Recently, I was consulting with a client in the fashion e-commerce space, I can’t disclose the brand name, but it’s a mid-to-large-scale apparel retailer with a broad online customer base. During our analysis sessions, a major pain point surfaced:

→ Customers loved the styles
→ but weren’t confident about size and fit
and that uncertainty turned into regret and returns.

And that created a measurable business ripple effect:

• high operational costs on returns

• increased customer support load

• reduced customer trust

• and most importantly, lower repeat purchase behavior

The client initially assumed the issue was informational, that customers “just needed clearer size guides.”

But I pushed for a deeper reframing:
this wasn’t a measurement problem, it was an emotional confidence problem.

Understanding the Psychology of “Did I choose the right size?”

I walked them through a behavioral insight we repeatedly see in apparel UX:

When a user guesses their size instead of knowing their size, they subconsciously build an emotional “uncertainty debt” into the purchase.

Meaning:
The moment the item arrives, the first reaction isn’t delight, it’s evaluation, doubt-checking, and comparison:

• “Does this actually fit me?”

• “Should I have picked a size up?”

• “Would that other size have been better?”

This mindset leads to:
→ higher return probability
→ lower product satisfaction
→ lower brand loyalty

So I proposed building a feature that attacks the uncertainty before the purchase even happens.

Solution Proposed: AI-Powered Fit Advisor Feature

I pitched an approach that uses AI not as a “novelty tool”, but as a confidence engine.

Instead of asking the customer to interpret a static size chart, we let the system understand the customer.

We introduced a guided 4-step interaction:

• Fit preference (snug / true / relaxed)

• Current usual size

• Height range

• Body build / proportions

Instead of forcing precise measurements, we used friendly, subjective input.

And behind the scenes, AI mapped the user to predictive size clusters.

The Key UX Philosophy

I explained to the client that:

People don’t think in centimeters, they think in how they feel in clothing.

So the interface didn’t ask:
“How many inches is your chest?”

Instead it asked:
“How do you prefer your sweaters to fit?”

That single shift, from numeric to emotional data, unlocked the entire experience.

The UI Flow I Designed

We created a lightweight modal that activates when the user taps:
“Not sure about your size? Get size guidance.”

Each screen was intentionally:
→ human
→ reassuring
→ judgement-free
→ low-cognitive-effort

The final recommendation screen didn’t just say:
“Your size: L”

It also gave social proof:

“94% of profiles similar to yours are happy with size L.”

That’s AI-driven confirmation, not persuasion, and the difference matters.

AI Explainability

One thing I insisted on during design reviews:
The user shouldn’t feel like:
“the machine picked my size for me.”
They should feel like:
“I understand why this size makes sense.”

So I included a “Why this size?” optional deep-dive revealing factors like:

• preference fit

• fabric stretch

• model patterning

• retention data

• return risk analysis

Even though many users won’t expand it.
the mere presence of explanation increases trust.

Results (summarized due to NDA)

We tracked the impact with real metrics, including return revenue loss reduction, customer confidence indicators, support deflection, and repeat purchase value. Since I can’t disclose the client name, I’ve anonymized the data, but these are actual improvements measured over the post-launch period.

After launching the AI Fit Advisor feature and collecting 6 weeks of post-release data, we observed measurable improvements across key indicators:

• 22.4% reduction in size-related returns

• 31% increase in confidence-based size selection
  (where users selected the AI-recommended size vs manually choosing)

• 18% increase in first-time buyer conversion
  (from product view → add to cart)

• 9.8% increase in repeat purchases within 45 days

• 28% increase in “fit satisfaction” ratings in reviews
  (e.g., “Fits perfectly”, “Exactly as expected”)

• 35% decrease in customer support contacts related to sizing
  (previously: “Does this run small?”, “Should I size up?”)

Additionally, during post-purchase survey sampling:

• 83% of users who used the Fit Advisor said they felt more confident about their choice

• 71% said they prefer AI sizing guidance over static size charts

Next Phase: What UX Problems Identified & What GenAI Can Solve Next

During post-launch evaluation of the Fit Advisor, we identified several opportunities for Phase 2 optimization.

1. The experience resets every time

Users must input preference again on their next visit.

Insight

Returning users shouldn’t repeat themselves.

Phase-2 Solution

Add persistent Fit Profile memory:

• stores preference

• stores successful sizes

• future recommendations become instant

This evolves into:

“We know your fit.”

2. Current recommendation is single-size

We give:
“We recommend L.”

But sizing isn’t binary.

Insight

Some users want nuance.

Phase-2 Solution

Introduce dual recommendation logic:

“L for classic fit, XL for relaxed, oversized fit.”

Giving choice within confidence.

3. Feature currently only lives on the product screen

Limiting discoverability.

Phase-2 Solution

Bring Fit Advisor to:

• checkout page size review

• order confirmation

• “recommended size for you” widget across catalog

• size suggestions during browsing

• personalized product feed:

“Only show items suitable for your body type.”

4. No integration of user-generated photos

Customers often prefer seeing clothing on real bodies.

Phase-2 Solution

Offer:
“See this fit on similar body types from real customers.”

Real-world grounding

5’4” • 145 lbs • relaxed fit → photo
5’8” • 165 lbs • true to size → photo

Final Note

This project reminded the client, and reaffirmed for me, that the best e-commerce UX doesn’t just help users buy, it helps them feel confident in their purchase decisions.

And when you design for confidence instead of conversion, conversion follows naturally.