Home About Contact Book a meeting
KovaAnalytics intelligence ArborData platform PlioAI invoice pipeline FinningDealer portal OnBuyMarketplace UX Design systemSIERA · components & tokens
HOME ABOUT

OnBuy

Checkout and product discovery for one of the UK's largest online marketplaces, streamlining the path from basket to confirmed order across millions of listings.

onbuy / home

Read time
~5 min read
Client
OnBuy
Timeline
2023
Role
Product designer
Team
Designer (me) 2 other designers 1 Product owner Engineering team 2 QA devs

The brief

OnBuy is one of the UK's largest online marketplaces, millions of listings from thousands of sellers, reaching millions of buyers. As the catalogue scaled, the journey from finding a product to confirming an order started to leak. Search returned overwhelming result sets, product pages buried the buyer-protection signals that earn trust on a marketplace, and a multi-step checkout shed conversions on mobile, where most of the traffic actually was.

My role was product discovery and UX: map where buyers hesitated and dropped, then redesign the discovery and basket-to-confirmation flow so the path was scannable, trustworthy and fast on a phone. I worked alongside a product manager and the engineering team, who built against the flows and requirements that came out of the design work.

8M+
Marketplace buyers the experience serves
Millions
Of live listings across thousands of sellers
Mobile-first
Where the majority of the journey happens
Basket → order
The flow this project rebuilt end to end
Problem

A marketplace lives on trust and momentum, and both were leaking between basket and confirmation.

Four problems tangled into one drop-off. None of them was a single broken screen. All of them were about what happens to a hesitant buyer between wanting something and paying for it.

Where it hurt

Checkout was a multi-step flow that asked too much, too early, and surfaced delivery and total cost too late. On mobile, where most buyers were, each extra step was another chance to abandon.

Who felt it

Buyers comparing across millions of listings, and the thousands of sellers who lost a sale every time a full basket was abandoned at the last step.

The trap

A marketplace earns the sale on trust. Buyer protection, returns and delivery promises existed but were buried below the fold, so the exact buyers who needed reassurance never saw it.

The hardest case

Discovery across millions of listings. Results had to stay scannable and filterable without overwhelming a buyer who only knew roughly what they wanted.

A dated seven-step OnBuy checkout with the order total hidden until the final step
The checkout this replaced, seven steps with the total hidden until the end
An OnBuy product page with buyer protection and returns relegated to footer links
Buyer protection lived in the footer, not where buyers hesitated
Research

Find where the journey leaks, and why hesitant buyers bounce before they pay.

A focused discovery loop: an analytics and heuristic review to locate the drop-off, then usability sessions to understand the hesitation behind it.

Q1

Where in the basket-to-confirmation flow do buyers actually drop, and how much of that is concentrated on mobile?

Q2

Which trust signals, buyer protection, returns, delivery, seller rating, change a hesitant buyer's mind, and where do they need to appear?

Q3

How do you make a result set of millions of listings feel scannable instead of infinite?

What the journey told us

Mapping the funnel, the steepest drop sat in the back half of checkout, not in discovery. Buyers were getting to the basket and stalling. In sessions, the pattern was consistent: people wanted the total cost, the delivery date and the safety net visible before they committed, not revealed one step at a time. Uncertainty, not price, was doing the damage.

"I'll add it to the basket, but if I can't see what delivery costs and whether I can send it back, I'll just go and check Amazon instead."Usability participant

That became the design principle for the flow: show the reasons to trust before you ask for the commitment. Cost, delivery and protection move up; steps come down.

01 · Discovery & user research

Search or browse? The data answered.

The product team flagged that a percentage of users were dropping out of checkout. I ran a study on Userbrain, recording real users finding a product and trying to buy it, each leaving a rating and a comment. The first thing it settled: how people actually want to find things.

Userbrain study

They search.

Users with a specific product in mind went straight to the search bar, not the menu. Search was the preferred path, and the comments backed it: "search works well for specific products". So search has to be fast, prominent and forgiving.

They browse.

The opposite held when users had nothing specific in mind. They were happy navigating the menu and leaning on page filters to see everything available. So the menu and taxonomy still carry real weight, which is exactly what the card sort in section 04 is about.

Voice of the users

I don't paraphrase user feedback into something tidier than it was. Here it is verbatim. Read together, the negatives cluster on two things: trust and friction. That clustering is the whole brief.

"Site looks untrustworthy, unsure about security"
trustusability test
"Too many checkout steps"
frictionusability test
"Branding feels dated compared to Amazon"
trustsurvey
"Checkout fields are repetitive"
frictionusability test
"Mobile site navigation is confusing"
navusability test
"Delivery options not clear enough"
claritysurvey
"Hard to find deals or bundle offers"
discoverysurvey
"Search works well for specific products"
keepusability test
"I appreciate the variety of sellers and UK-based options"
keepsurvey
The loudest negative isn't about steps. It's "unsure about security". Users were abandoning as a precaution against identity theft, not out of impatience.
Search and seller variety are already working. Don't break what users came to praise.

Three problems, one root

Trust. Users called the site "dubious" and "blank in places". The fear was looking like a scam, so they bailed.

Friction. Too many steps, repetitive fields, unclear delivery. Each one adds a reason to second-guess the purchase.

Retention. Even after a sale, OnBuy struggled to be remembered. A weak brand leaves no imprint to come back to.

It wasn't the number of clicks. It was the signals.

The easy read was "too many steps, so cut steps". Userbrain said something sharper. Users weren't getting the cues they needed to believe this wasn't a scam: Trustpilot badges, secure-payment gateways, the company logo and a consistent colour scheme. It needed to look like a user hadn't left one site and gone to another just to pay.

My read: fix friction and trust together, but trust leads. A three-step checkout still fails if step two feels fraudulent. So the ranking was clear: trust first, friction second, with search-first journeys for known items and browse-and-filter for discovery.

02 · Personas

Two shoppers, one shared doubt.

The research split cleanly into two behaviours, the buyer who arrives knowing exactly what they want, and the buyer who's browsing for ideas. I built a persona for each. Different journeys, but the same thing makes them close the tab: a flicker of "is this site safe?"

2 personas
PT
Persona 01 · The decided buyer

Priya, the search-first shopper

28–45Mobile-firstTime-poor
Behaviour
Arrives with a specific product and types it straight into search.
Compares a couple of sellers on price and delivery, then commits fast.
Shops on her phone in stolen minutes, so any snag ends the session.
What she wants
Search that finds the right item first time.
A short, predictable checkout with no repeated fields.
Clear delivery options before she commits.
What stops her
Checkout that looks unfamiliar at the payment step.
Repetitive fields that make a quick buy feel like admin.
"If the pay screen looks even slightly off, I'm out. It's not worth the risk for £30."
MC
Persona 02 · The browser

Marcus, the deal-led explorer

22–38Desktop + mobileBargain-hunter
Behaviour
No specific item in mind; happily navigates the menu and filters.
Hunts for deals, bundles and UK-based sellers he can trust.
Will abandon a cart and come back, or not, depending on the impression.
What he wants
A menu and taxonomy that matches how he thinks, not internal org charts.
Easy ways to find deals and compare prices.
A brand he'll actually remember next time.
What stops him
Confusing mobile navigation that hides what he's hunting for.
A site that feels dated next to Amazon, so he doesn't trust the deal.
"If I can't tell it apart from a dodgy site, the discount doesn't matter. I'll just use Amazon."

Where the personas overlap

Priya needs speed: a clean, short checkout that doesn't make her re-enter what she's already given. Marcus needs findability: a menu and filters built around shopper mental models, not OnBuy's internal structure. Both need trust: visible proof it's safe to pay. Without it, neither persona finishes, whatever else you fix. That is the shared design target the rest of the work aims at.

03 · Journey map & drop-off

Mapping exactly where they leave.

To fix abandonment you have to see it stage by stage. I mapped Priya's checkout journey against what she's thinking and where she stalls, then turned the recordings into a drop-off funnel. The biggest fall sits right at payment, the moment trust gets tested.

journey + funnel
Find
Product
Cart
Details
Payment
Action
Searches a specific item, scans results
Checks price, seller and delivery
Adds to bag, starts checkout
Enters details across several steps
Reaches the pay screen
Thinking
🙂 "Search actually works here"
🤔 "Delivery options aren't clear"
🙂 "Right, let's get this done"
😕 "Why is it asking the same thing twice?"
😟 "This looks off. Is my card safe here?"
Friction
Low. The praised path.
Delivery clarity gap
Low
Repetitive fields, too many steps
No trust signals, feels like a new site

Drop-off funnel

Product viewedentered checkout flow
Search → product
100
Added to cartintent confirmed
Cart
82
Details enteredrepetitive fields bite
Details
61
Reached paymentthe trust test
Payment
44
Completedorder placed
Done
31
The steepest single drop is details → payment, where the screen stops looking like OnBuy and starts feeling like a stranger asking for a card.
The shape is the point: the steepest fall is at the payment step, where trust gets tested.

What we changed: friction and trust

Fewer steps, no repeats. Working with the product owners, I cut the number of steps and stripped the "unnecessary" information the old flow demanded. Address and contact details stopped being asked twice. A quick buy started to feel quick again.

Make the pay screen feel like home. Carry the OnBuy logo, colours and Trustpilot badges all the way into payment, with visible secure-gateway cues. The brief in one line: a user should never feel they've left one site and landed on another just to pay.

Discovery

From a leaky funnel to a fast, trustworthy path to order.

Two moves shaped the redesign: make a catalogue of millions of products feel navigable, and win on trust at the exact moments a marketplace is most exposed.

04 · Card sorting & IA

35 million products, one mental model.

Research showed browsers rely on the menu and filters, so the taxonomy had to match how shoppers group things, not how the business does. I ran a card sort on the top-level categories to surface where OnBuy's structure fought the user's. "Mobile navigation is confusing" was an IA problem before it was a UI one.

open card sort

How participants grouped the categories

Participants sorted product cards into groups that made sense to them and named each group. Solid borders are placements people agreed on. Dashed amber cards are the ones that kept jumping groups, the friction points worth redesigning around.

Home & Living
Furniture
Kitchen & Dining
Garden & Outdoor
Home Office ↔ Tech
Tech & Electronics
Phones & Accessories
Computing
TV & Audio
Smart Home ↔ Home
Health & Beauty
Skincare
Fragrance
Vitamins
Electricals ↔ Tech
Family & Leisure
Toys & Games
Sports & Fitness
Pet Supplies
Baby ↔ Health
Four clean clusters emerged. Most products had an obvious home, which means a flatter, shopper-led top level is achievable.
The movers (Smart Home, Home Office, Beauty Electricals, Baby) live between two groups. Solution: let them appear in both, with filters doing the disambiguating.

The old menu problem

Deep, business-shaped categories meant shoppers tunnelled the wrong branch and bounced.
On mobile the depth was worse, hence "navigation is confusing".
Cross-category items had exactly one home, so half the users never found them.

The IA the sort points to

A flatter top level built on the four shopper clusters.
Ambiguous items listed in both sensible places, resolved by filters.
Search stays front and centre for the decided buyer; the menu finally serves the browser.

Why card sort, not opinion

Category trees are where everyone has a view and nobody has evidence. A card sort replaces the loudest opinion in the room with how actual shoppers group products. The movers are the gold, they tell you exactly which items need to live in two places instead of forcing a false single home.

This ties straight back to research: the browser persona lives in the menu. Fix the taxonomy and you fix the navigation complaint at its root. The output: four shopper-led clusters, four cross-category movers, and a flatter, dual-homed IA where filters do the tie-break.

05 · Competitor & trust analysis

Beating Amazon on trust, not scale.

OnBuy can't out-scale Amazon or eBay, and it doesn't need to. But in a market that competitive, looking less trustworthy is fatal, and users said the brand felt dated and dubious. I benchmarked the trust signals the big players use, then turned them into a rebrand brief aimed at one feeling: this is safe, and worth coming back to.

rebrand brief

Trust-signal benchmark

MarketplaceConsistent brand to paymentVisible trust badgesClear delivery infoMemorable identity
OnBuybefore rebrandBreaks at payUnderusedUnclearForgettable
AmazonSeamlessStrongExplicitIconic
eBayConsistentBuyer protectionVaries by sellerEstablished

The trust signals the rebrand had to deliver

Continuity to payment: Logo, colours and layout carry unbroken into the pay screen, so users never feel handed off to a stranger. Proof badges: Trustpilot ratings and secure-gateway marks placed where doubt peaks, at details and payment. A bolder identity - A stronger colour palette and cleaner hierarchy so OnBuy reads as established, not "in its infancy".

Users abandoned as a precaution against identity theft. That's not a friction problem, it's a credibility problem, and a competitive market punishes it fast. After competitor analysis and best-practice research, the output was a new design system: a bolder brand palette, fresh components and clearer hierarchy, all aimed at trust and recognition.

One study set the whole direction

Userbrain found the drop-off. Personas gave it two faces. The journey map and funnel pinned it to the payment step. The card sort fixed the navigation it surfaced. Competitor analysis turned "looks dubious" into a concrete trust brief. Every method answered a question the one before it asked.

My read on OnBuy: the trap was treating abandonment as a step-count problem. It was a trust problem wearing a friction costume. Cut steps and you help. Earn trust and you actually keep the sale.

Product

The screens that carried the buyer from browse to confirmed order.

Discovery, product detail and a condensed checkout, designed mobile-first and built to make trust visible at every step.

Impact

A faster, more trustworthy path from basket to order.

+18%
Mobile checkout completion
−3
Steps removed from the basket-to-confirmation flow
−12%
Basket abandonment on mobile
8M+
Buyers the redesigned experience reaches

The clearest signal was behavioural: moving cost, delivery and buyer protection ahead of the commitment changed where buyers hesitated. The drop-off that had sat in the back half of checkout flattened once the flow stopped asking people to advance a step just to learn what they were agreeing to.

Expected outcomes

↓
Checkout abandonment

Fewer steps plus visible trust signals remove both reasons users were dropping at payment.

↑
Customer retention

A memorable, trusted brand leaves an imprint, so shoppers come back instead of defaulting to Amazon.

Other Work

The rest of the work, from greenfield builds to systemic redesigns.

SIERA Design System
Design system30+ shared componentsThe design system behind the SIERA platform: tokens, components, patterns and the design-to-dev workflow.
Kova
Kova67% of managers off Power BIAnalytics and data-quality intelligence for internal and external stakeholders. Turning raw portfolio data into the answers fund managers were actually asking.
Verde
Verde32% less reconciliation timeA live, governed, traceable ESG reporting workspace. From raw portfolio data to filed disclosure, with a paper trail on every number.
Arbor
Arbor66.7% faster onboardingRebuilding the entity foundation the whole asset platform stands on. From the data model up through onboarding and the interface.
Plio
Plio86% parsing accuracyAI-assisted invoice ingestion with guardrails, not a black box. Confidence-graded and human-in-the-loop across a multi-country portfolio.
Finning UK
Finning UKSelf-serve quoting and financeEnterprise dealer portal and finance calculator for a major Caterpillar equipment dealer. Turning multi-step sales conversations into a self-serve quote and repayment flow.