Product classification is not filing. It is the structure that decides whether an engineer with a specification in mind finds your part in twenty seconds. Or buys it from someone else. In a catalogue of tens of thousands of SKUs, an inconsistent structure is not untidy. It is lost revenue that never appears in any report, because nobody measures the searches that returned nothing useful.
The practices below hold up across B2B distributors and wholesalers. In these businesses the buyer often knows more about the product than the person who catalogued it.
Start product classification from the buyer, not your internal structure
B2B buyers do not browse. They arrive with a specification, a replacement to match, or a problem that needs solving today.
Classify by function and application, not only by product type. A lighting buyer searches “high bay IP65 emergency”, not a brand. A maintenance engineer searches by the failed part, not by the family it belongs to.
Mirror procurement logic. Look at how customers group items on orders, framework agreements, and bills of materials. Those groupings tell you more about useful category boundaries than any internal convention.
Use customer language. If the trade calls it armoured cable, do not bury it behind an internal label like SWA LV conductors. Both can exist as synonyms, but the customer’s term is the one that surfaces.
The cheapest research available is your own people. Sales, technical support, and key account managers hear how customers describe products every day. An hour with them beats a week of modelling, and it is the first thing we do on taxonomy and attribution work.
Design a taxonomy that survives growth
The job is to get a professional buyer from a broad area to a precise shortlist in very few steps.
Start with broad, obvious families: electrical, fasteners, PPE, HVAC. Add three to five levels of depth where the range justifies it. Electrical, then cables and wiring, then power cable, then armoured power cable. Depth beyond five levels usually means attributes are being used as categories, which is the most common structural fault we find.
Keep the data model mutually exclusive and collectively exhaustive. Every product sits in exactly one place, and nothing has nowhere to go. That is a separate question from what the website shows. Merchandising views, “shop by application” pages, and campaign collections can slice the same products many ways, as long as the underlying classification stays single-valued.
Leave room to grow. New ranges, regions, and own-brand lines arrive constantly. A structure with no space for them gets rebuilt, and rebuilding classification is expensive in ways that are easy to underestimate.
Attributes do more work than categories
In technical ranges, the filters matter more than the category name at the top of the page. A buyer who has reached “armoured power cable” still needs core count, cross-sectional area, and voltage rating before they can choose.
Three attribute groups carry most of the load. Technical values such as pressure rating, torque, thread pitch, lumen output, and temperature range. Standards and approvals including ISO, EN, UL, ATEX, IP ratings, and sector schemes. Regulatory fields such as RoHS, REACH, and country of origin.
Enforce controlled vocabularies and consistent units. Decide once whether the value is A2-70 or A2 70, mm or millimetres, and lock it to a picklist rather than free text. Free text is where filters go to die. One supplier writing “Stainless Steel” and another writing “stainless” produces two filter options for one material. Buyers then assume the range is smaller than it is.
Handle variants, kits, and assemblies with discipline
B2B catalogues are families, not single products. Pumps with motor options. Valves in several bore sizes. Luminaires with different optics and gear.
Use the parent and child structure properly. The parent holds the shared narrative, the media, and the application detail. The children hold only what differentiates them: size, material, rating, pack size. Duplicating the description across forty children guarantees forty versions that drift apart.
Display the family together so a buyer moves from range to exact configuration without jumping between records. Kits and assemblies get their own classification, with the component SKUs listed underneath, because procurement needs to see what sits inside a single line item.
Product classification and industry standards
Most B2B businesses publish into distributor platforms, marketplaces, and customer procurement systems, all of which want their own structure.
Map your internal taxonomy to UNSPSC, eCl@ss, or ETIM where the sector expects it. Hold those external codes in the PIM alongside your own categories, as attributes of the product rather than as a replacement structure.
That distinction matters more than it sounds. Adopting a standard as your internal navigation is a mistake we see regularly. These classifications were designed for procurement and data exchange, not for helping a buyer choose, and their category names rarely match how customers speak. Map to them, publish to them, and keep your own structure for the storefront. The automotive equivalent is worth reading if you sell parts, since ACES and PIES in the automotive aftermarket work the same way.
Remember that categories get consumed downstream. URLs, channel feeds, entitlement rules, and reporting all reference them. Changing a category structure after launch is never just a website change, which is another argument for designing it properly the first time.
Product classification at scale needs more than spreadsheets
Manual classification works for a few hundred products. It does not survive continuous supplier data onboarding at distributor volumes, where new ranges and changed specifications arrive weekly.
A PIM gives you one place to hold the taxonomy, the attribute model, the standards mappings, and the channel formats. Validation rules stop incomplete records going live, so a product without a voltage rating cannot reach a filter that depends on it.
AI-assisted classification now handles a real share of the repetitive decisions, proposing categories and flagging missing technical fields from learned patterns. It does not replace product experts. It moves them onto the edge cases, the new ranges, and the judgement calls, which is where their knowledge is actually worth something. Across our product data services, the automated share keeps rising and the review step does not disappear.
Product classification is a living capability
Structures decay. Regulations change, ranges evolve, and the language customers use moves with them.
Read your own search data. Zero-result searches, unused filters, and terms that return the wrong family are the clearest evidence you will get about where the structure fails. Review navigation and terminology periodically with sales, customer service, and a few key customers.
Fix legacy data in waves, starting with the highest-value and highest-traffic categories. A big-bang reclassification of everything at once stalls, and the categories nobody searches can wait.
Above all, decide who approves a new category and who says no. Without that, categories multiply until the structure means nothing, which is the subject of governing the taxonomy over time.
Where this leaves you
Product classification is the least visible and most consequential work in a B2B catalogue. Done properly it disappears, and buyers simply find things. Done badly it shows up as search abandonment, support calls, and quotes that go elsewhere. We see the same structural faults repeatedly, and have written up the wider set of patterns we see in industrial distribution and in electrical wholesale.
If your catalogue is hard to search, your filters are inconsistent, or classification is still done by hand, book a thirty-minute discovery call. We will talk it through against your structure. Our PIM and PXM services cover the taxonomy work as well as the platform.