Lou & Grey

Checkout Redesign & Optimization From 0.9% to 3% conversion, how a structural redesign and iterative testing transformed checkout for a major e-commerce brand

my role: user experience design | experimentation team

As part of a dedicated experimentation team, I served as the sole User Experience Designer supporting Lou & Grey, a women’s apparel brand with a persistent conversion problem. Conversion held steady at a 0.9%, well below industry benchmarks, and the business knew something was broken, but not what or where.

My role was end-to-end: partnering with stakeholders to form hypotheses; designing and prototyping test variants, and interpreting results to inform the next round of experiments. No single test produced the lift; it was the compounding effect of many small, well-targeted changes over time.

The Problem

Lou & Grey's checkout was converting at 0.9%, users were abandoning at every stage, not one catastrophic moment. The challenge was identifying which friction points to address first and in what sequence. Analytics and session recordings told a clear story before we ran a single test.

Four distinct friction patterns emerged:

Checkout Flow Friction
Too many steps between cart and confirmation, users lost momentum.
Formfield Overload
Cognitive load was high before users committed to buying.
Trust and Security
Session recordings showed hesitation at payment fields.
Cart Abandonment
Funnel data showed a drop-off pattern from cart to checkout.

Design Process

Hypothesis-driven, not opinion-driven

Every experiment started with a formal hypothesis before a single design was created. This discipline, shared across the team, kept the program focused on learning, not just shipping variants.

Data synthesis
Combined funnel analytics with session-recording analysis to identify specific drop-off moments and user behaviors. Looked for patterns across sessions, not individual outliers.
Hypothesis formation
Wrote structured hypotheses: "We believe [change] will [outcome] because [data insight]." Partnered with the analyst to prioritize by potential impact and ease of isolation.
Variant design
Designed test variants with one variable changed at a time. Maintained brand fidelity while isolating the specific element being tested, ensuring results were attributable to the change, not noise.
Prototype and QA
Built high-fidelity prototypes for stakeholder alignment, then partnered with developers through implementation to ensure test variants matched design intent exactly; a corrupted variant produces meaningless data.
Results analysis and iteration
Reviewed results with the analyst after each test. Win or loss, the result informed the next hypothesis. Losing tests were often as valuable as winning ones; they ruled out explanations and sharpened the model of what users actually responded to.