Fernway wants to replace its vanilla oat yogurt and has two candidate formulas. An in-home use test puts all three products in real kitchens, blind and coded, so people eat them the way they normally would. Every household first tastes all three, one per day in a rotated order (sequential monadic), then lives with one of them for four weeks. The first taste tells you what wins a spoonful. The four weeks tell you what still gets eaten.
Control 482
The product on shelf today. Included so both new formulas are judged against what buyers already know, not against each other alone.
Formula A 317
New culture blend and added pea protein for a thicker, spoonable texture. Sugar unchanged at 9 g per cup.
Formula B 956
Smoother texture and a rounder vanilla, with sugar raised from 9 g to 12 g per cup.
Who took part Measured
116 of 120 households finished all four weeks (97%). Two moved, one lost fridge space, one stopped answering.
Screener criteria
Placement and rotation
Six tasting orders, 20 households each, so every product appears first, second and third equally often. After tasting, households are split into three cells of 40, balanced on Fernway buyers and children.
| Order | Day 1 | Day 2 | Day 3 | HH |
|---|
Schedule, week by week
What each household does
Why video and not a longer diary: people do not write down the morning they skipped breakfast yogurt. They will show you the cup they pushed to the back of the fridge.
This is the ballot each household answered after each tasting cup, three times over three days. Overall, flavor and texture liking use the standard 9-point scale. Sweetness, thickness and vanilla flavor use just-about-right scales: the middle box is the target, and the two sides say which way a product misses. The penalty analysis at the bottom asks how much liking each miss actually costs.
1. Overall, how much do you like or dislike this yogurt?
2. How much do you like or dislike the flavor?
3. How much do you like or dislike the texture?
4. How would you describe the sweetness? just about right
5. How would you describe the thickness? just about right
6. How would you describe the vanilla flavor? just about right
7. What, if anything, did you like or dislike about this yogurt?
Liking means · first taste, n=120 each Measured
| 9-point scale | 482 | 317 | 956 |
|---|---|---|---|
| Overall liking | 6.4 | 6.9a | 7.2ab |
| Flavor liking | 6.5 | 6.7 | 7.4ab |
| Texture liking | 6.1 | 7.3ac | 6.9a |
| Top two box, overall (8 or 9) | 28% | 40% | 48% |
| Preferred after all three | 20% | 35% | 45% |
Letters mark a mean significantly higher than that product at 95% confidence: a = 482, b = 317, c = 956. Preference: 24, 42 and 54 of 120 households.
Overall liking, full scale Measured
At first taste, 956 wins on every liking measure but texture. If the test stopped here, it would be the launch.
Just-about-right results · first taste, n=120 each Measured
The two "too little" boxes and the two "too much" boxes are combined. A product is in good shape on an attribute when about 70% or more say just about right.
Penalty analysis · what each miss costs Measured
Each dot is one way a product misses: for example, "317 not sweet enough." Across: the share of households who said so. Up: how much lower their overall liking was than the households who said just about right. Dots in the shaded corner are worth fixing: at least 20% of people, and at least half a point of liking lost.
Flagged misses Measured
| Product, miss | Share | Liking | Drop |
|---|
Liking is the overall mean among households in that group; the drop is against the just-about-right group for the same product and attribute. 482 too thin is the largest single penalty in the study, and both new formulas fix it.
956 too sweet is flagged even at first taste. One cup, 30% already say too sweet. Week by week shows what happens next.
After tasting, each household lived with one product for four weeks, 40 households per product. Every week they rated it and reported how many of their 8 cups they ate. 956 held for two weeks, then fell in week three, on liking and on cups eaten at the same time. 317 went the other way. The videos say why.
956 liking, week 2 to 3
−0.8
7.0 to 6.2, then 6.0 in week 4
956 cups eaten, week 4
2.9 of 8
down from 4.8 in week 1
317 cups eaten, week 4
4.7 of 8
up from 4.2 in week 1
Videos recorded
201
from 104 of 116 households
Overall liking by week, 9-point Measured
Cups eaten per week, of 8 delivered Measured
Counted from the leftover-cup photo at each check-in, not from memory.
Videos by week · went right vs went wrong Measured
956 "went wrong" videos jump from 7 in week 2 to 17 in week 3. Households were never asked about sweetness in the video prompt. They brought it up on their own.
956 · what went wrong, weeks 3 and 4 Measured
31 videos, each coded to its main point.
Sweetness, asked again at week 4 Measured
Top bar: day of tasting, same cell (n=40). Bottom bar: week 4 (n=39 and 38). "Not sweet enough" for 317 halved on its own. "Too sweet" for 956 nearly doubled.
Video diary clips Illustrative
Transcribed from phone videos, lightly trimmed. Household numbers are study IDs.
Weekly check-in table Measured
| Product | Week 1 | Week 2 | Week 3 | Week 4 | Change |
|---|
Top line: overall liking. Second line: cups eaten of 8. Third line: households answering that week.
The readout answers the question Fernway asked: which formula to launch, and what to fix first. It puts the first-taste result next to the fourth-week result, because they point in different directions.
Recommendation
Launch 317,
after a vanilla lift
317 was second at first taste and first by week four: liking rose from 6.9 to 7.3, households ate more of it each week, and 59% would probably or definitely buy it. 956 won the first spoonful, then lost 40% of its servings by week four once the sweetness wore thin. Do not launch the first-taste winner. Fix 317's light vanilla, keep the sugar where it is, and confirm in a short follow-up test.
Scorecard Measured
| Measure | 482 current | 317 thicker | 956 sweeter |
|---|---|---|---|
| Overall liking, first taste (n=120) | 6.4 | 6.9 | 7.2 |
| Preferred after tasting all three | 20% | 35% | 45% |
| Overall liking, week 4 | 6.2 | 7.3 | 6.0 |
| Cups eaten of 8, week 4 | 3.6 | 4.7 | 2.9 |
| Would probably or definitely buy, week 4 | 38% | 59% | 34% |
| Households finishing (of 40) | 39 | 39 | 38 |
| Flagged just-about-right misses | Too thin, vanilla weak | Not sweet enough, vanilla weak | Too sweet |
Purchase intent: 15 of 39, 23 of 39 and 13 of 38 households. The week-4 gap between 317 and 956 on liking and cups eaten is significant at 95% confidence.
What to fix, in order
What 956 is still good for
956 had the best flavor liking at first taste and the most "went right" videos in week 1, many from children. Households called it dessert. That points to an occasion, not a replacement: a smaller cup positioned after dinner may be worth a separate concept test.
Next step
Where this test could mislead