Case study · Ecommerce · checkout trust

The checkout worked. The story users told themselves about it did not.

A direct-to-consumer homeware brand, mid-six-figure monthly sessions, checkout conversion in line with category benchmarks, and reviews calling the same checkout “dodgy”, “clunky” and “unsafe”. The engagement was commissioned to scope a redesign. It ended by cancelling one.

Before

3.1@0.74

After · 7 weeks

3.4@0.79

Delta

+0.3

What the measurement said

The Belief surface came back at 4.0 behaviour against 2.3 perception. Checkout completed reliably, payment drop-off was 6% (healthy for the category), and yet perception of the same flow was among the worst readings on the card. Phantom friction, concentrated in the trust signals: payment happened on a plainly-styled third-party page with no brand continuity, no security evidence, and a total that appeared for the first time on the final step.

Rebuilding a flow that completes at 94% has no headroom. The complaint was about the moment of doubt at the card field.

The four surfaces at baseline

SurfaceBehaviourPerceptionQuadrant
Task3.73.5Flow
Commitment3.03.1Watch
Access3.23.3Watch
Belief4.02.3Phantom

The treatment

The planned interaction redesign was cancelled: behaviour said the flow worked. What shipped instead: the payment page restyled to the brand, the total surfaced from the basket onward, security and returns evidence placed at the card-number field, and delivery expectations stated before checkout began instead of after payment. Copy and styling. The flow's steps were not touched, and the cancelled redesign had been budgeted at six times the cost of what shipped.

What did not work

The badge row

A generic trust-badge strip, the padlock-and-logos standard answer, was tried first and moved nothing in a three-week test. Interviews suggested users read it as decoration because it appears on every site, trustworthy or not. What moved perception was specific evidence in context: the returns-policy sentence next to the card field outperformed the entire badge row. Generic reassurance failed; placed evidence worked.

2.3 → 3.5
Perceived trust reading
6% → 5%
Payment drop-off, small as predicted
0
Screens rebuilt

Composite, seven weeks apart

3.1@0.74 3.4@0.79

“Unsafe” and “dodgy” disappeared from new reviews within the window, and users praised “the redesign” when nothing about the interaction had changed. The flow was fine all along; only the story changed.

Next case · Subscription news · loud friction Find your gap