and feedback were one factor, so their individual effects cannot be separated here. The baskets were hypothetical. The result does not establish an interaction for your store.
4. Keep a no-improvement result possible
📈 Recommendation
Treat a mission-based sort as a testable proposal. Keep a comparison group and decide what evidence would justify retaining the change.
Measure the intended choices directly, and check whether shoppers notice any explanation of the order.
Keep customer control and access to familiar products; revise the approach if the desired shift does not appear.
🎓 Findings
Shi and colleagues (2023; online 2022) tested carbon-footprint ordering with 1,842 online-panel participants choosing a meal for two in a simulated supermarket. Products appeared in random order or in lower-carbon-first order, with or without an explanation.
Covert ordering produced no statistically clear improvement in lower-carbon choices. The difference between overt and covert ordering was also unclear. The primary funder summary suggests that prior preferences may have outweighed position.
🧠 Why it works
Appearing first is an opportunity to be inspected, not a guarantee of being chosen. Familiar preferences can still dominate. An unchanged outcome should inform the decision rather than disappear from the report.
✋ Limitations
This was a six-category simulated task, not a live sales trial. The null result does not prove that carbon-based sorting never works, and it does not negate the health-ordering findings in different settings.
💡 The takeaway
Choose a clear objective for the first order, then test the actual choices.
Compare your current default with one plausible alternative. Keep a visible way to change the order, measure the intended customer outcome alongside completed purchases, and check the labels and feedback that accompany the list.
📚 Sources and study types
E. Godden, N. Dens, B. Coppens, L. Thornton (2025). Can we improve the healthiness of online food purchases through the Nutri-Score and site design?. Food Policy, 134, 102899. Randomized seven-condition virtual-supermarket experiment; Prescribed shopping list in a virtual store; no assumed real paid-order effect.
Eva Valenčič, Emma Beckett, Clare E. Collins, Barbara Koroušić Seljak, Tamara Bucher (2024). Changing the default order of food items in an online grocery store may nudge healthier food choices. Appetite, 192, 107072. Randomized simulated online-grocery task; No payment/delivery; category and product order changed together; substantial exclusions.
Soye Shin, Bibhas Chakraborty, Xiaoxi Yan, Rob M. van Dam, Eric A. Finkelstein (2022). Evaluation of Combinations of Nudging, Pricing, and Labeling Strategies to Improve Diet Quality: A Virtual Grocery Store Experiment Employing a Multiphase Optimization Strategy. Annals of Behavioral Medicine, 56(9), 933–945. Randomized full-factorial hypothetical grocery experiment; Hypothetical baskets; labels/feedback one combined factor; no evidence of universal synergy.
Zhuo Shi, Michael Ratajczak, Katie Thornton, Phil Jones, Ayla Ibrahimi Jarchlo, Natalie Gold (2023). Testing the impact of overt and covert ordering interventions on sustainable consumption choices: A randomised controlled trial. Appetite, 181, 106368. Preregistered randomized three-arm simulated-supermarket trial; Simulated meal-for-two task; null result does not prove no effect in other stores.
