Most merchants don’t set out with the aim of complicating life by building a confusing category tree. It just… happens, slowly at first, and then all with gathering speed. One day, a structure that worked perfectly well for 1,000 products becomes a maze with no exit at 50,000. Internal teams are in constant disagreement about where SKUs belong. Suppliers classify inconsistently. And the customers? For them, finding what they want in “three clicks or less.” Is maddeningly out of reach. Every “quick fix” you apply makes the next fix even harder.
This issue is rarely a strategy problem. It’s more a structural one caused by organic growth without adequate governance. Over time, your taxonomy drifts away from how customers shop, how products relate to each other, and what your operational systems (PIM, ERP, search, and now AI) need in order to work reliably. Our article digs deeper into why this happens and what practical steps you can take to address this structural deficit.
The moment your categories stop being a navigation system
A category tree is meant to do two jobs simultaneously:
- Help customers discover and compare (findability and conversion)
- Help the business control product data (classification, schema, rules, automation)
When the tree drifts, it fails you in both ways. At first, symptoms are small:
- Multiple “homes” for the same product (“Smart Home” under Home and Electronics)
- Redundant labels that mean the same thing (“Home Automation,” “Smart Home,” “LED Lighting”)
- A growing “Misc/Other” dumping ground
- Inconsistent attributes (“Size” here, “Dimensions” there) which end up breaking filters and impacting on search
- Deep subcategory tunnels that restrict discovery instead of enabling it
If your teams need to spend more time debating placement than actually listing products, the structure, as opposed to supporting work, is creating a greater volume of it.
Why categories drift in mature catalogues
Taxonomy drift is the predictable outcome of four forces acting together.
1) Your structure mirrors internal logic, not shopper logic
Internal org charts seep into navigation: departments, supplier demands, or campaign needs become permanent branches. But remember, customers don’t shop using your internal constraints as their criteria. They shop by recognition, intent, and comparison.
2) “Just one more category…” becomes a habit
A new range arrives that “doesn’t quite fit,” anywhere, so a new bucket is created. Your latest seasonal collection requires prominent placement, so it becomes a top-level category and then somehow never gets retired. One of your key suppliers wants a ‘shop-in-shop,’ so you dutifully graft on another branch. You could argue that, in isolation, each of these decisions is rational. However, aggregate them and they have a corrosive effect on your performance.
3) You over-categorise (and then wonder why visitors bounce)
Over-categorisation creates a paradoxical ‘agony of indecision.’ Splitting “summer dresses” from “sundresses” might feel highly precise, but it simply forces shoppers into narrow corridors. It also increases the chance they choose the “wrong” corridor and reach the assumption that you don’t stock what they want.
This is where, ironically, so many category trees inadvertently become more restrictive. You haven’t enhanced navigation – you’ve made it more difficult.
4) Markets shift from features to outcomes
Think of fast-moving categories like AI, smart home, health, or sustainability). Potential purchasers are increasingly thinking in terms of outcomes and use cases, not simply product definitions. For instance, “AI-powered CRM” is a feature label; “Sales Automation” is an outcome label. If your taxonomy describes what something is rather than what it does, you’re missing customer intent, and your categories will stop mapping cleanly to how people actually search.
The hidden cost: PIM, search, SEO, and AI tools all depend on structure
Once categories start to drift, the damage isn’t confined to navigation.
- Search and filters degrade because attributes aren’t governed consistently across categories (“filter fragmentation”)
- SEO authority splits when the same concept exists in multiple branches (“duplicate home effect”)
- Analytics results mislead because performance is scattered across overlapping segments
- PIM implementations slow down because you’re unable to clearly define stable category scopes, rules, inheritance, or thresholds for data validation
- AI tools become less accurate because these models need to interpret consistent patterns. If your taxonomy is inconsistent, all automated tasks do is inherit and work off those inconsistencies.
So, a broken category tree isn’t just an operational (and revenue-slashing) pain in the neck. It makes every downstream system fundamentally less reliable.
What’s actually broken: the category tree is doing work filters should do
Many businesses use subcategories to encode attributes like size, material, colour, season, brand, or compatibility, – but that’s how they end up struggling with deep menus and thin category pages.
A more fit-for-purpose pattern in most catalogues is:
- Broader, stable categories for primary grouping (what it is / what it’s for)
- Faceted navigation (filters) for attributes which vary across products (size, finish, brand, compatibility)
This approach considerably reduces the risk of menu sprawl as well as improving discovery and making schema design easier. This is because attribute logic is located where it should be – in structured data, rather than in naming conventions.
What ‘good’ looks like: a deliberate evolution
A scalable taxonomy isn’t ever truly “finished.” It’s governed. And good governance acts when circumstances change. A mindful approach to your taxonomy evolution needs:
- Clear principles
Decide what your taxonomy optimises for. Example: “We categorise by function first, then by audience.” Or: “No use of category names based on internal departments.”
- Defined category boundaries
Each category has a documented scope, examples, and rules for what belongs, and what doesn’t. This will eliminate overlap.
- A lightweight governance process
Someone has ownership of the taxonomy. Changes are proposed, reviewed, approved, and documented. These procedures are not bureaucratic red tape – they are foundational for safeguarding that all-important principle – consistency.
- The rule of logical exhaustion
Take a typical instance: As a footwear retailer, if your structure supports “Men’s Shoes” and “Women’s Shoes,” you must logically have a deliberate plan for “Unisex.” If you don’t, exceptions will quietly eat away at the tree like termites!
- Regular cycles of category tree pruning
Merge redundancies. Retire dead branches. Rename categories to match real, up-to-date search behaviour (For instance, customers may look for “Alexa-type speakers,” not “Internet-enabled smart speakers”). You can’t solve drift once. It needs monitoring and adjusting repeatedly.
- Alignment between taxonomy and schema
Once schema and taxonomy are working off the same song sheet, category decisions will connect directly to attributes, validation rules, and enrichment requirements. These are the circumstances where a PIM solution can genuinely start to deliver value.
A simple 3-way UX test: does your structure make sense to a stranger?
Let’s say you were to hire a new merchandiser tomorrow. Would they be able to classify consistently without having to resort to internal ‘tribal knowledge’? Or, if a supplier joined next week, could they map products without endless messages back-and-forth? Finally, If a customer arrived via organic search, would your category page actively help them to compare, as opposed to landing and wondering where to go next?
If the answer to any of these is “mmm…not really,” the last thing your categories need is yet more exceptions. What they need is a controlled plan for evolution.
If your categories feel bloated, duplicated, or full of “Misc/Other,” it’s high time to stop the drift. Get in touch with us today at Start with Data and we’ll discuss how we can conduct a Product Data Structure Audit for your outfit so you can get a clear map of what’s broken, why it’s happening, and what to fix first.