Episode 2: Why Your Category Structure Is Quietly Hurting Ecommerce Performance

‍

Category structure is one of those topics that rarely gets the attention it deserves. It's not flashy. It doesn't show up in a UX presentation or a marketing campaign. But get it wrong, and it quietly undermines almost everything else you're trying to do — from product discovery to team efficiency to AI search performance.

We see the same pattern repeatedly across clients in building supplies, retail, and beyond: businesses that have skipped the hard work of categorising at a granular level and are now paying the price. Here's what that actually looks like in practice — and what you can do about it.

The Real Purpose of Category Structure

Most teams think of category structures as a way to organise products. And yes, that's part of it. But the more important function is this: your category structure determines what data fields your products get.

That single decision flows downstream into everything — which attributes get collected, what your filters show, how well search performs, whether comparison tables work, and how easy it is for your team to enrich product data at scale. Category structure isn't a taxonomy exercise. It's a data architecture decision.

What Goes Wrong: A Real Example

A building supplies business came to us mid-replatform and PIM project. When they started loading data, they hit a wall: their entire bathrooms range — showers, taps, baths, heated towel rails, basins, the lot — had been dumped into a single "Bathrooms" category.

The result? 140 attributes in that one category, covering every possible product type underneath it. The operational impact alone was significant: team members were scrolling through 120+ attributes trying to work out which ones applied to a basin versus a shower versus a heated towel rail, with no mandatory fields to guide them. Fill rates were low because most attributes simply didn't apply to most products.

But the customer impact was just as damaging. Filters returned poor results. Search — both traditional and AI-driven — struggled to surface the right products because there wasn't enough context or specificity in the structure. A customer looking for a heated towel rail in a specific size and finish essentially couldn't find it.

The Other Extreme: Going Too Granular

It's also possible to swing too far the other way. We've seen data sets where individual categories were things like "corded hoover" and "uncorded hoover." That's not a category — that's an attribute value. Over-categorisation creates its own problems: fragmented data, maintenance overhead, and categories so narrow they stop being useful.

The goal is the right level of specificity — granular enough that the attributes assigned to a category are relevant and consistent for the products within it, but not so granular that you're essentially creating a new category for every product variant.

How to Fix It: Work Backwards From What You Have

If you're sitting on a bloated top-level category with hundreds of products and dozens of mismatched attributes, the approach is to reverse-classify what's there.

Start by reviewing everything in that category. Group products by type — not by how they were originally organised, but by what they actually are. For a bathrooms category, you might identify eight to ten logical next-level categories. Some of those will then have their own subcategories, depending on how different the attribute needs are at that lower level.

From there, for each subcategory, define which attributes are mandatory — the fields that every product in that category must have — and which are optional. Don't try to cover every conceivable spec. For a wash basin, there might be 40 possible attributes, but realistically 10 to 15 are genuinely important for helping a customer make a buying decision. Think about what falls into three buckets: regulatory requirements, buying decision information, and logistical data (dimensions, materials, compatibility).

Once the structure is right, data enrichment becomes dramatically faster — both for your team and for AI-assisted workflows. It's the difference between asking someone to search through 140 fields and giving them 15 clearly relevant ones to complete.

The Customer Experience Impact

When category structure is done well, the effects on the front end are clear and immediate:

• Filters work properly — returning real, relevant values that help customers narrow down their choices
• Search improves — both traditional keyword search and AI-powered contextual search. A customer searching for "women's waterproof hiking boot" needs that product to live in a subcategory that reflects that specificity, not just under "Footwear"
• Product detail pages have the right specs — because the right attributes were defined and collected
• Comparison tables function correctly — which only works when attributes are defined at a consistent, low level

The inverse is also true: a product buried in an overly broad category, with patchy attribute data, is effectively invisible — to human shoppers and to AI search alike.

When to Tackle This

The honest answer: before you do anything else.

If you're about to start a PIM implementation or an e-commerce replatform and you haven't done the taxonomy, classification, and attribute schema work first — stop. Do that work before you build anything. Skipping it means you'll be unpicking it later, almost certainly at double the cost, once everything is live and changes are harder to make.

If you're not in the middle of a major project, the good news is you can do this incrementally. Pick one category, audit it, restructure it, define the mandatory attributes, and start enriching. Roll it out category by category. The results are visible quickly — better filters, better search, a noticeably improved buying experience — which also makes it much easier to make the case internally for doing the rest.

A useful exercise: ask your leadership team to go and find a product on your own website as if they were a customer. It's a surprisingly effective way of surfacing just how broken the experience can be — and of getting the investment needed to fix it.

Three Things to Take Away

1. Category level determines product quality. The specificity of your category structure controls which attributes get assigned, which controls the quality of everything downstream.
2. Attributes drive discoverability. Well-defined, mandatory attributes at the right category level means filters work, search performs, and customers can actually find and evaluate what they're looking for.
3. You don't have to do it all at once. If you're doing a replatform or PIM project, yes — do it all upfront. But if not, a category-by-category approach works. Start somewhere, see the results, and build from there.

‍

‍