DemoAn example study: real packs, invented results.
Chapter 5 of 5
Read the report
The report says what shoppers did, with a margin on every figure, and whether the question is settled. The same report can be shared by link or as a PDF.
Example data from a demonstration study. The packs are real products; the study, the shoppers and every number below are invented and say nothing about these brands.
New pack on the snack shelf
Packaging Standout · 40 visits analysed (0 excluded)
The answer
Research question: Do more shoppers notice the new pack first than the current pack on the same shelf?
Yes. With the new pack, 67% of the eye-tracked shoppers looked at Chex Mix Traditional Snack Mix first (12 of 18), against 6% with the current pack (1 of 18): a gap of 61 points. The gap holds after correcting for the other comparisons in the study. The new pack wins the first look on this shelf. Each figure carries a margin of up to ±20 points at this size, so read the size of the gap as approximate; its direction is not in doubt.
67%
looked at Chex Mix first with the new pack
±20 points, 18 eye-tracked shoppers
6%
with the current pack
±12 points, 18 eye-tracked shoppers
Settled
the gap holds after the correction
40 visits analysed
First look is measured by webcam eye tracking, on the shoppers who switched it on. The evidence is in the sections below: open the one you want.
Version A against version BEvery measure side by side for the current pack and the new pack, with the margin on each.ShowHide
Real shoppers · 2 versions
Version B: new pack is ahead of Version A: current pack on "first look on the target" (67% ±20 against 6% ±12), and the difference holds after correcting for the other comparisons.
Same conclusion with excluded visits included: no visit was excluded.
| Metric | Version A: current packThe shelf as it is today. | Version B: new packThe same shelf with the new pack. | Difference | Test | Verdict |
|---|---|---|---|---|---|
| What did shoppers do when free to shop?: bought Chex Mix Traditional Snack Mixhigher is better | 5% ±111 of 20 shoppers | 20% ±174 of 20 shoppers | B: new pack less A: current pack: 15 points (95% interval -7 to 37) | Fisher's exact testp = 0.342, corrected 1.000 | No difference detected |
| What did shoppers do when free to shop?: picked up Chex Mix Traditional Snack Mixhigher is better | 5% ±111 of 20 shoppers | 30% ±196 of 20 shoppers | B: new pack less A: current pack: 25 points (95% interval 1 to 47) | Fisher's exact testp = 0.091, corrected 0.915 | No difference detected |
| Looked at Chex Mix Traditional Snack Mix (target) in free shoppinghigher is betterConfidence grade: Fair. Gaze from a webcam; validated tracked shoppers only. Estimated shares that allow for the measured eye-tracker error. | 26% ±195 of 18 validated tracked shoppers | 71% ±1913 of 18 validated tracked shoppers | B: new pack less A: current pack: 44 points (95% interval 12 to 66) | Two-proportion z-testp = 0.008, corrected 0.123 | No difference detected |
| First look on Chex Mix Traditional Snack Mix (target) in free shoppinghigher is betterConfidence grade: Fair. Gaze from a webcam; validated tracked shoppers only. Estimated shares that allow for the measured eye-tracker error. | 6% ±121 of 18 validated tracked shoppers estimated to look at a product | 17% ±173 of 18 validated tracked shoppers estimated to look at a product | B: new pack less A: current pack: 11 points (95% interval -12 to 34) | Fisher's exact testp = 0.603, corrected 1.000 | No difference detected |
| First shelf approached in free shopping: Salty snacks, bay 4no better or worseThe shelf approached first by most shoppers overall. A shelf counts as approached when the shopper stayed in the zone in front of it for at least 1 s, slowed to under 0.5 m/s there, or picked up a product from it; walking past does not count. | 75% ±1815 of 20 shoppers with store-flow records | 100% ±820 of 20 shoppers with store-flow records | B: new pack less A: current pack: 25 points (95% interval 4 to 47) | Fisher's exact testp = 0.047, corrected 0.613 | No difference detected |
| First side of the store in free shopping: leftno better or worseThe side of the first shelf the shopper approached. The other shoppers went right, straight ahead or to no shelf. A shelf counts as approached when the shopper stayed in the zone in front of it for at least 1 s, slowed to under 0.5 m/s there, or picked up a product from it; walking past does not count. | 15% ±153 of 20 shoppers with store-flow records | 0% ±80 of 20 shoppers with store-flow records | A: current pack less B: new pack: 15 points (95% interval -4 to 36) | Fisher's exact testp = 0.231, corrected 1.000 | No difference detected |
| Visited Salty snacks, bay 4 (the shelf with the target)higher is betterA visit is a stay of 2 seconds or more in the zone in front of the shelf. | 90% ±1418 of 20 shoppers whose store has the shelf | 100% ±820 of 20 shoppers whose store has the shelf | B: new pack less A: current pack: 10 points (95% interval -8 to 30) | Fisher's exact testp = 0.487, corrected 1.000 | No difference detected |
| Looked at Chex Mix Traditional Snack Mix (target)higher is betterConfidence grade: Fair. Gaze from a webcam; validated tracked shoppers only. Estimated shares that allow for the measured eye-tracker error.More shoppers looked at Chex Mix Traditional Snack Mix in Version B: new pack than in Version A: current pack: 84% versus 27%, in this study. | 27% ±195 of 18 validated tracked shoppers | 84% ±1615 of 18 validated tracked shoppers | B: new pack less A: current pack: 56 points (95% interval 23 to 74) | Two-proportion z-testp = under 0.001, corrected 0.015 | Difference detected |
| First look on Chex Mix Traditional Snack Mix (target)higher is betterConfidence grade: Fair. Gaze from a webcam; validated tracked shoppers only. Estimated shares that allow for the measured eye-tracker error.More first looks went to Chex Mix Traditional Snack Mix in Version B: new pack than in Version A: current pack: 67% versus 6%, in this study. | 6% ±121 of 18 validated tracked shoppers estimated to look at a product | 67% ±2012 of 18 validated tracked shoppers estimated to look at a product | B: new pack less A: current pack: 61 points (95% interval 31 to 79) | Two-proportion z-testp = under 0.001, corrected 0.003 | Difference detected |
| Found ithigher is better | 80% ±1716 of 20 shoppers who tried | 95% ±1119 of 20 shoppers who tried | B: new pack less A: current pack: 15 points (95% interval -7 to 37) | Fisher's exact testp = 0.342, corrected 1.000 | No difference detected |
| Time to find, median among finderslower is betterIt was found faster in Version B: new pack than in Version A: current pack: median 11.4 s versus 21.2 s, in this study. | 21.2 sn = 16 shoppers who found it | 11.4 sn = 19 shoppers who found it | B: new pack less A: current pack: -8.88 s (95% interval -14.44 to -3.98)typical difference (median of all pairwise differences) | Mann-Whitney U testp = under 0.001, corrected 0.016 | Difference detected |
| Found it, of all shoppers askedhigher is betterShown with the median time to find over all shoppers asked. A shopper who gave up or reached the time cap counts as not yet found up to the moment they stopped; "not reached" means fewer than half found it. | 80% ±17; median time to find 23.2 s16 of 20 shoppers asked | 95% ±11; median time to find 11.4 s19 of 20 shoppers asked | B: new pack less A: current pack: 15 points (95% interval -7 to 37) | Fisher's exact testp = 0.342, corrected 1.000 | No difference detected |
| Time to find relative to own pace, medianlower is betterEach time set against that shopper's own usual pace (task medians pooled over the versions). 1.0x is the usual pace; below is faster.Set against each shopper's own pace, it was found faster in Version B: new pack than in Version A: current pack: median 0.7x versus 1.4x, in this study. | 1.4xn = 16 shoppers who found it | 0.7xn = 19 shoppers who found it | B: new pack less A: current pack: -0.59 x (95% interval -1.00 to -0.26)typical difference (median of all pairwise differences) | Mann-Whitney U testp = under 0.001, corrected 0.017 | Difference detected |
| Self-rated ease of finding, mean (1 to 7)higher is betterA stated answer, asked right after the find task ended, found or not. 1 = Very difficult, 7 = Very easy."Self-rated ease of finding, mean (1 to 7)" was higher in Version B: new pack than in Version A: current pack: mean 5.15 points versus 3.00 points, in this study. | 3.00 pointsn = 20 shoppers who answered | 5.15 pointsn = 20 shoppers who answered | B: new pack less A: current pack: 2.15 points (95% interval 1.49 to 2.81)difference of meanstypical shift (median of all pairwise differences) 2.00 points (95% interval 2.00 to 3.00) | Mann-Whitney U testp = under 0.001, corrected under 0.001 | Difference detected |
| Made at least one wrong picklower is better | 10% ±142 of 20 shoppers with a result | 0% ±80 of 20 shoppers with a result | B: new pack less A: current pack: -10 points (95% interval -30 to 8) | Fisher's exact testp = 0.487, corrected 1.000 | No difference detected |
| Which version did they choose from?: chose Chex Mix Traditional Snack Mixno better or worseNo target is bound to this task: this is the product chosen most overall. | 15% ±153 of 20 visits in the version | 40% ±208 of 20 visits in the version | B: new pack less A: current pack: 25 points (95% interval -3 to 48) | Two-proportion z-testp = 0.077, corrected 0.843 | No difference detected |
| Decision time, medianno better or worse | 10.9 sn = 20 shoppers who chose, with a usable time | 11.0 sn = 20 shoppers who chose, with a usable time | B: new pack less A: current pack: 0.20 s (95% interval -1.52 to 1.93)typical difference (median of all pairwise differences) | Mann-Whitney U testp = 0.776, corrected 1.000 | No difference detected |
| Decision time relative to own pace, medianno better or worseEach time set against that shopper's own usual pace (task medians pooled over the versions). 1.0x is the usual pace; below is faster. | 0.9xn = 20 shoppers who chose, with a usable time | 1.0xn = 20 shoppers who chose, with a usable time | B: new pack less A: current pack: 0.24 x (95% interval 0.06 to 0.41)typical difference (median of all pairwise differences) | Mann-Whitney U testp = 0.010, corrected 0.147 | No difference detected |
| How likely are you to buy Chex Mix Traditional Snack Mix?: rating, mean (1 to 5)higher is better | 2.85 pointsn = 20 shoppers who answered | 3.40 pointsn = 20 shoppers who answered | B: new pack less A: current pack: 0.55 points (95% interval 0.02 to 1.08)difference of meanstypical shift (median of all pairwise differences) 1.00 points (95% interval 0.00 to 1.00) | Mann-Whitney U testp = 0.060, corrected 0.718 | No difference detected |
| After the visit: ease of finding products, mean (1 to 7)higher is betterA stated answer. 1 = Very difficult, 7 = Very easy. | 4.15 pointsn = 20 shoppers who answered | 3.90 pointsn = 20 shoppers who answered | A: current pack less B: new pack: 0.25 points (95% interval -0.29 to 0.79)difference of meanstypical shift (median of all pairwise differences) 0.00 points (95% interval 0.00 to 1.00) | Mann-Whitney U testp = 0.357, corrected 1.000 | No difference detected |
| After the visit: interest held by the visit, mean (1 to 7)higher is betterA stated answer. 1 = Not at all, 7 = Completely. | 4.45 pointsn = 20 shoppers who answered | 3.90 pointsn = 20 shoppers who answered | A: current pack less B: new pack: 0.55 points (95% interval 0.11 to 0.99)difference of meanstypical shift (median of all pairwise differences) 1.00 points (95% interval 0.00 to 1.00) | Mann-Whitney U testp = 0.019, corrected 0.267 | No difference detected |
Each value shows its unit and the number it rests on; a share also shows its 95% margin in points. Shares are tested with Fisher's exact test when an expected count is under 5 and with the two-proportion z-test otherwise; times and ratings with the Mann-Whitney U test. The difference column gives the effect size and its 95% interval: for a share the difference in points between the two versions, for a time or a rating the typical shift (the median of all pairwise differences, Hodges-Lehmann), beside the difference of means where the row shows a mean.
21 rows were tested. The p-values are corrected together with the Holm method, and a row counts as significant only when its corrected p is under 0.05. A version is marked as better only on rows where higher or lower is better, and only once the study's key comparison is settled. A sentence under a row is written only when its difference holds after the correction; it describes what was observed between the versions shoppers were assigned to in this study, not why. Results describe behaviour in this virtual store: use them to compare the versions, not to forecast in-store levels.
Visits per version: A: current pack 20, B: new pack 20. The split is within what an equal split would give by chance (chi-square p = 1.000).
Why shoppers did it3-minute interviews after the walk, grouped into themes, each answer checked against what the store recorded.ShowHide
After the walk, each shopper answers a short interview about moments of their own visit: the product they picked straight away, the one they kept looking at, the one they came back to. The answers are grouped into themes, and each theme is a count of shoppers.
Why shoppers did it
From 22 completed interviews with the shoppers in this report. Each answer is labelled, then the labels are grouped into themes, so a theme's size is a count of shoppers. Quotes are verbatim.
Wanted a snack mix
6 of 22 shoppersShoppers wanted a mix of pieces and chose the bag that showed it clearly.
“I wanted a bit of everything, pretzels and the rye chips, and the bag shows all of it”
P-102Final choicepicked Chex Mix Traditional Snack Mix during the task “I was not sure it was the traditional one, the name is written sideways”
P-206Hesitatedlooked at Chex Mix Traditional Snack Mix for a while and did not take ittyped “I went back to it because I wanted a mix rather than one kind of chip”
P-105Came back to itcame back to Chex Mix Traditional Snack Mix after looking at other products and took it New pack easy to spot
5 of 22 shoppersThe new bag stood out next to the other blue bags, even low on the shelf.
“the new bag has the pieces right on the front, it jumped out so I just took it”
P-101Quick pickput Chex Mix Traditional Snack Mix in the basket almost straight away “easy to find this time, the new pack is brighter than the Doritos next to it”
P-104Quick pickput Chex Mix Traditional Snack Mix in the basket almost straight away “the red badge on the new bag is easy to spot even from a distance”
P-109Final choicepicked Chex Mix Traditional Snack Mix during the task Price felt high for the size
5 of 22 shoppersSome products were left on the shelf because the price felt high for the pack.
“I compared the price per box and the crackers were not worth it for me”
P-107Hesitatedlooked at Cheez-It Reduced Fat Crackers for a while and did not take it “the avocado oil chips looked good but 4.49 felt a lot for one bag”
P-103Looked, not takenkept looking at Boulder Canyon Sea Salt Kettle Chips but did not pick it uptyped “I checked a few prices and went for the one with the fairer price”
P-209Compared pricescompared several products and their prices on the shelf Current bag blends in
5 of 22 shoppersThe current bag blended in with the blue bags around it on the lower shelf.
“I was not sure it was the traditional one, the name is written sideways”
P-206Hesitatedlooked at Chex Mix Traditional Snack Mix for a while and did not take ittyped “I could not find it at first, it blends in with the blue Doritos bags”
P-201Slow to findtook a while to find Chex Mix Traditional Snack Mix “the old bag is all blue stripes, my eyes went past it twice”
P-203Slow to findtook a while to find Chex Mix Traditional Snack Mix Bought the usual brand
4 of 22 shoppersSome shoppers took the chips they always buy without looking further.
“I always buy the blue Doritos at home, I did not really look at others”
P-108Quick pickput Doritos Cool Ranch Tortilla Chips in the basket almost straight away “I usually buy Doritos so I grabbed the blue bag without thinking”
P-202Final choicepicked Doritos Cool Ranch Tortilla Chips during the tasktyped “a snack mix goes with everything so that is my usual pick”
P-210Final choicepicked Chex Mix Traditional Snack Mix during the tasktyped
Answers are checked too
Answer contradicts the visit
1 of 22
Answers the store recording does not back up, shown under Said vs. did. They count against the visit's quality score.
Typed answers that read as AI-written
0 of 6
Answers that look pasted or machine-written are flagged and lower the visit's quality score.
What shoppers didVisit length, distance and products handled per shopper, and the products most picked up and bought in free shopping.ShowHide
How did they behave?
Real shoppers · 40 visits
Visits with a cart add (free shopping)
67.5% ±14
27 of 40 visits
Net basket size
1.30
95% interval 0.92 to 1.68, n = 40 visits
Products picked up
1.93
95% interval 1.55 to 2.30, n = 40 visits
Total gaze dwell (s)
5.5
95% interval 4.9 to 6.1, n = 36 visits
What did they pick?
Real shoppers
Free-shopping stage only: no target was named. Each stage counts visits. Looked at is counted over the visits with eye tracking only. Bought is the product in the cart when free shopping ended. Nothing from practice or from the tasks that followed is counted.
Every measure in detail4 measures, one per part of the visit: open the one you want.ShowHide
What did shoppers do when free to shop?ShowHide
Real shoppers · Data from the Free shopping stage.
n = 40 shoppers (analysed visits). No target was named in this stage: shoppers were asked to shop as they normally would. Every share is a share of shoppers, with its 95% margin in points and its count.
Basket at the end of free shopping, per shopper: median 1 unit, mean 1.20 units (40 shoppers; an empty basket counts as 0; 27 bought something). Basket value: median $3.19, mean $4.07 (40 baskets with a known value).
First pick: the first product the shopper picked up. Top-2 pick: the first or second different product picked up. Handling time: from pick-up to put-back or add to cart, time with the tab hidden left out. Bought: in the cart when free shopping ended. Pick-to-buy: of the shoppers who picked the product up, those who had it in the cart at the end.
In this store picking a product up opens it for a closer look, so "Looked closer" equals "Picked up" for every product and is not shown as its own column.
Turned the pack: the shopper dragged the pack to turn it themselves, out of the shoppers whose holds of the product recorded pack sides. The pack also turns slowly by itself while it is held, so a side that came round by itself is not counted. Saw the back or a side: it then faced the shopper for 1 s or more.
Looking: 36 shoppers of 40 had eye tracking whose accuracy was measured and passed. "Looked at" is the estimated share of those 36 who looked at the product on the shelf, allowing for eye-tracker error; "First look" is the estimated share of first looks, over the about 36 estimated to look at a product; "Look time" is the expected look time per shopper over all 36 and is a secondary detail. The same figures, with the number of looks, the confident-look counts and their confidence grade, are in the looking table below.
First pick
Bought
Pick-to-buy
Each bar has its own denominator: the shoppers who picked that product up.
Where shoppers looked, by product
Gaze comes from a webcam. Looked at, first look and looks are estimates that allow for the measured eye-tracker error: each look counts with the probability that it was on the product, so a look at a small pack is not lost when the gaze estimate spreads over its neighbours. Figures marked Uncertain should be read with care.
n = 36 tracked shoppers whose gaze accuracy was measured and passed (validated). A shopper who looked at nothing still counts. "Looked at" is the estimated share of shoppers who looked at the product, allowing for eye-tracker error: for each shopper, the chance that one or more of their looks was on it, averaged over all shoppers. "First look" is the estimated share of first looks that went to it, and "Looks" the expected number of looks per shopper. "Looked at" is over all 36; "First look" is over the about 36 estimated to look at a product on the shelf; "Looks" is over all 36, and "came back" the estimated share with two or more.
These visits were recorded before the store kept a probability for every settled stretch of gaze, so the estimates are approximate: each confident look counts with its probability, and the rest of the gaze probability counts as looks of the shopper's typical look length (about 1.0 s here).
Confident looks (secondary): looks where at least half of the gaze probability fell on this one product. They miss looks at small packs more often than looks at big blocks, so they are shown for reference only.
Picked up and bought are the same 36 shoppers in the same free-shopping stage (bought: in the cart when free shopping ended). Looking at a product is not the same as taking it in: read every looking figure beside what those shoppers then did.
Look time is a secondary detail: it is not interpretable on its own (a longer look can mean a clearer pack or a harder one to read), so read it beside the share who looked, picked up and bought.
Confidence: Typical look probability = the median, across shoppers, of each shopper's median probability that their looks credited to the product were really on it. Clear: 80% or more. Fair: 65% up to 80%. Uncertain: under 65%, or fewer than 5 shoppers with a look.
Typical gaze error when a look was made: about 1.0 product width (median over 36 shoppers, from 1.0 to 1.0).
Looked at, picked up, bought
Each group is one product; the three bars share one denominator. The margins are in the table.
Where did shoppers go?ShowHide
Real shoppers · Data from the Free shopping stage.
Floor plan
- 1Salty snacks, bay 4: 90% visited (18 of 20)
- 2Salty snacks, bay 3: 45% visited (9 of 20)
Shelves visited: each zone in front of a shelf is shaded by the share of shoppers who visited it (a stay of at least 2 s there). Time on the floor: each 0.5 m cell is shaded by the share of a shopper's time spent in it, averaged so that every shopper counts the same.
n = 40 shoppers with store-flow records.
First shelf approached
n = 40 shoppers with store-flow records.
Measured in free shopping, before any task names a product or a place. A shelf counts as approached when the shopper stayed in the zone in front of it for at least 1 s, slowed to under 0.5 m/s there, or picked up a product from it; walking past does not count.
How the first shelf came to count: stayed 0, slowed down 38, picked up a product 0.
First shelf zone entered at all, also in passing: Salty snacks, bay 4 35, Salty snacks, bay 3 3.
Time from Start to arriving at that first shelf: median 2.7 s (95% interval 2.7 s to 2.7 s; middle half 2.7 s to 2.7 s), n = 38.
First side of the store
The side is where the first shelf approached stands, seen from the entrance in the direction shoppers face on entry: within 20 degrees of that direction is straight ahead. It is the place of the shelf, not a turn of the shopper.
Shelves visited
| Shelf | Side | First approached | Visited | Visits per shopper | Time at the shelf, visitors | Time at the shelf, all shoppers |
|---|---|---|---|---|---|---|
| Salty snacks, bay 4 | Straight ahead | 88% ±10 (35 of 40) | 95% ±8 (38 of 40) | 0.95 (all 40); 1.00 among visitors | median 12.8 s (95% interval 9.5 s to 13.9 s), n = 38 | mean 11.5 s (95% interval 10.0 s to 13.0 s), n = 40 |
| Salty snacks, bay 3 | Left | 8% ±9 (3 of 40) | 43% ±15 (17 of 40) | 0.42 (all 40); 1.00 among visitors | median 9.2 s (95% interval 4.2 s to 10.7 s), n = 17 | mean 3.5 s (95% interval 2.1 s to 4.9 s), n = 40 |
A visit is a stay of at least 2 s of active time in the zone in front of the shelf. Time at the shelf is all active time in that zone over stays of 0.5 s or more (shorter is walking through); shoppers who never stood there count as 0 s in the mean over all shoppers.
Share of the store's shelves a shopper visited: median 50% (95% interval 50% to 100%; middle half 50% to 100%), n = 40; mean 69%.
From visiting a shelf to buying from it
| Shelf | Visited | Picked up something there | Bought something from there | Picked up without a visit |
|---|---|---|---|---|
| Salty snacks, bay 4 | 95% ±8 (38 of 40 shoppers) | 89% ±10 (34 of 38 who visited) | 76% ±14 (26 of 34 who visited and picked up) | 0 |
| Salty snacks, bay 3 | 43% ±15 (17 of 40 shoppers) | 71% ±20 (12 of 17 who visited) | 83% ±20 (10 of 12 who visited and picked up) | 0 |
Each step is counted among the shoppers of the step before it. Bought: a product the shopper added from that shelf is in the cart at the end of free shopping.
Picked up without a visit: shoppers who picked up a product of the shelf without a stay of 2 s in the zone in front of it (from further away, or in passing).
Path and time in free shopping
| Measure | Value |
|---|---|
| Path length | median 3.9 m (95% interval 3.9 m to 5.5 m; middle half 3.9 m to 5.5 m), n = 40 |
| Time in free shopping (active) | median 19 s (95% interval 16 s to 23 s; middle half 13 s to 25 s), n = 40 |
| No movement at all | 5% ±8 (2 of 40 shoppers) |
Distances are virtual metres in a desktop store: use them to compare versions of a store, not as real walking distances. The automatic movement through the entrance is not counted.
A visit with a path under 0.5 m is flagged "no movement". It is reported and stays in every figure.
How fast did shoppers find it?ShowHide
Real shoppers · Data from the Tasks stage.
| By | Found, version A: current pack | Found, version B: new pack |
|---|---|---|
| 5 s | 0% ±8 (0 of 20) | 0% ±8 (0 of 20) |
| 10 s | 5% ±11 (1 of 20) | 35% ±19 (7 of 20) |
| 15 s | 20% ±17 (4 of 20) | 70% ±19 (14 of 20) |
| 20 s | 35% ±19 (7 of 20) | 85% ±15 (17 of 20) |
| 30 s | 65% ±19 (13 of 20) | 95% ±11 (19 of 20) |
| 45 s | 80% ±17 (16 of 20) | 95% ±11 (19 of 20) |
| 60 s | 80% ±17 (16 of 20) | 95% ±11 (19 of 20) |
| 69.6 s | 80% ±17 (16 of 20) | 95% ±11 (19 of 20) |
Shown the pack image and name.
35 of 40 shoppers found it (88% ±10); 5 did not.
Time to find, among the 35 of 40 who found it:
Median 14.8 s, average 17.1 s (35 shoppers; 2 times capped to the shopper's own pace limit; raw average 17.1 s).
95% interval of the average: 14.3 s to 19.8 s.
Relative to own pace: 0.7x (B: new pack) vs 1.4x (A: current pack). Below 1.0x is faster than the shopper's usual pace. Mann-Whitney U test p = under 0.001.
Times count only while the task was running (pauses removed).
Median time to find over all 40 shoppers asked: 16.7 s (95% interval 12.8 s to 21.5 s). 35 of 40 found it; 5 gave up and count as not yet found up to the moment they stopped.
By version: A: current pack 16 of 20 found it, median over all asked 23.2 s · B: new pack 19 of 20 found it, median over all asked 11.4 s.
The median over all shoppers asked is a Kaplan-Meier estimate: a shopper who gave up, reached the time cap or left counts as not yet found up to the moment they stopped, and it reads "not reached" whenever fewer than half found the product. The chart is the observed share of all shoppers asked who had found it by each time; it ends at the share who found it.
Found it without a wrong pick: 33 of 40 shoppers (83% ±12). 2 of 40 picked up a wrong product at least once (5% ±8); mean wrong picks per shopper 0.05.
Where the target stood at fieldwork: A: current pack: shelf level 2 of 4 (counted from the bottom), 3 facings · B: new pack: shelf level 2 of 4 (counted from the bottom), 3 facings. Shelf level and facings are known drivers of find time: compare find times only between like positions.
Self-rated ease ("How easy or difficult was it to find this product?", 1 = Very difficult, 4 = Neither easy nor difficult, 7 = Very easy; a stated answer): mean 4.08, 95% interval 3.61 to 4.54, median 4, n = 40 shoppers who answered. Asked right after the task ended, found or not.
Among the 35 who found it: mean 4.43. Among the 5 who did not: mean 1.60.
Answered 6 or 7 (the two easiest points): 18% ±12 (7 of 40).
By version: A: current pack mean 3.00 (n = 20) · B: new pack mean 5.15 (n = 20). Difference A: current pack less B: new pack: -2.15 (95% interval -2.81 to -1.49), Mann-Whitney U test p = under 0.001.
How easy shoppers said it was to find
After the visitShowHide
Real shoppers · Data from the Questions stage.
About the visit
| Question | Scale | Answers | Mean | Answers per scale point |
|---|---|---|---|---|
| How easy or difficult was it to find products in the store? | 1 to 7: 1 = Very difficult, 4 = Neither easy nor difficult, 7 = Very easy | 40 | 4.03, 95% interval 3.76 to 4.29 | 1: 0 · 2: 0 · 3: 14 · 4: 11 · 5: 15 · 6: 0 · 7: 0 |
| How much did the visit hold your interest? | 1 to 7: 1 = Not at all, 4 = Moderately, 7 = Completely | 40 | 4.17, 95% interval 3.94 to 4.41 | 1: 0 · 2: 0 · 3: 8 · 4: 17 · 5: 15 · 6: 0 · 7: 0 |
n = 40 shoppers who answered the closing questions. Stated answers, given after the visit. Each mean carries its 95% interval; no mean is compared with a fixed number.
Ease of finding products, by version: A: current pack mean 4.15 (n = 20) · B: new pack mean 3.90 (n = 20). Difference A: current pack less B: new pack: 0.25 (95% interval -0.29 to 0.79), Mann-Whitney U test p = 0.357.
Interest held by the visit, by version: A: current pack mean 4.45 (n = 20) · B: new pack mean 3.90 (n = 20). Difference A: current pack less B: new pack: 0.55 (95% interval 0.11 to 0.99), Mann-Whitney U test p = 0.019.
About the products each shopper picked up
| Product | Shoppers who rated it | Liking (1 to 5) | Buy intent (1 to 5) | Value for money (1 to 5) |
|---|---|---|---|---|
| Boulder Canyon Sea Salt Kettle Chips | 17 | mean 3.53, 95% interval 2.92 to 4.14 (n = 17); answers 1: 0 · 2: 4 · 3: 5 · 4: 3 · 5: 5; answered 4 or 5: 47% ±21 (8 of 17) | mean 3.29, 95% interval 2.67 to 3.92 (n = 17); answers 1: 0 · 2: 7 · 3: 1 · 4: 6 · 5: 3; answered 4 or 5: 53% ±21 (9 of 17) | mean 3.59, 95% interval 3.01 to 4.16 (n = 17); answers 1: 0 · 2: 4 · 3: 3 · 4: 6 · 5: 4; answered 4 or 5: 59% ±21 (10 of 17) |
| Cheetos Crunchy | 14 | mean 3.29, 95% interval 2.52 to 4.05 (n = 14); answers 1: 0 · 2: 6 · 3: 2 · 4: 2 · 5: 4; answered 4 or 5: 43% ±23 (6 of 14) | mean 3.43, 95% interval 2.76 to 4.10 (n = 14); answers 1: 0 · 2: 4 · 3: 3 · 4: 4 · 5: 3; answered 4 or 5: 50% ±23 (7 of 14) | mean 3.29, 95% interval 2.63 to 3.94 (n = 14); answers 1: 0 · 2: 4 · 3: 5 · 4: 2 · 5: 3; answered 4 or 5: 36% ±22 (5 of 14) |
| Popchips BBQ Potato Chips | 10 | mean 3.50, 95% interval 2.59 to 4.41 (n = 10); answers 1: 0 · 2: 3 · 3: 2 · 4: 2 · 5: 3; answered 4 or 5: 50% ±26 (5 of 10) | mean 3.80, 95% interval 2.80 to 4.80 (n = 10); answers 1: 0 · 2: 3 · 3: 1 · 4: 1 · 5: 5; answered 4 or 5: 60% ±26 (6 of 10) | mean 2.90, 95% interval 2.19 to 3.61 (n = 10); answers 1: 0 · 2: 5 · 3: 1 · 4: 4 · 5: 0; answered 4 or 5: 40% ±26 (4 of 10) |
| Lay's Classic Potato Chips | 8 | mean 3.25, 95% interval 2.18 to 4.32 (n = 8); answers 1: 0 · 2: 3 · 3: 2 · 4: 1 · 5: 2; answered 4 or 5: 38% ±28 (3 of 8) | mean 3.38, 95% interval 2.49 to 4.26 (n = 8); answers 1: 0 · 2: 2 · 3: 2 · 4: 3 · 5: 1; answered 4 or 5: 50% ±28 (4 of 8) | mean 3.25, 95% interval 2.66 to 3.84 (n = 8); answers 1: 0 · 2: 1 · 3: 4 · 4: 3 · 5: 0; answered 4 or 5: 38% ±28 (3 of 8) |
| Chex Mix Traditional Snack Mix | 7 | mean 3.29, 95% interval 2.41 to 4.17 (n = 7); answers 1: 0 · 2: 1 · 3: 4 · 4: 1 · 5: 1; answered 4 or 5: 29% ±28 (2 of 7) | mean 4.29, 95% interval 3.59 to 4.98 (n = 7); answers 1: 0 · 2: 0 · 3: 1 · 4: 3 · 5: 3; answered 4 or 5: 86% ±24 (6 of 7) | mean 3.71, 95% interval 2.69 to 4.74 (n = 7); answers 1: 0 · 2: 1 · 3: 2 · 4: 2 · 5: 2; answered 4 or 5: 57% ±30 (4 of 7) |
| Cheetos Crunchy Flamin' Hot | 6 | mean 3.17, 95% interval 1.94 to 4.39 (n = 6); answers 1: 0 · 2: 2 · 3: 2 · 4: 1 · 5: 1; answered 4 or 5: 33% ±30 (2 of 6) | mean 3.67, 95% interval 2.40 to 4.94 (n = 6); answers 1: 0 · 2: 1 · 3: 2 · 4: 1 · 5: 2; answered 4 or 5: 50% ±31 (3 of 6) | mean 3.17, 95% interval 2.13 to 4.20 (n = 6); answers 1: 0 · 2: 2 · 3: 1 · 4: 3 · 5: 0; answered 4 or 5: 50% ±31 (3 of 6) |
| Cheez-It Reduced Fat Crackers | 5 | mean 3.80, 95% interval 2.44 to 5.00 (n = 5); answers 1: 0 · 2: 1 · 3: 0 · 4: 3 · 5: 1; answered 4 or 5: 80% ±29 (4 of 5) | mean 3.00, 95% interval 1.76 to 4.24 (n = 5); answers 1: 0 · 2: 2 · 3: 1 · 4: 2 · 5: 0; answered 4 or 5: 40% ±33 (2 of 5) | mean 2.80, 95% interval 1.76 to 3.84 (n = 5); answers 1: 0 · 2: 2 · 3: 2 · 4: 1 · 5: 0; answered 4 or 5: 20% ±29 (1 of 5) |
| Doritos Cool Ranch Tortilla Chips | 5 | mean 4.40, 95% interval 3.72 to 5.00 (n = 5); answers 1: 0 · 2: 0 · 3: 0 · 4: 3 · 5: 2; answered 4 or 5: 100% ±22 (5 of 5) | mean 3.80, 95% interval 2.18 to 5.00 (n = 5); answers 1: 0 · 2: 1 · 3: 1 · 4: 1 · 5: 2; answered 4 or 5: 60% ±33 (3 of 5) | mean 4.40, 95% interval 3.29 to 5.00 (n = 5); answers 1: 0 · 2: 0 · 3: 1 · 4: 1 · 5: 3; answered 4 or 5: 80% ±29 (4 of 5) |
| Pringles Original | 1 | mean 5.00 (n = 1); answers 1: 0 · 2: 0 · 3: 0 · 4: 0 · 5: 1; answered 4 or 5: 100% ±40 (1 of 1) | mean 5.00 (n = 1); answers 1: 0 · 2: 0 · 3: 0 · 4: 0 · 5: 1; answered 4 or 5: 100% ±40 (1 of 1) | mean 2.00 (n = 1); answers 1: 0 · 2: 1 · 3: 0 · 4: 0 · 5: 0; answered 4 or 5: 0% ±40 (0 of 1) |
Each shopper rated only the products they had picked up, so every product has its own n and the shoppers behind two products are not the same people: do not rank the products on these means.
Liking: "How much do you like or dislike it?" 1 = Dislike a lot, 2 = Dislike a little, 3 = Neither like nor dislike, 4 = Like a little, 5 = Like a lot.
Buy intent: "Would you buy it?" 1 = Definitely would not buy, 2 = Probably would not buy, 3 = Might or might not buy, 4 = Probably would buy, 5 = Definitely would buy.
Value for money: "How good or poor is its value for money?" 1 = Very poor, 2 = Poor, 3 = Neither good nor poor, 4 = Good, 5 = Very good.
"Value for money" is a rating of value on five points. It is not a price and not a willingness to pay.
How the closing questions were answered
| Surveys with a score | Median score (0 to 100) | Lowest to highest | Under 70 |
|---|---|---|---|
| 40 | 90 | 90 to 90 | 0 of 40 |
The score starts at 100 and loses points when the survey was answered very fast or the ratings were all (or nearly all) the same. It describes how the survey was filled in, not what shoppers think.
No survey was flagged.
Where shoppers lookedEye tracking heatmap of the shelf, from the shoppers who switched their webcam on.ShowHide
Gaze from a webcam. 36 of 36 tracked visits had their gaze accuracy measured and passed; only those are drawn.
Version A: current pack
n = 18 validated tracked shoppers. Tracker: webcam. Typical accuracy: about 1.0 product width. Areas without colour had little or no gaze recorded on them; it does not mean they were out of view.
Version B: new pack
n = 18 validated tracked shoppers. Tracker: webcam. Typical accuracy: about 1.0 product width. Areas without colour had little or no gaze recorded on them; it does not mean they were out of view.
Free shopping, version A: current pack: looking figures per product
Gaze comes from a webcam. Looked at, first look and looks are estimates that allow for the measured eye-tracker error: each look counts with the probability that it was on the product, so a look at a small pack is not lost when the gaze estimate spreads over its neighbours. Figures marked Uncertain should be read with care.
| Product | Looked at | First look | Looks | Picked up | Bought | Looked at, then picked up, then bought | Look time (secondary) | Confidence |
|---|---|---|---|---|---|---|---|---|
| Cheetos Crunchy | 82% ±17 (an estimated 14.7 of 18 tracked); confident looks: 94% (17 of 18) | 39% ±21 (an estimated 7 of 18 estimated to look at a product); confident looks: 39% (7 of 18); in confident looks typically the 2nd product looked at (median over the 17 shoppers with a confident look at it) | 1.09 expected per shopper, 95% interval 0.80 to 1.38 (all 18, zeros included); came back: 23% ±18 (an estimated 4.1 of 18); confident looks: 1.39 per shopper | 39% ±21 (7 of 18 tracked) | 33% ±20 (6 of 18 tracked) | An estimated 14.7 looked at it; 6.4 of those picked it up; 5.7 of those bought it. 0.6 picked it up without a look | 1.03 s per shopper (expected, from the look probability) | Fair (typical look probability 78%) |
| Boulder Canyon Sea Salt Kettle Chips | 71% ±19 (an estimated 12.7 of 18 tracked); confident looks: 83% (15 of 18) | 17% ±17 (an estimated 3 of 18 estimated to look at a product); confident looks: 17% (3 of 18); in confident looks typically the 3rd product looked at (median over the 15 shoppers with a confident look at it) | 0.92 expected per shopper, 95% interval 0.63 to 1.20 (all 18, zeros included); came back: 21% ±18 (an estimated 3.8 of 18); confident looks: 1.17 per shopper | 44% ±21 (8 of 18 tracked) | 28% ±19 (5 of 18 tracked) | An estimated 12.7 looked at it; 7.2 of those picked it up; 4.3 of those bought it. 0.8 picked it up without a look | 0.84 s per shopper (expected, from the look probability) | Fair (typical look probability 78%) |
| Cheetos Crunchy Flamin' Hot | 45% ±21 (an estimated 8.2 of 18 tracked); confident looks: 56% (10 of 18) | 17% ±17 (an estimated 3 of 18 estimated to look at a product); confident looks: 17% (3 of 18); in confident looks typically between the 2nd and the 3rd product looked at (median over the 10 shoppers with a confident look at it) | 0.52 expected per shopper, 95% interval 0.25 to 0.78 (all 18, zeros included); came back: 6% ±13 (an estimated 1.1 of 18); confident looks: 0.67 per shopper | 17% ±17 (3 of 18 tracked) | 11% ±15 (2 of 18 tracked) | An estimated 8.2 looked at it; 1.7 of those picked it up; 0.9 of those bought it. 1.3 picked it up without a look | 0.45 s per shopper (expected, from the look probability) | Fair (typical look probability 77%) |
| Popchips BBQ Potato Chips | 43% ±21 (an estimated 7.8 of 18 tracked); confident looks: 56% (10 of 18) | 6% ±12 (an estimated 1 of 18 estimated to look at a product); confident looks: 6% (1 of 18); in confident looks typically between the 3rd and the 4th product looked at (median over the 10 shoppers with a confident look at it) | 0.46 expected per shopper, 95% interval 0.23 to 0.69 (all 18, zeros included); came back: 3% ±11 (an estimated 0.5 of 18); confident looks: 0.61 per shopper | 11% ±15 (2 of 18 tracked) | 6% ±12 (1 of 18 tracked) | An estimated 7.8 looked at it; 1.7 of those picked it up; 0.9 of those bought it. 0.3 picked it up without a look | 0.47 s per shopper (expected, from the look probability) | Fair (typical look probability 74%) |
| Doritos Cool Ranch Tortilla Chips | 41% ±21 (an estimated 7.3 of 18 tracked); confident looks: 50% (9 of 18) | 11% ±15 (an estimated 2 of 18 estimated to look at a product); confident looks: 11% (2 of 18); in confident looks typically the 4th product looked at (median over the 9 shoppers with a confident look at it) | 0.41 expected per shopper, 95% interval 0.20 to 0.62 (all 18, zeros included); came back: 0% ±9 (an estimated 0 of 18); confident looks: 0.50 per shopper | 22% ±18 (4 of 18 tracked) | 11% ±15 (2 of 18 tracked) | An estimated 7.3 looked at it; 3.2 of those picked it up; 1.6 of those bought it. 0.8 picked it up without a look | 0.42 s per shopper (expected, from the look probability) | Clear (typical look probability 85%) |
| Lay's Classic Potato Chips | 35% ±20 (an estimated 6.3 of 18 tracked); confident looks: 44% (8 of 18) | 6% ±12 (an estimated 1 of 18 estimated to look at a product); confident looks: 6% (1 of 18); in confident looks typically between the 4th and the 5th product looked at (median over the 8 shoppers with a confident look at it) | 0.35 expected per shopper, 95% interval 0.15 to 0.56 (all 18, zeros included); came back: 0% ±9 (an estimated 0 of 18); confident looks: 0.44 per shopper | 28% ±19 (5 of 18 tracked) | 11% ±15 (2 of 18 tracked) | An estimated 6.3 looked at it; 3.3 of those picked it up; 0.8 of those bought it. 1.7 picked it up without a look | 0.39 s per shopper (expected, from the look probability) | Clear (typical look probability 82%) |
| Pringles Original | 30% ±20 (an estimated 5.4 of 18 tracked); confident looks: 39% (7 of 18) | 0% ±9 (an estimated 0 of 18 estimated to look at a product); confident looks: 0% (0 of 18); in confident looks typically the 3rd product looked at (median over the 7 shoppers with a confident look at it) | 0.30 expected per shopper, 95% interval 0.10 to 0.49 (all 18, zeros included); came back: 0% ±9 (an estimated 0 of 18); confident looks: 0.39 per shopper | 6% ±12 (1 of 18 tracked) | 6% ±12 (1 of 18 tracked) | An estimated 5.4 looked at it; 0.9 of those picked it up; 0.9 of those bought it. 0.1 picked it up without a look | 0.35 s per shopper (expected, from the look probability) | Fair (typical look probability 73%) |
| Chex Mix Traditional Snack Mix (target) | 26% ±19 (an estimated 4.6 of 18 tracked); confident looks: 33% (6 of 18) | 6% ±12 (an estimated 1 of 18 estimated to look at a product); confident looks: 6% (1 of 18); in confident looks typically the 4th product looked at (median over the 6 shoppers with a confident look at it) | 0.30 expected per shopper, 95% interval 0.06 to 0.54 (all 18, zeros included); came back: 4% ±12 (an estimated 0.7 of 18); confident looks: 0.39 per shopper | 6% ±12 (1 of 18 tracked) | 6% ±12 (1 of 18 tracked) | An estimated 4.6 looked at it; 1 of those picked it up; 1 of those bought it | 0.29 s per shopper (expected, from the look probability) | Fair (typical look probability 75%) |
| Cheez-It Reduced Fat Crackers | 21% ±18 (an estimated 3.8 of 18 tracked); confident looks: 28% (5 of 18) | 0% ±9 (an estimated 0 of 18 estimated to look at a product); confident looks: 0% (0 of 18); in confident looks typically the 3rd product looked at (median over the 5 shoppers with a confident look at it) | 0.21 expected per shopper, 95% interval 0.04 to 0.38 (all 18, zeros included); came back: 0% ±9 (an estimated 0 of 18); confident looks: 0.28 per shopper | 0% ±9 (0 of 18 tracked) | 0% ±9 (0 of 18 tracked) | An estimated 3.8 looked at it; 0 of those picked it up; 0 of those bought it | 0.20 s per shopper (expected, from the look probability) | Fair (typical look probability 72%) |
n = 18 tracked shoppers whose gaze accuracy was measured and passed (validated). A shopper who looked at nothing still counts. "Looked at" is the estimated share of shoppers who looked at the product, allowing for eye-tracker error: for each shopper, the chance that one or more of their looks was on it, averaged over all shoppers. "First look" is the estimated share of first looks that went to it, and "Looks" the expected number of looks per shopper. "Looked at" is over all 18; "First look" is over the about 18 estimated to look at a product on the shelf; "Looks" is over all 18, and "came back" the estimated share with two or more.
These visits were recorded before the store kept a probability for every settled stretch of gaze, so the estimates are approximate: each confident look counts with its probability, and the rest of the gaze probability counts as looks of the shopper's typical look length (about 1.1 s here).
Confident looks (secondary): looks where at least half of the gaze probability fell on this one product. They miss looks at small packs more often than looks at big blocks, so they are shown for reference only.
Picked up and bought are the same 18 shoppers in the same free-shopping stage (bought: in the cart when free shopping ended). Looking at a product is not the same as taking it in: read every looking figure beside what those shoppers then did.
Look time is a secondary detail: it is not interpretable on its own (a longer look can mean a clearer pack or a harder one to read), so read it beside the share who looked, picked up and bought.
Confidence: Typical look probability = the median, across shoppers, of each shopper's median probability that their looks credited to the product were really on it. Clear: 80% or more. Fair: 65% up to 80%. Uncertain: under 65%, or fewer than 5 shoppers with a look.
Typical gaze error when a look was made: about 1.0 product width (median over 18 shoppers, from 1.0 to 1.0).
Free shopping, version B: new pack: looking figures per product
Gaze comes from a webcam. Looked at, first look and looks are estimates that allow for the measured eye-tracker error: each look counts with the probability that it was on the product, so a look at a small pack is not lost when the gaze estimate spreads over its neighbours. Figures marked Uncertain should be read with care.
| Product | Looked at | First look | Looks | Picked up | Bought | Looked at, then picked up, then bought | Look time (secondary) | Confidence |
|---|---|---|---|---|---|---|---|---|
| Chex Mix Traditional Snack Mix (target) | 71% ±19 (an estimated 12.7 of 18 tracked); confident looks: 83% (15 of 18) | 17% ±17 (an estimated 3 of 18 estimated to look at a product); confident looks: 17% (3 of 18); in confident looks typically the 3rd product looked at (median over the 15 shoppers with a confident look at it) | 0.91 expected per shopper, 95% interval 0.61 to 1.22 (all 18, zeros included); came back: 18% ±17 (an estimated 3.3 of 18); confident looks: 1.17 per shopper | 33% ±20 (6 of 18 tracked) | 22% ±18 (4 of 18 tracked) | An estimated 12.7 looked at it; 5.6 of those picked it up; 3.8 of those bought it. 0.4 picked it up without a look | 0.83 s per shopper (expected, from the look probability) | Fair (typical look probability 77%) |
| Boulder Canyon Sea Salt Kettle Chips | 70% ±20 (an estimated 12.5 of 18 tracked); confident looks: 78% (14 of 18) | 28% ±19 (an estimated 5 of 18 estimated to look at a product); confident looks: 28% (5 of 18); in confident looks typically the 2nd product looked at (median over the 14 shoppers with a confident look at it) | 1.02 expected per shopper, 95% interval 0.64 to 1.39 (all 18, zeros included); came back: 27% ±19 (an estimated 4.8 of 18); confident looks: 1.28 per shopper | 39% ±21 (7 of 18 tracked) | 17% ±17 (3 of 18 tracked) | An estimated 12.5 looked at it; 6.7 of those picked it up; 3 of those bought it. 0.3 picked it up without a look | 0.95 s per shopper (expected, from the look probability) | Clear (typical look probability 83%) |
| Cheetos Crunchy | 66% ±20 (an estimated 11.8 of 18 tracked); confident looks: 83% (15 of 18) | 33% ±20 (an estimated 6 of 18 estimated to look at a product); confident looks: 33% (6 of 18); in confident looks typically the 3rd product looked at (median over the 15 shoppers with a confident look at it) | 0.76 expected per shopper, 95% interval 0.52 to 1.00 (all 18, zeros included); came back: 11% ±15 (an estimated 1.9 of 18); confident looks: 1.00 per shopper | 28% ±19 (5 of 18 tracked) | 17% ±17 (3 of 18 tracked) | An estimated 11.8 looked at it; 4.2 of those picked it up; 2.5 of those bought it. 0.8 picked it up without a look | 0.66 s per shopper (expected, from the look probability) | Fair (typical look probability 76%) |
| Popchips BBQ Potato Chips | 50% ±21 (an estimated 9 of 18 tracked); confident looks: 61% (11 of 18) | 6% ±12 (an estimated 1 of 18 estimated to look at a product); confident looks: 6% (1 of 18); in confident looks typically the 3rd product looked at (median over the 11 shoppers with a confident look at it) | 0.65 expected per shopper, 95% interval 0.33 to 0.96 (all 18, zeros included); came back: 15% ±16 (an estimated 2.6 of 18); confident looks: 0.83 per shopper | 39% ±21 (7 of 18 tracked) | 11% ±15 (2 of 18 tracked) | An estimated 9 looked at it; 5.3 of those picked it up; 0.7 of those bought it. 1.7 picked it up without a look | 0.60 s per shopper (expected, from the look probability) | Fair (typical look probability 76%) |
| Cheez-It Reduced Fat Crackers | 38% ±20 (an estimated 6.8 of 18 tracked); confident looks: 44% (8 of 18) | 6% ±12 (an estimated 1 of 18 estimated to look at a product); confident looks: 6% (1 of 18); in confident looks typically the 4th product looked at (median over the 8 shoppers with a confident look at it) | 0.41 expected per shopper, 95% interval 0.16 to 0.66 (all 18, zeros included); came back: 4% ±11 (an estimated 0.6 of 18); confident looks: 0.50 per shopper | 22% ±18 (4 of 18 tracked) | 11% ±15 (2 of 18 tracked) | An estimated 6.8 looked at it; 3.6 of those picked it up; 1.8 of those bought it. 0.4 picked it up without a look | 0.39 s per shopper (expected, from the look probability) | Clear (typical look probability 85%) |
| Lay's Classic Potato Chips | 30% ±19 (an estimated 5.3 of 18 tracked); confident looks: 39% (7 of 18) | 0% ±9 (an estimated 0 of 18 estimated to look at a product); confident looks: 0% (0 of 18); in confident looks typically the 4th product looked at (median over the 7 shoppers with a confident look at it) | 0.30 expected per shopper, 95% interval 0.10 to 0.49 (all 18, zeros included); came back: 0% ±9 (an estimated 0 of 18); confident looks: 0.39 per shopper | 11% ±15 (2 of 18 tracked) | 11% ±15 (2 of 18 tracked) | An estimated 5.3 looked at it; 0.9 of those picked it up; 0.9 of those bought it. 1.2 picked it up without a look | 0.25 s per shopper (expected, from the look probability) | Fair (typical look probability 73%) |
| Pringles Original | 27% ±19 (an estimated 4.9 of 18 tracked); confident looks: 33% (6 of 18) | 6% ±12 (an estimated 1 of 18 estimated to look at a product); confident looks: 6% (1 of 18); in confident looks typically the 3rd product looked at (median over the 6 shoppers with a confident look at it) | 0.32 expected per shopper, 95% interval 0.06 to 0.57 (all 18, zeros included); came back: 4% ±12 (an estimated 0.8 of 18); confident looks: 0.39 per shopper | 0% ±9 (0 of 18 tracked) | 0% ±9 (0 of 18 tracked) | An estimated 4.9 looked at it; 0 of those picked it up; 0 of those bought it | 0.26 s per shopper (expected, from the look probability) | Clear (typical look probability 81%) |
| Cheetos Crunchy Flamin' Hot | 26% ±19 (an estimated 4.7 of 18 tracked); confident looks: 33% (6 of 18) | 6% ±12 (an estimated 1 of 18 estimated to look at a product); confident looks: 6% (1 of 18); in confident looks typically the 2nd product looked at (median over the 6 shoppers with a confident look at it) | 0.26 expected per shopper, 95% interval 0.07 to 0.45 (all 18, zeros included); came back: 0% ±9 (an estimated 0 of 18); confident looks: 0.33 per shopper | 6% ±12 (1 of 18 tracked) | 0% ±9 (0 of 18 tracked) | An estimated 4.7 looked at it; 0.8 of those picked it up; 0 of those bought it. 0.2 picked it up without a look | 0.21 s per shopper (expected, from the look probability) | Clear (typical look probability 80%) |
| Doritos Cool Ranch Tortilla Chips | 14% ±16 (an estimated 2.6 of 18 tracked); confident looks: 17% (3 of 18) | 0% ±9 (an estimated 0 of 18 estimated to look at a product); confident looks: 0% (0 of 18); in confident looks typically the 4th product looked at (median over the 3 shoppers with a confident look at it) | 0.14 expected per shopper, 95% interval 0.00 to 0.30 (all 18, zeros included); came back: 0% ±9 (an estimated 0 of 18); confident looks: 0.17 per shopper | 0% ±9 (0 of 18 tracked) | 0% ±9 (0 of 18 tracked) | An estimated 2.6 looked at it; 0 of those picked it up; 0 of those bought it | 0.09 s per shopper (expected, from the look probability) | Uncertain: fewer than 5 shoppers with a look (typical look probability 88%) |
n = 18 tracked shoppers whose gaze accuracy was measured and passed (validated). A shopper who looked at nothing still counts. "Looked at" is the estimated share of shoppers who looked at the product, allowing for eye-tracker error: for each shopper, the chance that one or more of their looks was on it, averaged over all shoppers. "First look" is the estimated share of first looks that went to it, and "Looks" the expected number of looks per shopper. "Looked at" is over all 18; "First look" is over the about 18 estimated to look at a product on the shelf; "Looks" is over all 18, and "came back" the estimated share with two or more.
These visits were recorded before the store kept a probability for every settled stretch of gaze, so the estimates are approximate: each confident look counts with its probability, and the rest of the gaze probability counts as looks of the shopper's typical look length (about 1.0 s here).
Confident looks (secondary): looks where at least half of the gaze probability fell on this one product. They miss looks at small packs more often than looks at big blocks, so they are shown for reference only.
Picked up and bought are the same 18 shoppers in the same free-shopping stage (bought: in the cart when free shopping ended). Looking at a product is not the same as taking it in: read every looking figure beside what those shoppers then did.
Look time is a secondary detail: it is not interpretable on its own (a longer look can mean a clearer pack or a harder one to read), so read it beside the share who looked, picked up and bought.
Confidence: Typical look probability = the median, across shoppers, of each shopper's median probability that their looks credited to the product were really on it. Clear: 80% or more. Fair: 65% up to 80%. Uncertain: under 65%, or fewer than 5 shoppers with a look. 1 product here is graded Uncertain: usually a single facing on a distant shelf, small on screen relative to tracking accuracy.
Typical gaze error when a look was made: about 1.0 product width (median over 18 shoppers, from 1.0 to 1.0).
Replay one visit (version A: current pack)
One visit at a time, never a pooled path. A stop is the middle of the product a look was credited to, not the exact point on the pack.
Based on 36 of 40 visits with eye tracking. Gaze comes from a webcam and is read at the level of a product or shelf zone. It shows where shoppers looked, not lab-grade eye tracking.
The view opens on free shopping, where no target was named. Pick a task under Stage to see that task alone: a task that named a product adds looks on that product. The funnel above covers free shopping alone.
Decision readout and trustThe hypothesis, the key figure with its margin, the decision grade and which visits were analysed.ShowHide
Hypothesis
More shoppers notice Chex Mix Traditional Snack Mix first in the new pack than in the current pack.
Trust: Directional · key figure Version B: new pack 67% ±20 (n = 18) against Version A: current pack 6% ±12 (n = 18) · decision grade: key comparison settled · 40 analysed real visits (smallest version decides the level) · 0 simulated shopper visits
Version B: new pack is ahead of Version A: current pack on "first look on the target" (67% ±20 against 6% ±12), and the difference holds after correcting for the other comparisons.
Measured
Free shopping, Where shoppers went, Time to find, After the visit
Real shoppers
40 analysed visits
Trust
Directional
Key figure: Version B: new pack 67% ±20 (n = 18) against Version A: current pack 6% ±12 (n = 18)
Decision grade
Version B: new pack is ahead of Version A: current pack on "first look on the target" (67% ±20 against 6% ±12), and the difference holds after correcting for the other comparisons.
Same conclusion with excluded visits included: no visit was excluded.
The key figure is known to ±20 points (the wider of the two versions): directional. Real and simulated results are always shown separately.
How to read this report. These results come from a desktop virtual store. Use them to compare versions and products. Times and shares describe behaviour in this store and are not forecasts of in-store levels. Gaze comes from a webcam and shows where shoppers looked.
43 started, 40 completed, 40 analysed. No completed visit was excluded.
Method and data qualityHow the study was run, how the figures are worked out, and how much you can trust them.ShowHide
Visits
- Visits: 43 started, 40 completed, 40 analysed.
- 3 visits started but did not finish (3 recorded as left before the end). They are not in any figure: this is drop-out, not an exclusion.
- 0 of 40 completed visits excluded. No completed visit was excluded.
- Median active visit time 243 s over 40 completed visits; a visit under 73 s counts as too fast.
- Every figure in this report rests on the analysed visits. "Show with excluded visits" recomputes the report on every completed visit.
Session script and stages
- Session script: version 2. Every shopper went through the same four stages in the same order: Practice, Free shopping, Tasks, Questions. Inside a stage the tasks ran in one fixed order, least revealing first. Answer options and product lists were shuffled per shopper and the order shown was recorded; task order was not shuffled.
- Median time per shopper in each stage (active time, pauses left out): Practice 1.2 min (n = 40); Free shopping 35 s (n = 40); Tasks 37 s (n = 40); Questions 47 s (n = 40).
- Practice: 40 of 40 shoppers completed every practice step (100%); 0 continued without completing it.
- Practice took place in a separate practice area with plain non-test items. Nothing done there is in any figure, funnel, journey or heatmap of this report; the practice time is used only to judge each shopper's own pace.
- Free shopping came before any task named a product: shoppers were asked to shop as they normally would. The cart from free shopping was saved and emptied before the tasks.
- Find cue: Find it among look-alikes: shown the pack image and name (40 of 40 shoppers).
Rules applied
- Rule set 2026-10-b, fixed before fieldwork: it was stored with the study when the study went live and has not changed since.
- These thresholds are our own rules, stated in full. They are fixed in advance and are never tuned after looking at a study's results.
- Visit too fast to be a real walk: Active visit time under 30% of the study median (median over all completed visits, versions pooled). Runs only with 5 or more completed visits.
- Recording fault: Records lost for good during the visit (lost writes minus replayed writes above 0), or the visit clock was flagged as unreliable.
- Repeat visit by the same shopper: When one shopper has several completed visits, only the earliest completed visit counts.
- Flagged as a likely bot at registration: The registration check flagged the shopper (quality flag "flagged") or its hidden form field was filled in.
- Straight-lined survey answers: Every rating in the exit survey was identical (4 or more ratings). Only this visit's stated answers (ratings and survey) are left out; what the shopper did stays in.
- Reading check not passed: When the study has the optional reading check on (one question that tells shoppers which answer to choose) and the shopper chose another answer. Only this visit's stated answers (ratings, questions and survey) are left out; what the shopper did stays in.
- Did not complete (time cap): A find, choice or walk-to-zone task with more than 300 s of active time counts as not completed.
- Accidental result: A find, choice or walk-to-zone task completed in under 0.3 s is left out for that shopper.
- Unusable time: A time that is missing, negative, not a number, or recorded on a clock flagged as unreliable is left out.
- Single look too long: A single look longer than 10 s is dropped.
- Too few valid gaze samples: A tracked visit with under 50% valid gaze samples is dropped from the looking figures only.
- Eye tracking not accurate enough: After calibration the error is measured on 5 fresh points. A median error above 12% of the window diagonal, after one automatic retry, switches eye tracking off for the visit (behaviour stays in). A tracked visit whose accuracy check did not pass is left out of the product-level looking figures.
- Product-level gaze from validated visits only: A tracked visit counts toward the product-level looking figures only when its gaze accuracy was measured on fresh validation points and passed. What the shopper did (picked up, bought) stays in every other figure.
- Legacy tracking, accuracy not measured: A tracked visit with no accuracy check on record (an older visit) is reported apart as legacy tracking and is never pooled with validated visits.
- Eye tracking drifted during the visit: The error is re-checked during the visit (re-check dots between stages, product pick-ups, button presses). While the rolling error stays above 12% of the window diagonal for 3 checks in a row, no looks are credited.
- Product-level look: Each gaze sample is shared out over the products in proportion to how likely it is on them given the error measured at that moment. A look at a product is counted only when that product has the highest probability, at least 50%, for 120 ms or longer, on the shelf (not the pack held in hand), outside the task intro and outside practice.
- Look without attribution in a validated visit: A look row of a validated visit that carries no probability attribution is left out.
- Confidence grade of a product's gaze figures: Typical look probability = the median, across shoppers, of each shopper's median probability that their looks credited to the product were really on it. Clear: 80% or more. Fair: 65% up to 80%. Uncertain: under 65%, or fewer than 5 shoppers with a look.
- Smallest group for a gaze density map: With fewer than 5 validated shoppers with a gaze track on a shelf, the map shows each shopper's track as a faint path, not a density.
- Time capped to the shopper's own pace limit: Expected time = the shopper's pace factor x the task median (task median needs 8 or more shoppers; pace factor = median of the shopper's other timed results relative to their task medians, 1 with fewer than 2). A time above 3x or below one third of the expected time is kept and capped at that limit for the average. Medians use the raw times.
- Trust level: Early read: fewer than 10 analysed visits in the smallest version, or a 95% margin of the key estimate above ±25 points. Directional: margin above ±10 and up to ±25. Solid: margin at or below ±10.
- Decision grade stamp: Earned only when the key comparison is settled at 0.05 with 10 or more analysed visits per version. One version: the leader against the runner-up (multinomial top-two test). A/B: the primary metric, on a metric with a direction, after the Holm correction.
- A gaze figure as the headline: A gaze figure can carry the headline or the decision grade stamp only when it is not graded Uncertain and rests on 10 or more validated tracked shoppers. Otherwise the headline uses the behavioural figure and says why.
- Gaze thresholds of rule set 2026-10-b the looking figures use:
- A look at a product: the product has the highest look probability, at least 50%, for at least 120 ms.
- Confidence grade of a product's looking figures: Clear from a typical look probability of 80%, Fair from 65%, Uncertain below.
- Fewest shoppers with a look for a grade above Uncertain: 5.
- Fewest validated tracked shoppers before a looking figure can be the headline: 10.
- Fewest shoppers with a gaze track on a shelf for a density map: 5.
- Density map grid on a shelf front: 200 cells across, 120 cells up.
- Blur of one gaze track point: the gaze error measured at that moment, out to 3 standard deviations, never under 0.5 cells.
- Typical look length of the looking estimates when no confident look gives one: 600 ms.
- Time one gaze track point stands for: at most 1000 ms; a lone point stands for 200 ms.
Checks on single results
- The same checks apply to this page: a single look longer than 10 s is dropped, and a tracked visit with under 50% valid gaze samples is left out of the looking figures (0 of 36 checked).
- The checks on timed task results (time cap, accidental results, pace limits) and their counts are listed in the full report.
Versions and assignment
- Versions: 2. Visits per version: A: current pack 20, B: new pack 20.
- Split check: the visits per version are within what an equal split gives by chance (chi-square test, p = 1.000).
- Assignment: alternating (balanced, not random): each new shopper got the version with the fewest real visits so far, and the first version on a tie, so the next version could be told from the last one. The version was fixed before the visit started, a returning shopper kept it, and each shopper saw one version.
Tests and intervals
- Shares carry a 95% margin in points (Wilson score interval); means carry a 95% interval (t-based up to 30 values).
- Times: the median comes first and uses the raw times. Averages use the times capped to each shopper's own pace limit, with the raw average beside them.
- Between two versions: shares by Fisher's exact test when an expected count is under 5, else the two-proportion z-test; times, ratings and pace scores by the Mann-Whitney U test (a rank test). All two-sided at 0.05; on the Compare tab the p-values are corrected together with the Holm method.
- Within one version: the leading product against the runner-up by the multinomial top-two test (both shares come from the same shoppers).
- Visits needed to settle a gap are worked out for the gap observed, at 80% power and 0.05 two-sided; between two versions, at the stricter level the Holm correction leaves the key comparison.
Eye tracking
- Eye tracking (webcam, optional, with the shopper's consent): 36 of 40 visits tracked.
- Calibration score (1 is a perfect fit on the calibration points): median 0.90, range 0.90 to 0.90, over 36 tracked visits.
- Calibration error (RMSE): not recorded for these visits.
- Sampling interval (time between two gaze samples): median 66 ms over 399 looks, about 15 samples a second.
- Share of valid gaze samples (samples with a usable face and gaze estimate): median 93%, range 90% to 96%, over 36 tracked visits.
- Webcam gaze is read at the level of a product or a shelf zone. It shows where shoppers looked; it does not show what they thought of it.
Gaze accuracy and product-level looking
- Tracker: webcam (no dedicated eye tracker). Tracked visits: 36 of 40 visits. Validated (gaze accuracy measured on fresh validation points and passed): 36. Accuracy check not passed: 0. Legacy tracking, accuracy not measured: 0. Dropped for too few valid gaze samples: 0.
- Product-level looking figures (looked at, first look, number of looks) rest on the validated visits only. Each gaze sample is attributed to products in proportion to how likely it was to be on them given the gaze error measured at that moment. Looked at, first look and number of looks are estimates that allow for that error: each look counts with the probability that it was on the product, so a look at a small pack is not lost when the gaze estimate spreads over its neighbours. Confident looks (at least half of the gaze probability on one product) are kept as a secondary figure.
- Validated gaze error (the typical miss on fresh validation points, as a fraction of the screen diagonal): median 3.1%, range 2.2% to 4.0%, over 36 validated visits. In degrees of visual angle: median 1.3, range 0.9 to 1.6 (36 visits).
- Gaze error over the visit, re-checked during the visit: per-visit median 3.0% (range 3.0% to 3.0%); largest error in any visit 7.0%.
- Gaze error in product-facing widths, when a look was made: median 1.0, range 1.0 to 1.0, over 36 shoppers. A value above 1 means the typical miss is wider than one facing: single facings are then graded Uncertain and blocks of several facings carry the product-level figures.
- Accuracy re-checks during the visit: median 8 per visit, range 5 to 11, over 36 visits.
- Time gaze was suspended because the error was too large: median 0.0 s per visit, range 0.0 s to 0.0 s. Nothing is recorded as a look during that time.
- Gaze sampling interval over the visit: not recorded (the interval inside each look is given above, where recorded).
- Gaze estimator version: hp-ridge-1 (36 visits).
- Gaze rules of rule set 2026-11-b, added after this study's rule set 2026-10-b was fixed and applied to every version alike:
- Looked at, first look and looks (estimates): A look is one settled stretch of gaze on the shelf of 120 ms up to 10 s, ended by a fast eye movement, a gap or the gaze moving more than 7% of the window; its probability on a product is the mean of that product's probability over the stretch. The looks of one task are then read together: the shopper's attention in the task is fitted to all of them (each look's probabilities read against the products' screen areas, with one pseudo-look spread by area as the prior), and each look is re-read with that attention as its prior, so repeated looks at one pack do not pile up on its neighbours. Looked at: for each shopper the chance that one or more looks were on the product (1 minus the product over the looks of 1 minus that re-read probability), averaged over all eligible shoppers. First look: the chance that the first look at any product was on it, over the expected number of shoppers with a look. Looks: the expected number per shopper. No 50% threshold applies, so a look at a small pack whose probability spreads over its neighbours still counts with the probability it has. Bigger blocks and more facings are not adjusted for. Visits recorded before the store kept these probabilities: each confident look counts with its probability, and the rest of the gaze probability counts as looks of the shopper's typical look length (the scope's when the shopper has no confident look, 600 ms when no shopper has one). The 95% interval of the expected looks per shopper adds each shopper's own uncertainty (the sum over their looks of p x (1 minus p), p the look's probability on the product) to the spread between shoppers.
- Visits under the gaze rules of rule set 2026-10-b (listed under "Rules applied"), applied to every version alike: Product-level gaze from validated visits only: 36. Eye tracking not accurate enough: 0. Legacy tracking, accuracy not measured: 0.
- Gaze quality by version: A: current pack: 18 of 18 tracked visits validated (100%), median validated gaze error 3.1% of the screen diagonal (n = 18); B: new pack: 18 of 18 tracked visits validated (100%), median validated gaze error 3.1% of the screen diagonal (n = 18).
- Gaze quality is similar across the versions. Our rule: the versions differ when the validated share is 15 points or more apart, or the median validated gaze error is at least 1.25 times larger in one version; versions with fewer than 5 tracked visits are not compared.
- Gaze density map: each track point is blurred by the gaze error measured at that moment and weighted by the time it stands for (at most three times the usual step of its row, and never more than 1 s); each shopper's map is divided by that shopper's tracked shelf-viewing time, so it sums to their share of that time on the shelf front, and the maps are averaged so every shopper counts once. Grid: 200 by 120 cells per shelf front.
- Looking at a product is not the same as taking it in: every looking figure is shown beside what the same shoppers picked up and bought. Look time and share of look time are not interpretable on their own and are shown as a secondary detail only.
Ease of use and comfort
- Ease of using the store controls ("How easy or difficult was it to use the controls of the store?", 1 = Very difficult, 4 = Neither easy nor difficult, 7 = Very easy): mean 5.33, 95% interval 5.01 to 5.64, median 5, n = 40 shoppers. Answers per point: 1: 0 · 2: 0 · 3: 0 · 4: 11 · 5: 10 · 6: 14 · 7: 5.
- Discomfort ("Did you feel any discomfort, such as dizziness or eye strain?"): No 39, A little 1, Yes 0, of 40 shoppers who answered. 1 reported some discomfort (3% ±6); each of those visits carries the flag discomfort_reported.
- The flag excludes no visit from any figure: no rule of the study's rule set uses it. It is there so the results can be checked without those visits.
- Sensitivity check, shoppers with discomfort against shoppers without (mean on 1 to 7): ease of using the controls: 6.00 (n = 1) with discomfort against 5.31 (n = 39) without; difference 0.69, Mann-Whitney U test p = 0.498; ease of finding products: 5.00 (n = 1) with discomfort against 4.00 (n = 39) without; difference 1.00, Mann-Whitney U test p = 0.269; interest held by the visit: 3.00 (n = 1) with discomfort against 4.21 (n = 39) without; difference -1.21, Mann-Whitney U test p = 0.148.
Setting, devices and dates
- Setting: a desktop virtual store in the browser, walked with keyboard and mouse.
- Devices and browsers: not recorded for these visits.
- Fieldwork: 2026-10-01 to 2026-10-02 (study went live on 2026-10-01).
- Measurement rule set: 2026-10-b.
This study follows established shopper-research protocols for immersive consumer research, using an immersive virtual store simulation (browser-based). Peer-reviewed research supports immersive virtual-store methods as a cost-effective alternative to physical store testing (Oliveira et al., 2020).
Eye tracking is optional. The study owner decides whether to switch it on, and each shopper can decline it. When a shopper accepts, gaze is recorded with the webcam after a short calibration at the start of the visit. Looking figures come only from visits with eye tracking; all other results use every analysed visit.
Webcam gaze is read at the level of a shelf zone or product area: enough to show which zone or product area shoppers looked at, but not to resolve points on a single pack with lab-grade precision. Time to first look and look time are computed per zone or product area across the visits with eye tracking.
The looking figures come from the shoppers in the example who switched their webcam on. Webcam gaze is optional and is not part of this demo.