Work/Gap 3D Fit
Case study ( 05 ) · Gap
Turning sizing uncertainty into a 61% add-to-bag decision.
Technical product designer on Drapr, a Y Combinator 3D-fit startup that Gap Inc. acquired. I led the redesign of the 3D Fit flow so shoppers could trust a size. An A/B test showed a 61% lift in add-to-bag.
Overview
Fit is the biggest friction in online denim. Showing it wasn't enough.
Drapr built a 3D avatar that lets shoppers try on clothes. Gap Inc. acquired it in August 2021 to improve fit across its brands, and called fit “the number one point of friction for customers” (Gap Inc. press release, 26 August 2021).
My hypothesis: a 3D avatar shows fit, but shoppers buy on trust. So I designed the flow around what a shopper needs to believe a size, then tested it against the old one.
Drapr: 3D avatar try-on
A Y Combinator startup. Gap acquired it in August 2021.
Hypothesis: ask less, prove more
Little upfront. Explain the size when trust is at stake.
Result: +61% add-to-bag
An A/B test of the new flow against the old one.
Problem
The old flow showed fit. Shoppers still felt risk.
Denim made sizing harder: one size had to explain multiple body zones without turning checkout into a long form.
Measurements were not enough
Charts were precise; comfort was not.
Fit was multi-zone
Waist, hip, thigh, calf, and inseam all mattered.
The answer needed a reason
A size recommendation without evidence felt risky.
Every extra input had a cost
More personalization could also mean more drop-off.
Decisions
I asked less upfront and showed more evidence when trust was at stake.
The key move was sequencing: what to ask, what to defer, and when to explain. I tested each call with participants before it shipped.
Decision 1: Accuracy vs completion rate
User need
Credible sizing.
Business need
Purchase momentum.
Constraint
Too many inputs caused drop-off.
Compromise
Baseline first, tuning optional.
Participants said they felt more confident seeing the numbers when they entered their measurements, especially those who knew a few of them, than relying on a visual approximation.
Keep the dimensions visible by default. It is less ambiguous than showing no numbers at all.
Complexity moved later, after the value was clear.
Decision 2: Explainability vs purchase velocity
User need
Evidence behind the size.
Business need
Fast purchase flow.
Constraint
Too much rationale slowed action.
Compromise
Fast answer, optional reason.
Almost everyone opened it out of curiosity, even when the size matched what they usually wear.
Add a brief explanation of the recommendation, especially when it falls outside their usual range, to build trust in the tool.
The recommendation became inspectable without becoming heavy.
Decision 3: Fit fidelity vs mobile performance
User need
Realistic body-zone fit.
Business need
Reach mobile shoppers.
Constraint
Fidelity slowed performance.
Compromise
Use copy to support visuals.
Participants knew to click the question mark for more information and valued the clarity of what fitted, loose and relaxed mean.
Add definitions for the fit descriptors, so shoppers know what each one feels like.
Fit copy became interface logic: short, local, decision-tied.
Solution
The final flow made fit easier to trust without adding friction.
Each step either improved the model or helped shoppers verify the recommendation.
How it works
- 01
Baseline input
How tall are you?
- 02
Fit calibration
- 03
Optional tuning
Further customize your avatar
- 04
Explainable fit
Why this recommendation, and what tight, fitted and loose feel like
Impact
An A/B test showed a 61% lift in add-to-bag.
The strongest signal was not more 3D realism. It was reducing sizing ambiguity at the moment shoppers decided whether to trust a size and add to bag.
Leading the team meant keeping four disciplines on one definition of success:
Add-to-bag: +61% (A/B test)
New flow against the old flow
Add-to-bag, indexed: 100 → 161 (Old → new flow)
Indexed to the old flow = 100. The same 61% lift, shown as a change, not a share.
Product
Connected confidence to commerce signals
Engineering
Balanced 3D assets with mobile performance
Research
Mapped trust breakdowns
Fit language
Aligned product, research, and merchandising terms
Reflection
I changed my mind about the job of 3D.
The visual was not the product. Confidence was what shoppers believed, understood, and felt ready to do next.
I would cut more today: every control should earn its place at the moment of doubt.
Book a call