Context
An unusual promise, with a £40 catch.
Bother was a grocery delivery startup with a deliberately narrow remit:
Non-perishable household essentials (not fresh produce) delivered to take the boring half of your shopping off your hands…
"We Bother, so that you don't have to," as the motto put it.
It was a genuinely new proposition, and that was the catch: new users couldn't easily tell what we actually sold.
The range was large and spread across many categories, but the value prop was unfamiliar enough that people didn't know what to look for.
And with a £40 minimum on every order, a first-time shopper had to understand the range and move across categories just to reach the point of checking out.
The thinking, then, was that to unlock greater average order values, and thus better overall conversion—
We'd need to design a better solution to achieve two product goals:
Expose the full range to new customers
Make it easier to find what you want
I led that work across web and native (iOS and Android), alongside a product manager, two full-stack engineers, and the commercial team.
But before I could show people the range, I had to work out exactly how the current design kept hiding it.
Problem
Most shoppers only ever saw the aisle they came for.
A quick survey with 25 customers and a design audit I ran surfaced the same issues—
All of which got in the way of product discovery:
1 · The homepage grid dropped customers straight into top-level product listing pages (PLPs)
With no way to drill down, and no exposure to the more specific categories underneath.
2 · On desktop, the sidebar reordered itself every time you switched PLP
So customers could never build a reliable mental model of where things lived.
It saw barely any interaction, and kept most of the category tree out of sight.
Together they added up to the same outcome:
Customers saw the one category they arrived for and little else.
That capped basket size and left first-time baskets stranded below the £40 minimum they needed to check out at all.
The range was all there. The current design just never let anyone see it.
So the question became how to put the whole range in front of shoppers without overwhelming them…
Process
The new menu had to show more, without feeling like more.
I brought the three limitations, and what we were trying to achieve, to engineering, product, and commercial, and ran a solution session that generated 30+ ideas.
We grouped them into three directions, and for each I noted the risks it exposed us to and the assumptions it would have to prove in testing (Does it feel familiar? Does it overwhelm?).
Those directions were:
1 · Horiztonal mega-menu
A common e-commerce pattern that shows the whole range at once.
2 · Sub-category carousels
Replacing top-level homepage carousels with ones a level deeper.
Pushing customers toward search, which converted well, from a new entry point.
I turned all three into prototypes and put them in front of 25 testers, each completing a set navigation task.
Then I asked something the task alone couldn't reveal: 'name everything you think Bother sells'.
That way I could measure whether the navigation was educating people about the range, not just moving them through it.
The result? The mega-menu won.
It had the fastest task completion, and fewest wrong turns.
But the test surfaced a new problem…
It exposed so many categories at once that it risked both overwhelming people and creating a menu too tall to hover away from on smaller desktop screens.
So the real fix wasn't in the menu at all…
I ran a session with commercial to show them the problem, and asked them to consider restructuring the category tree to cut the total number of categories.
I gave them hard constraints: a three-level depth limit, and a cap on sub-categories per top-level category.
While that was underway, I explored four executions of the mega-menu pattern:
B: L1s as tabs in a floating panel, L2s on hover
C: L1s in a persistent second bar, L2s on hover
One more round of usability testing pointed clearly to D — the strongest, and the most consistently recognised.
Mobile needed a different approach.
There, the problem was access, not structure: the cross-category menu buried behind the burger icon already worked well—
It just saw almost no use (8% tap-through), and nothing re-surfaced it once a shopper was inside a listing page.
So instead of testing competing directions, I added an explicitly labelled "Browse Aisles" entry point in-page on every PLP.
Existing mobile menu
Easy to navigate menu hidden behind unlabelled burger icon menu, which got little interaction 8% tap-through.
New menu entry point
Added an explicitly labelled 'Browse Aisles' button in-page on every PLP to facilitate category discovery.
It was low-risk, so we chose speed, building it and validating with an A/B test rather than usability testing first.
Finally, I ran the updated desktop prototype (now populated with the new architecture) through the weekly customer interviews the PM, Lorin Minxhozi, and I were running, watching people complete a real shop and asking questions as they went.
Then I mapped the use cases in Figma, documented the decisions, built the navigation as reusable design-system components, and presented the finished work at a company All-Hands.
With the architecture finally doing the hard work, the menu itself could afford to be simple…
Solution
Everything on offer, always one step away.
On desktop, we shipped an intuitive, familiar mega-menu.
This allowed:
The commercial, marketing, and product teams to feature each of their own high-level menu items
Main L1 categories to be revealed using a hover interaction on 'Browse Aisles'
Sub-categories (L2s & L3s) to appear when an L1 was clicked.
That click-versus-hover distinction was the crucial part: it let customers scan the top-level categories without being buried under everything beneath them.
It's the principle of progressive disclosure: you reveal detail only when someone chooses to go looking for it…
On mobile web and native, the existing menu, already a good interaction, got sharpened rather than replaced:
New icons against the L1 categories, and clearer labelling on the back arrow, so people always knew where "back" would take them…
A new homepage entry point and a persistent category button across every category page, so the menu was discoverable throughout the journey rather than hidden in the top bar…
A context-aware menu that opened to wherever you already sat in the category tree, instead of resetting you to the top every time…
The range was finally all visible. The only question left was whether shoppers would do anything with it…
The Impact
Once they could find it, they bought it.
During the usability testing stage: the new navigation designs produced positive signs:
Customers who strongly agreed they could find what they needed rose from 29% to 37%.
When asked to name what Bother sold, customers on the new design listed 15% more of the range.
And once launched to 20% of users behind a feature-flag, that understanding showed up in customer behaviour:
'Browse Aisles' button triggered:
From within a listing page, rising from 8.1% to 16.3%
Order conversion had:
As new basket value went >£40 went from 0.86% to 2.4%
The work shipped, and the numbers moved.
But three years on, there are a few things I'd do differently…
Learnings
What I'd do differently now in 2026
I'd test less, and trust the pattern more. In Inside the Box, David Epstein makes the case that we spend the most time agonising over the decisions that matter least — precisely because their options look the most alike. A mega-menu is a well-worn e-commerce pattern, and the evidence for it was strong early. I think my newness at the company, and the urge to look confident, had me running more tests than a choice like this really warranted.
I'd build the prototypes in Claude Code. A luxury I didn't have in 2023, but it's where I now save the most time, and, more importantly, where I get the most honest insight. Product designers design software people use, not flat mockups people admire. The faster I can get an interactive prototype in front of someone, the more of their feedback is about what actually matters: how it feels to use.
I'd recognise sooner that the architecture was the design. My job is to find the highest-leverage decision and start there — and here it wasn't the menu, it was getting commercial to restructure the category tree inside a strict depth-and-breadth constraint. The menu only worked once the architecture beneath it did. Unlocking that first would have won the business a quick, visible result — and made the case for everything that followed.
“Always a fresh take - Justin is a pragmatic, critical thinker. There's always a clear process behind his designs and what problems they are trying to solve.”
Henriette Brand