Checkout flow redesign

+29% conversion rate achieved with 3 subtle changes

Introduction

I worked as a product designer at Zent Care, a wellness e-commerce. At the time we were struggling with a low conversion rate, despite having good traffic.

Stakeholders couldn't pin down the cause. There were talks about the products themselves, the product pages, brand credibility or even the tone of the copy.

As a product designer I decided to dig deeper and look into the numbers.

E-commerceProduct analysisFlow redesign

Peeling the problem apart

The true challenge is finding the root

01 · The symptom

Low conversion rate

Traffic was healthy and carts were filling up, but the revenue never followed.

I performed a funnel analysis to understand our users' behavior and identify the cause of the problem.

02 · Why 01

High dropout rate

Breaking the funnel down step by step put the drop exactly before the checkout step.

Carts filled at a healthy rate, then almost nobody finished checkout. So it was a flow problem.

03 · Why 02

Checkout struggles

Session recordings showed the same pause over and over. Users wanted to check out and got stuck.

The pause landed in the same spot every time, so I mapped where people were actually clicking.

04 · Why 03

Disabled button

A dense cluster of taps landed on the pay button while it was still disabled and nothing happened.

Almost all of them missed the consent checkbox above it. Grey plus a lock read as locked or broken.

05 · The root cause

No feedback

A consent checkbox held the button disabled and grey. Users didn't understand why, because we never gave them any feedback on their actions.

So most of them assumed the shop was broken and left, with a finished cart one click away.

Key findings

No feedback on action
Problematic checkbox placement

How I got there

Fundamental parts of my process in this project

Funnel analysis

Broke sales down step by step. Carts were filling, but almost no one was reaching checkout.

That gave me a clear view of where the struggle was. Then I needed to identify it.

Session recordings

Recordings showed exactly where users got stuck.

They missed the consent checkbox and couldn't tell why the button didn't work.

Understanding the audience

Many of our shoppers are older and can easily miss small details on screen.

Without feedback, they might never spot the checkbox and never finish the order.

Stakeholders presentation

Walked the team through the funnel, the recordings and our audience.

With the data on the table, I got a yes on the redesign.

What could fix it

Considered options that would solve our problem

Version 1. The consent checkbox flashes red with a notification after a blocked attempt

Pros

  • Users are told exactly what is blocking them after a failed attempt
  • Leaves the flow itself untouched

Cons

  • The button still looks grey and locked until someone taps it
  • The consent checkbox still blocks the way into checkout

What shipped, and why

Simplest solution that eliminates every bit of friction

The shipped checkout. A primary pay button, with consent moved into checkout

Why this solution?

  • The main goal was to remove every possible source of friction before checkout.
  • The pay button became a clear primary button, so it never looks disabled.
  • We removed the lock icon to keep the button simple, even though we hadn't confirmed that assumption with users.
  • Consent moved to checkout, where most e-commerce sites keep it.
  • The checkbox highlights if someone tries to pay without accepting it.

What changed

The results and the learnings

+29%conversion rateThe month after release against the month before.
3×lower abandonment rateAt checkout, over the same two months.

What I'd do differently

I had to ship all three changes at once. Given more room, I'd run them as sequential tests to isolate which one actually drove the result, and interview users to confirm the lock icon was really the problem, not just my assumption.

The lesson

Always trace a metric back to its source. A conversion number means little without the full context around it.

Next caseJuwenalia festival app