Experiment Design · Consumer Research
Does a clothing company that offers customization or tailoring lead to a surge in consumer purchasing? We ran a randomized pilot to find out, got a promising number, and decided it didn’t prove anything yet. Then we designed the experiment that could, written as a class pitch addressed to the CEO of Fast Retailing.
My part
The idea, and the setup research, from working the alterations counter
Why it matters
UNIQLO did $20.2B in sales; North American revenue grew 32.8%
Pilot
Qualtrics survey, 121 students
Scaled design
Store-level randomized trial, 6–8 months
Pitch addressed to
Tadashi Yanai, CEO of Fast Retailing
The pilot
Participants were randomly shown one of two versions of the same four product listings — $39.99 (20% off), delivered to campus — and asked whether they’d buy. The only difference was the offer to alter the length.
Control
These UNIQLO Wide Straight Jeans are available in Light Blue and Dark Blue at a 20% discount. One jean will cost $39.99. If you decide to buy, payment will be sent through Venmo, and the pants will be delivered to campus within a week. Would you purchase these jeans?
Available in 30-inch waist or 34-inch waist, with a standard 34-inch length.
Treatment
These UNIQLO Wide Straight Jeans come with a customizable option to alter the length, available in Light Blue and Dark Blue at a 20% discount. One jean will cost $39.99. A minimum of 2 inches must be removed for the alteration to be completed. If you decide to buy, payment will be sent through Venmo, and the pants will be delivered to campus within a week. Would you purchase these customizable jeans?
Available in 30-inch waist or 34-inch waist, with a standard 34-inch length (adjustable).
Pair 1 as it appeared in each survey. Highlighted text is what only the treatment group read — including a condition: at least two inches had to come off.
Results
Treatment participants said yes more often. But with 121 people, which pants someone was looking at mattered more than which group they were in.
Said yes to buying
Share of each group who would buy, across all four pairs of pants.
| Group | Said yes | Participants | Rate |
|---|---|---|---|
| Control | 8 | 62 | 12.9% |
| Treatment | 12 | 59 | 20.3% |
Were the groups comparable?
| Control | Treatment | |
|---|---|---|
| Participants | 62 | 59 |
| Average age | 21.1 | 22.32 |
| Share male | 49.2% | 65.1% |
| Would buy | 12.9% | 20.3% |
Close on age; sixteen points apart on gender. That gap is the first thing a skeptic would point at.
Purchase rate by pair of pants
Each participant saw four pairs. The spread between pairs is larger than the spread between groups.
Read off the team’s chart, so accurate to about a point. Pairs aren’t named because the two surveys listed the pants in different orders.
| Pair | Control | Treatment |
|---|---|---|
| Pair 1 | ≈15.3% | ≈12.9% |
| Pair 2 | ≈20.3% | ≈21.0% |
| Pair 3 | ≈6.8% | ≈19.5% |
| Pair 4 | ≈1.7% | ≈14.5% |
What we learned
The pilot’s most useful output wasn’t the purchase rate. It was a list of everything besides tailoring that was deciding the answer.
Four in five treatment participants still said no. The team read the pilot as no clear effect: whatever the tailoring option did, it wasn’t enough on its own to change a decision.
Purchase rates swung far more between pairs than between conditions. Price, style, and simply not being in the market for pants were doing most of the work.
Participants said they doubted the pants would ever arrive if they paid. A hypothetical purchase through a form measures trust in the form as much as interest in the product.
Age was close across groups, but the treatment group skewed more male (65% vs 49%). With 121 people, a gap like that is hard to separate from the effect being measured.
The pitch · addressed to Tadashi Yanai, CEO of Fast Retailing
A live experiment fixes the pilot’s two biggest problems at once: real pants in front of real shoppers, and no Venmo leap of faith.
Treatment stores
Customers can have their clothes tailored to fit — pants, shirts and more.
Control stores
Customers buy the sizes UNIQLO stocks, as they do today.
778
Total sample
389 per group
80%
Power
Chance of detecting a real effect
5%
Significance
False-positive rate accepted
Variance estimated from the pilot’s purchase rates: 12.90% control, 20.34% treatment. The pilot didn’t prove the effect — it sized the test that could.
Non-compliance
Customers in treatment stores may never use the tailoring.
EffectDilutes the measured effect.
FixTrack uptake and analyze users and non-users separately to recover the effect of actually using it.
Spillovers
Control-store customers hear about tailoring and travel to a treatment store.
EffectInflates control, shrinks the gap.
FixDraw geographic boundaries between clusters and monitor where participants shop.
Seasonality
Holiday and back-to-school spikes land unevenly.
EffectSkews the estimate.
FixRun across every relevant season.
Moving the bottom line
An 8% sales lift from tailoring would be about $1.616B on UNIQLO’s $20.2B.
That 8% is the team’s projection for the pitch, not a result the pilot produced. The pitch was upfront that running it would be expensive.
One service, four angles
This experiment sits at one step of a much longer service: the moment someone decides to buy. Switch the lens to see what happens to an order after that — and how close the service already runs to its limit.
Tailoring Purchase Experiment covers 1 of 12 steps.
Customer
Floor staff
Specialist
Manager
Send one lockstitch order through the service and watch where the minutes go.
More work inDataStrategyUX research
Next Project
← All workLike work like this? Tell me what you're building.
Say hi →