Shopper research: choose the shelf or range change worth testing
Follow a worked shopper research example from competing packs and shelf space to substitution, a range decision and the retail evidence needed next.
Shopper research investigates how people find, compare and choose products in a defined shopping situation. For a roasted-chickpea brand, it can help decide whether clearer shelf information or a smaller bag deserves a retail test. The useful output connects the proposed offer, range or shelf change to the buying task it should improve.
A snack that looks promising on its own must compete with a familiar bag, a different pack size and the option to buy nothing. A shopper can like the idea and still leave it behind. Before changing the range, establish which part of that choice the proposal could address.
The proposal needs an operating consequence
Imagine ReadyBite, a fictional grocery brand selling roasted chickpeas. Its category team has two proposals: explain the existing bag more clearly at the shelf, or add a smaller bag for a first purchase. The retailer will consider a limited test, but the brand has no additional shelf width. Every new facing must take space from an existing one. A facing is one product front visible across the shelf. The offers and possible buyer pathways below are part of this worked example.
The team needs to decide which proposal deserves that retail test. The brand wants new buyers; the retailer wants a useful category offer. A smaller bag that mainly replaces purchases of the brand's larger bag may serve neither objective well enough. Clearer communication could solve a misunderstanding while leaving the main reason for rejection untouched.
Write the choice in operational terms: keep the current range, test a shelf explanation, or test a small pack within the same fixture. That gives the research a practical consequence. Our CPG market research guide covers the broader launch proposition; here, the question is what changes when the product competes for a place in a real shopping basket.
If you first need to decide which products to stock, our TURF analysis guide shows how to compare unique consideration coverage under launch constraints. Use the shopping study to investigate what the selected range would replace in the basket.
Build a complete, feasible choice
ReadyBite's test concerns a small ambient snack fixture in one grocery format. The shopping occasion is buying a packaged snack for today or the next few days. The choice includes these offers, with ordinary shelf labels and no temporary promotion:
Offer | Pack and shelf price |
|---|---|
ReadyBite roasted chickpeas | 120 g for $3.00; $2.50 per 100 g |
Store-brand salted crisps | 120 g for $2.40; $2.00 per 100 g |
Familiar-brand trail mix | 150 g for $4.50; $3.00 per 100 g |
ReadyBite small bag, candidate only | 40 g for $1.20; $3.00 per 100 g |
The three current products each occupy two 10 cm facings on a 60 cm fixture. Both ReadyBite bag sizes fit a 10 cm facing. The small-pack candidate replaces one large-bag facing, leaving one facing for each ReadyBite size; competitor space stays fixed. The retailer can stock and replenish either arrangement. These constraints are part of the example, including the space cost of adding the new bag.
The research must also allow another product elsewhere in the shop, postponement and no snack purchase. If every response has to choose one of the displayed products, the exercise cannot reveal rejection of the fixture as a whole. Ask for the person's usual snack and purchase occasion before showing the candidates, so you can distinguish a plausible switch from a response to an unfamiliar research display.
Keep the category boundary explicit. This fixture does not represent every snack the store sells. A shopper whose normal choice is fruit or a chilled drink can take that route; the study should record it rather than force them into the packaged-snack set.
Communication and assortment are different changes
Compare three shelf conditions:
Condition | What changes from the current fixture |
|---|---|
Current range | Two ReadyBite 120 g facings, alongside the crisps and trail mix |
Clearer shelf explanation | Same products, prices and facings; a strip below ReadyBite reads “Roasted chickpeas · four 30 g servings” |
Small-pack range | One ReadyBite 120 g facing becomes one 40 g facing; no new explanation strip |
The serving statement is assumed accurate for the fictional product. In an actual study, the wording must match the product evidence and intended pack. Our claims testing guide explains how to investigate the promise that wording creates.
Randomly assign each eligible shopper to one condition for their first encounter. Keep the image quality, fixture size, competing offers, availability and instructions matched. This gives each arrangement an unaided first encounter. A later direct comparison can explain trade-offs, but it answers a different question: people now know alternatives that an ordinary shopper would not have seen together.
Ask what they would choose and why before drawing attention to the smaller bag or the shelf strip. Then investigate what they thought the product was, how much they would get, what they expected to do with it and which alternative they would otherwise buy. Save the exact display version with the answer. “The small pack won” is unhelpful if the stimulus also moved the brand to eye level or added a promotion.
For a narrower decision about pack artwork and interpretation, use our packaging design testing guide. Match the shopping encounter and competing offers to the design question you need to resolve.
This attention to competing offers is consistent with Ipsos' description of virtual retail testing, which distinguishes shelf and communication scenarios from assortment changes and substitution. A competitive context makes the task more relevant; it does not make a stated or modeled choice an observed sale.
A new selection has an origin
Consider four possible pathways the study should be able to distinguish. They are alternative mechanisms to investigate, with different consequences for the same proposed small bag.
A shopper new to ReadyBite changes from the crisps to the small bag. They understand the large bag but do not want that much unfamiliar food. The smaller purchase fits the immediate occasion. This would support investigating a trial-size offer. It would not, by itself, tell you whether pack size, the lower total outlay or some combination caused the switch: the proposed product changes both size and total price.
A regular ReadyBite buyer moves from 120 g to 40 g. The new bag may be convenient, but the brand has exchanged an existing purchase for a smaller one. That is cannibalization: an offer takes demand from another product in the same range. It belongs in the decision alongside recruitment from competitors. A count of new small-pack selections would hide the distinction.
A shopper understands the shelf explanation and still chooses trail mix. They want something to share with another person and prefer the familiar mix. Improved comprehension has done its job without solving this purchase occasion. If the team reports only understanding, it could mistake clearer communication for a stronger offer.
A shopper chooses nothing from the fixture. They wanted a fresh snack and were never considering these products. A different pack might have no bearing on that decision. Keep this account distinct from someone who wanted roasted chickpeas but could not find them.
These pathways do not establish how common any response is. Their purpose is to make the eventual result interpretable. In a study, report the source and base for each group, preserve the outside choices, and identify where a new selection came from. If the data cannot connect a response to its prior or alternative choice, call the source of switching unknown.
A consumer decision tree makes the substitute visible
The four pathways become more useful when the team can see which condition a shopper protects and which one they will trade away. A consumer decision tree is a proposed hierarchy of those choices within a particular category and occasion. For ReadyBite, the useful question is whether the shopper protects the product, the pack size, the cash outlay or something outside this fixture when the first choice isn't available.
Don't infer the hierarchy by sorting a few explanations into a diagram. “I wanted a small bag” could describe a hard limit, a mild preference or a reason supplied after the choice. Ask about the shopping task before exposure, record the actual or declared choice, then use a feasible unavailability task to investigate what changes. Keep the initial choice and later substitution encounter distinct.
For the snack fixture, the following is a candidate structure to test, with counts deliberately unfilled. It organizes the existing example rather than supplying a measured hierarchy.
Eligible snack-shopping tasks [base N0]
Packaged snacks in this fixture are acceptable [N1 / N0]
ReadyBite is in the acceptable set [N2 / N1]
Preferred ReadyBite size is available [N3 / N2]
Chosen item and quantity at displayed price [N4 / N3]
Preferred ReadyBite size is unavailable [N5 / N2]
Other ReadyBite size / competitor / elsewhere / postpone / none
ReadyBite is outside the acceptable set [N6 / N1]
Crisps / trail mix / another product elsewhere / postpone / none
Nothing in this fixture is acceptable [N7 / N0]
Product elsewhere / postpone / no snack purchase
Every split needs mutually exclusive coded outcomes and a parent base. N1 plus N7 should recover N0; N2 plus N6 should recover N1. Keep unknown or unanswerable paths as separate branches when they occur. A person who considers several products contributes to one final choice path, while their consideration set remains a separate record. Otherwise the same shopper can accidentally swell several branch counts.
The “preferred size unavailable” branch is where substitution becomes inspectable. Record the preferred item, the attribute the shopper says they need, the removed item, remaining prices and availability, chosen replacement and quantity, outside choice and reason. For example, a ReadyBite buyer might accept another ReadyBite size, choose the $2.40 crisps rather than the $3.00 large bag, or buy something elsewhere. A strict $1.20 ceiling leaves no other displayed product within budget when the small bag is absent. Those are competing possibilities to investigate, not findings about these buyers.
Testable hierarchy | Evidence that would support or challenge it |
|---|---|
Brand or product first, size second | Buyers continue choosing ReadyBite when their preferred size is absent, rather than protecting size through a competitor. |
Pack size or immediate outlay first | Buyers switch brand to preserve the needed quantity or spending limit. Size and total price still need separating if both change. |
Occasion first, fixture second | Buyers leave the set when none fits the task, even if they like a displayed brand. |
Compare these candidate explanations with transaction patterns where the records, product availability and permissions allow it. NIQ's assortment approach explicitly considers product substitutability and demand transfer. That supports asking where a lost selection goes; it doesn't turn the small proposed tree above into observed demand.
For a SKU-removal decision, use the current preferred-SKU buyer base and report the destination of every affected choice, including unknown and outside options. The original item stays visible in the baseline, then is absent in the removal condition while the rest of the feasible offer is held clear. Keep repeated tasks identifiable and address order or learning effects in the design. A stated switch, a modeled switch and a sale during an actual stockout are different evidence.
The tree should change the retail test. A product-first pattern makes another size a plausible substitute. A quantity-first pattern makes that less safe. A large outside branch warns the team that removing a modest-selling item could lose a distinct shopping occasion. Check those hypotheses against actual baskets, availability and contribution before treating the tree as permission to remove a product.
The mechanism determines the next action
For ReadyBite, take the small bag forward only if the evidence supports a useful buying occasion that the current range does not serve. Separate potential new-to-brand trial from moves out of the large bag, and inspect the reasons behind both. Investigate the clearer strip when misunderstanding the product or quantity is the obstacle. If the main rejection concerns taste or texture, a shelf explanation cannot establish that the product experience will satisfy it.
The two proposals are commercially realistic packages of changes. A small bag has a different size, total price and unit price; it also reduces large-bag facings. If you need to isolate one of those factors, design an additional feasible comparison. Our pricing research guide develops the choice of method when the unresolved question concerns the product's price. Do not attribute an entire range result to one factor merely because it is the team's preferred explanation.
The handoff should make responsibilities clear. The brand's insights team owns the explanation of the choices and their evidence. Brand and product teams own the proposed pack or communication revision. The retailer's category and operations teams check space, stock depth, availability and the ability to replenish the tested arrangement. Finance checks the commercial consequence using actual pack economics. No single research preference score supplies all four answers.
Retail purchase records can then answer a different part of the problem. dunnhumby's assortment approach uses loyalty and point-of-sale data at store level alongside shelf constraints. Where those records and permissions are available, they can help investigate what people actually bought and what changed around the range intervention.
Retail evidence answers the remaining range question
Before putting the candidate in stores, agree what would make the change useful. For the small bag, inspect new-to-brand purchasing where it can be measured, movement from the existing large bag, total brand and category units and value, and contribution using the actual economics. Where repeat purchase matters, allow enough follow-up to observe it. Keep units and value separate: one larger pack and one smaller pack are not equivalent volumes of food.
Check whether the intended fixture was present. A missing large bag could create small-pack purchases that look like preference. Record stockouts, compliance with the shelf arrangement, promotions and material competitor changes. With appropriate comparable stores or a randomized store test, assess the candidate against the current arrangement over comparable periods. Differences in store traffic or availability should not silently become credit for the new range.
A rendered shelf image also cannot establish how easy the physical bag is to open, whether the food tastes good, or whether the product stays satisfactory after opening. Use actual product experience for those questions. Likewise, if the decision depends on whether people notice an item while moving through a busy aisle, obtain evidence from an appropriate human shopping task or store observation.
Related guides: retail market research.
From ReadyBite’s fixture to an inspectable study
With Vensia, you can compare how data-grounded synthetic audiences respond to defined offers and shopping scenarios, then investigate the reasons and audience differences behind the modeled choices. For a question like ReadyBite's, that can help distinguish a communication problem, a plausible trial occasion and a proposed change that mainly shifts choice within the range.
ReadyBite’s three feasible shelf conditions and usual snack occasion give our team a defined comparison to shape and interpret. Modeled responses remain distinct from observed transactions, physical shopping and product use; our guide to synthetic and human respondents explains how to choose the evidence for each question.
The retailer has offered one limited test, so the study needs to explain why a candidate should take that slot. In a Vensia demo, we can specify the three matched layouts and outside choices, then agree how to return unaided understanding, modeled choice, reasons and the alternative each selection would replace. That makes a new-to-brand occasion distinguishable from a move out of ReadyBite’s large bag, with unknown origins kept visible.
Related guides: clothing brand research.
