Let’s take the best-case scenario. You’ve invested in a high-performance eCommerce platform. Your User Interface is polished, pages load fast, and the search looks very smart. However, the moment you go live and customers start using the filters, the shopper experience breaks down: visitors select “Cordless” but see corded models; they choose “Under £50” and get… an empty page; they pick “Navy” but half the range is missing.
When this unfortunate worst-case scenario happens, internal teams usually point the finger at the platform or the UX. In most cases though, the real cause is simpler and harder to fix: the structural foundations of your product data can’t support the queries customers are posing.
Filters are the primary way shoppers interact with your catalogue. Faceted navigation doesn’t ‘invent’ insight by magic. It reads attribute values, groups them into options, and narrows the set of results. The filter panel acts as a live diagnostic of your taxonomy, product attributes, variant modelling, and how your data is governed. If these foundations aren’t consistent, all the UI does is faithfully display those inconsistencies.
The impact of broken filters on commercial performance
A broken filter is even worse than no filter. If a customer can’t filter, they might scroll. If they do filter and the results look wrong, they very quickly start to lose trust in your credibility.
It’s reflected in their behaviour:
- They abandon category pages faster (because “nothing matches”)
- They give up on using filters entirely (because they don’t believe them)
- Your high-margin items become hard to find (because the relevant attributes aren’t filter-ready)
- You end up fixing things manually (simply creating more exceptions and drift)
Why filters fail: the structural reasons
1) The normalisation gap
If the same concept appears in multiple formats, the filter treats them as separate choices. For instance:
- “10mm”, or “10 mm”, or “1cm”, or “10 Millimetre”
- “Navy,” or “Navy Blue,” or “Midnight,” or “Dark Blue”
- grams mixed with kilograms; inches mixed with centimetres…and on and on
You’re left with a chaotic facet list (with far too many options) or a range filter that ignores text-heavy values. It isn’t a UI issue, rather, a failure to standardise units, formats, and vocabularies at the point of entry.
2) Attributes and “silent failures” (missing data = missing products)
Filters will only work on what you explicitly tag. If SKUs are missing filterable attributes, they effectively become invisible the moment a customer filters. For example:
- If 40% of drills lack a voltage value, “18V” hides part of the range
- If size is only in the product description, “Size 10” will return incomplete results
- If material isn’t structured, “Cotton” becomes unreliable (or absent)
That’s why ‘completeness’ isn’t the goal. Your real aim is filter-readiness: A state where every attribute used for faceting is populated and validated, per category.
3) Over-generalised taxonomy and schemas
Filters are dependent on context. The word ‘Power’ has different meanings across product types. The risk of using only one generic attribute set everywhere is that you create:
- irrelevant filters (like ‘brand’ for houseplants, for example)
- mixed units (‘volts’ and ‘watts’ in one facet)
- a set of confusing options which don’t help customers make a decision
That’s why strong filtering needs category-level schemas. That is: which attributes matter here? Which ones are required? Which attributes are variant-level, and which need to be controlled vocabularies.
4) Broken variant and relationship modelling
Many catalogues don’t model product relationships consistently: Think of parents, variants, bundles, kits, compatibility, regional versions. In many cases, filters then behave unpredictably because the system can’t ‘see’ the product correctly.
Common patterns resulting in failure include:
- The parent has attributes, but variants don’t (or vice versa)
- variant attributes (size/colour) are stored inconsistently across children
- bundles/kits are mapped as standalone products without the correct inheritances
- compatibility is captured in free text rather than as structured relationships
If relationships aren’t modelled, filters can’t reliably narrow down options for customers.
5) Governance gaps and “filter decay”
Even if you clean data once (however thoroughly!), filter quality degrades without continuous monitoring built into the governance framework. This is especially the case when supplier imports introduce new terms and blanks.
Without adequate controls, you end up with:
- uncontrolled value lists (“navy” multiplies into 12 variants)
- duplicated and obsolete facets
- undocumented category rules
- ad hoc fixes which just create yet more exceptions later
Filters sit at the intersection between UX, merchandising, and product data management. If no one owns the definitions and standards, there’s no-one maintaining the level of quality of the filter experience.
‘Smart search’ won’t save your bacon
AI-powered search is able to interpret intent, but it still basically relies on the same underlying structure to verify and index results. So, if “Waterproof” is recorded as:
- Boolean logic (Y/N) on some products
- “water-resistant” as free text on others
- “10,000mm hydrostatic head” elsewhere
then both your filters and search will misrepresent the range. In a sentence – Better tooling can expose the inconsistency faster; what it can’t do is compensate for that inconsistency.
What “filter-first” alignment looks like
To make filters reliable, remember to treat attributes as navigation coordinates, not just as ‘extra info.’ Genuine alignment means:
- Standardising units and formats at ingestion (not once they’re in the storefront)
- Controlling vocabularies for filterable attributes (avoiding the chaos of free-text entries)
- Defining category-level schemas (relevance, elements required, rules for variants)
- Validating “ready-for-web” criteria before products go live
- Governing supplier onboarding with mapping to established internal standards
The practical fix: a structure audit
A comprehensive structure audit should give you the answers to three questions:
- Are our taxonomy and category rules consistent and channel–aligned?
- Are filterable attributes normalised, controlled, and complete by category?
- Are variants and relationships modelled so the platform can facet correctly?
When these foundations are fully-aligned, your filters stop being a barrier to conversion and provide the fastest route from search intent to product.
Are your filters misrepresenting what you sell? Reach out to us today at Start with Data to arrange a Product Data Management Structure Audit. Because the quickest path to enhancing your commercial performance is aligning structurally across taxonomy, attributes, variants, and product data governance.