← Back to Portfolio
AI · CRO · Internal Tool

UX Intel

An AI-driven UX and CRO analysis system for 11 e-commerce brands

I designed and built a system that combines GA4, Microsoft Clarity, and Claude to automatically find, rank, and measure conversion fixes, replacing 90–115 hours of manual analysis every month.

RoleLead UI/UX Designer (Design + Build)
StackGA4 · Clarity · Claude
StatusIn use
90–115 hrs
Analyst work replaced monthly
~$15/mo
Running cost
11
Brands covered
At a Glance

The Project in Brief

Problem

CRO analysis was manual, slow, and fixes were rarely measured after shipping.

Impact

Replaced 90–115 analyst hours per month on Custom Offsets alone, worth about $50–80K a year in labor, for ~$15/month.

The Problem

Finding Fixes Took Too Long

Too Much Data, Too Little Time

Finding friction meant manually going through GA4 reports and Clarity session recordings for each brand.

No Clear Priorities

Issues weren't ranked by business impact, so teams couldn't tell what to fix first.

Fixes Weren't Measured

Once a fix shipped, nobody systematically checked whether it actually improved conversion.

How It Works

From Raw Data to a Measured Fix

GA4 + Clarity Data
Automated Data Pull
Claude Analysis
Ranked Fix List
Team Ships Fix
Impact Measured
  1. GA4 + Clarity data. Each brand's GA4 reports and Microsoft Clarity session data are the raw input.
  2. Automated data pull. The system pulls that data automatically instead of someone exporting reports by hand.
  3. Claude analysis. Claude reviews the data to find friction points and recommend fixes.
  4. Ranked fix list. Fixes are ranked by business impact, so teams know what to work on first.
  5. Team ships fix. The team picks fixes from the list and ships them.
  6. Impact measured. After release, the system tracks each fix to see whether it improved conversion.
Design Rationale

Key Design Decisions

Measure Every Shipped Fix

Context

Once a fix shipped, nobody systematically checked whether it actually improved conversion.

Decision

The system tracks each fix after release.

Outcome

Results are proven rather than estimated.

Results & Impact

What It Delivers

90–115 hrs
Analyst work replaced monthly (Custom Offsets)
$50–80K/yr
Labor value saved
~$15/mo
Running cost
  • Covers 11 e-commerce brands
  • Every shipped fix is measured for conversion impact

On a high-volume brand, even a 1% improvement in purchase rate is a major revenue gain, and UX Intel is built to measure exactly that. This section will be updated as measured results come in.

Reflection

What I Learned

Trust Comes From Evidence

Recommendations got adopted when each one came with the data behind it.

Measure, Don't Assume

Closing the loop after shipping is what turns CRO from opinions into results.