Burberry

A £600 scarf shouldn't take three minutes and six seconds to buy — least of all from a brand whose in-store promise is stillness and quiet care. This is how Burberry mobile checkout went from 3:06 to 2:01 — and why every decision was really about how much friction a luxury purchase can survive before it stops feeling like one.

Walk into the Regent Street flagship and an associate meets you at the door, offers a seat and a glass of champagne. Nothing feels like a transaction. That's the point — it's what the brand sells above the leather.

Buy the same scarf on Burberry.com and the checkout sends you sideways for an address, sideways again for a delivery date, sideways again for payment. By the time the pay button appears, the moment has passed.

That line topped every deck. It flips the usual model: in a checkout, every extra beat is a chance to close the tab, doubt, or simply notice how much work they were doing. In luxury, that noticing is fatal — the whole promise is that you shouldn't have to.

Before proposing a change, I walked the whole flow and tallied it. Not “does it feel long?” — how many decisions, inputs and context-shifts. The numbers told the story:

Thirty inputs was a tax the customer paid for the brand's operational convenience. Six redirects meant no one had mapped this end-to-end — ten small, sensible experiences that added up to one senseless one.

That set the bar before I drew a screen: if a change didn't cut sections, questions or redirects, it wasn't a “quick win.”

Most checkout advice is tuned for a price-sensitive, comparison-shopping customer who forgives friction if the price is right. The Burberry customer isn't that: they're paying £600 for a scarf whose functional equivalent costs £30. They're buying the feeling — and friction that interrupts it doesn't just cost conversion, it strains the brand promise.

Four effects were doing quiet damage:

The analytics team was sitting on numbers that justified the project alone. I pulled the shape in numbers first.

Quant told me where customers dropped and how often — not what they were doing when they dropped. So I ran a small, sharp study to answer the why:

One blocked user in thirteen isn't a statistic — it's a P0. The same friction across tens of thousands of sessions is the −25% year-on-year the analytics only showed the surface of. Qual triangulated quant — which is why both exist.

The “quick wins” redesign wasn't a rebrand — it was eight specific decisions, each targeting an effect that was costing completion.

Decision 01 · Visual Hierarchy

One clear focal point per screen. Everything else, quieter.

The old checkout used body copy for every instruction — nothing told the eye where to go first. The redesign restored the type system: one primary heading per screen, supporting text below. The difference between an obvious next step and a guessed one.

Decision 02 · Peak-End Rule

Error messages that don't taste like the peak they'll become.

A generic error is the peak the customer remembers — champagne turning into a call centre. Every message got rewritten to answer two questions: what went wrong, and what do I do now? No codes, no blame — a peak you can recover from without leaving.

Decision 03 · Cognitive Load

Every question earned its place — or was cut.

For each of the 30+ questions I asked: does the business need this to fulfil the order right now? Anything that didn't clear the bar moved to post-purchase — where the customer has already transacted and is happy to share.

Decision 04 · Chunking

Semantic groups that respect the 7±2 rule.

Grouping fields by meaning — name together, address lines together, card number in the shape of a card — turns a form into a sequence of small, obvious commitments. Thirty fields in six chunks of five respects Miller's 7±2; thirty in one block doesn't. Same information, half the felt load.

Decision 05 · Default Bias

Mark the exceptions, not the rule.

The old form asterisked every mandatory field — and when almost everything is mandatory, the asterisks are noise. The redesign flipped it: mandatory is the unmarked default, optional gets called out. Faster to read, and permission not to fill something in.

Decision 06 · Transparency & Trust

If we ask, we say why.

Phone number is the classic: customers resist it, but there's a real reason (SMS delivery updates). The old checkout asked without explaining; the redesign gave a one-line why under the label — and a way to skip it for email. Extractive versus considerate.

Decision 07 · Recognition Over Recall

Show me my address. Don't ask me to remember it.

The billing-address section blocked our test user — empty fields filled from memory, labels subtly different from shipping. The redesign wired in address lookup as the primary input: recognise your address in a dropdown instead of reconstructing it on a phone keyboard. This one change alone would have unblocked her.

Decision 08 · Signifiers

The button has to look like a button.

One screen's primary CTA didn't read as interactive — flat, no elevation. Participants reached the end and stalled, unsure there was a next step. Every primary CTA got a consistent, unmissable treatment. Small change, measurable clearing of the stall.

None of this matters if the checkout only works for perfect vision, full mobility and no assistive tech — especially in luxury, where the customer base skews wider in age and ability. Accessibility is first-order here, not compliance.

So every one of the eight decisions was re-tested through an accessibility lens before shipping: logical tab order, labels readable by screen readers, fieldsets with legends, focus that never jumps under you, an accessible equivalent for every gesture.

Not a lab best-case. An average across 13 luxury shoppers on iPhone X, moderated and remote. Same task, same test cards, same brief.

On causation: the checkout shipped inside a broader web-and-app programme, so the numbers below reflect it landing within that — not an isolated experiment.

The Burberry app won Best Mobile App (Glossy Awards 2017) and Best App (Drapers Digital Festival 2018) — for the app as a whole, not this project, though checkout was a load-bearing part of it.

Beyond the numbers, three things I'm proud of:

Defensible screen-by-screen. Every decision had a behavioural rationale I could put on a slide. Name the effect and you can argue the change.

The billing-address block was fixed as a priority. A single blocked user drove a real change in the shipped flow — the kind of thing a quant-only programme quietly deprioritises.

The playbook was reusable. The same eight decisions informed later flows at Burberry and other brands. Systems outlive projects.

1. I'd tie every design decision to a psychology principle from Day 1.

I wrote the psychology up after making the decisions — backwards. Now I lead with the behavioural frame and let it drive the sketch. Design that starts with a named effect is already halfway defended.

2. I'd design the measurement rig before shipping, not after.

Everything shipped as one bundle, so I can't tell you which decision owns which slice of the +3%. Now I sequence the ship with per-change A/B tests. A bundle that lifts numbers is nice; knowing which part lifted them is repeatable.

3. I'd argue harder for a luxury-specific research recruit.

Thirteen luxury owners was a strong recruit — but it was thirteen. For a funnel with this much revenue on it, I'd run a larger, segmented study (gifters vs. self-buyers, international vs. domestic). More weight in the boardroom, fewer hedges.

4. I'd write about luxury friction as a category, not a checkout.

The real insight isn't about checkout or Burberry: friction has different weight at different price points. Luxury needs its own friction budget — the drag the promise can carry before it erodes brand value, not just conversion. A framework I've taken to other clients since.

Burberry is the project I return to whenever I'm about to accept that a checkout just is the way it is. Friction is never neutral — the price tag depends on what the customer thought they were buying.

Staring at a checkout, an onboarding flow, or anywhere your product promise and friction curve are in tension? My inbox is open.