Information architecture for 2,000+ products.
Bike Stamford Brook is an independent London bike shop whose e-commerce site relied almost entirely on search, which made a catalogue of more than 2,000 products hard to explore. I led the information architecture and interaction design, translating research into a category and filter model for the website and the shop’s inventory, documented and ready to build. The business chose not to invest in the build, so the work ended at the model.
- Information architecture
- Card sorting
- Competitor benchmarking
- Workshop facilitation
- Prototyping
- Testing
- Client
- Bike Stamford Brook, independent bicycle retailer, London
- My role
- Product Designer, leading IA and interaction design
- Type
- B2C e-commerce
- Team
- 2 UX designers and 2 stakeholders
What came out of it.
- A category and filter model for more than 2,000 products, covering navigation, filter logic and the back-end inventory structure, documented in one source for the business and developers. See Define
- Our own assumption tested and overturned: customers preferred a single route into the catalogue with persistent filters over repeated category entry points. See Develop
- Built from an open card sort weighed against competitor patterns, the shop’s product knowledge and what the back end could support. See Define
- Not taken live: the business chose not to invest in the build, so the work ended at the model. See Outcome
- Constraints
- No meaningful category structure to start from and most pages returning 404s
- Participant recruitment without incentives slowed the research
- The structure had to be one the business could maintain itself and that the back end could support
Customers gave up online and went to the shop.
In the shop, customers got expertise, trust and a sense of community. Online, none of that came through. Navigation was poorly structured, so products were hard to discover; many pages were broken or returned 404 errors; stock visibility was unreliable; and product information gave customers little to go on.
People could not browse with confidence before a visit, so they gave up on the site and went to the shop instead. The challenge was therefore not to tidy the interface. The underlying product structure and navigation had to support a large, complex catalogue.
My contribution
I led the information architecture and interaction design: shaping the category and filter structure, running the research, card sorting and design workshops, and aligning stakeholders on the direction. The model defined the navigation, filter logic and back-end inventory structure proposed for implementation.
Why the site didn’t feel like the shop.
I started by evaluating the existing site and benchmarking other bike retailers, to understand exactly where the online experience fell short of the shop and whether the gap was content or structure.
- Heuristic evaluationThe existing website
- Competitor benchmarkingOther bike retailers’ sites
- Open card sortingGrouping a 2,000+ product catalogue
- WorkshopsStakeholder alignment and exploration
What the audit exposed
- Broken journeys: most pages returned a 404 error.
- Weak navigation: products were not reachable by browsing, so people searched for them on another website.
- Limited interaction feedback: no rollover states or indication of what was selected.
- Brand-led marketing copy rather than the language customers used.
- Nothing of the clean, tidy, precision-based feel of the shop.
Competitors made browsing easy
The retailers I benchmarked consistently supported browsing rather than searching: clear category hierarchies, filters that persisted as people refined results, and product information detailed enough to compare options.
What this meantBike Stamford Brook had plenty of products and plenty of brand content. What it lacked was a structure people could explore. The problem was architecture, not a shortage of content.
Information architecture at scale.
Customers wanted to research online before committing to higher-value purchases in store. That, and the audit findings, set four priorities for the new structure.
- 01Clear, industry-specific categories
- 02Honest, unbiased product information
- 03Accurate stock levels
- 04Research online before buying in store
From card sort to taxonomy
With no meaningful category structure to start from, I ran an open card sort to learn how customers naturally grouped the catalogue.
The sort was a useful starting point, but it could not resolve the catalogue on its own: participants’ groupings worked for familiar items, while many products fitted none of them neatly. So I treated the card-sort results as one input and weighed them against the category patterns competitors had settled on, the shop’s own product knowledge and stakeholder input, deciding case by case where a product belonged and which attributes should become filters rather than categories. The resulting category and filter model was documented in a spreadsheet so the business and the development team could discuss and implement it from a single source.
Turning the structure into a journey
With the taxonomy and filters in place, I mapped the proposed journey: search, shop, wishlist and services as distinct routes that all lead to the basket. Setting them out in one flow made it possible to review the intended paths with stakeholders and developers before anything was built.
Testing our own assumptions.
With the category model taking shape, I explored alternative navigation and product-listing layouts, starting from an assumption I wanted to test: that offering more entry points into the catalogue would make products easier to discover.
Fewer routes, clearer choices
The homepage options set repeated category links against one clear primary navigation; the listing options set filters across the top against filters alongside the products. I explored them with stakeholders, then tested both pairs with customers. Our assumption was wrong: customers preferred a single route into the catalogue, refined by filters that stayed visible while browsing, to repeated category entry points that competed with the main navigation.
The decisionOne primary navigation, supported by side-mounted filters, rather than repeated category routes. Fewer competing paths, a coherent structure for exploring a large catalogue, and a predictable next step at every level.
How the options differed
The two comparisons below show the alternatives as customers tested them. Option B in each pair became the direction: a single primary navigation on the homepage, and side-mounted filters on product listings.
A taxonomy the business could maintain.
The engagement ended with a research-backed category and filter model covering the website’s navigation, its filtering logic and the back-end inventory structure, documented so the business could implement it. Where products had been reachable only through search, it set out a coherent, maintainable way to organise and discover more than 2,000 items. The business chose not to invest in building the site at that point, so the model was not taken live.
Card sorting alone could not structure a catalogue this large; it told me how people think about familiar products, not where the awkward ones belong. The model only held together once I combined it with competitor analysis, the shop’s product knowledge and what the back end could support.
- Recruit participants earlier: without incentives, recruitment slowed the research more than any other factor.
- Check back-end and technical constraints earlier in the IA work, so the structure is designed around them rather than reworked to fit.
Let’s make something work better.
Available for senior UX and interaction design opportunities. See how I work.