Automatic product categorisation places every product in the right node of your taxonomy using what is known about it: the title, the description, the supplier's category and any attributes already captured. SKULaunch classifies against your own category tree, not a generic one, and every placement is reviewed before it is approved.
The problem
Category placement drives almost everything downstream. It decides which attributes a product needs, which filters it appears under, which template a supplier gets and where a customer will find it. Yet thousands of SKUs typically sit in "Uncategorised", a catch-all node, or the supplier's own category, which rarely matches yours. Classifying by hand means a person reading each product and choosing from a tree of hundreds of nodes. It is slow, and two people will not make the same choice.
The backlog also hides. Products in the wrong node do not show up as errors. They show up as filters that miss half the range, search results that look thin and attribute gaps nobody can explain.
Before and after
An illustrative example from a builders merchant's mixed supplier file.
Before: how the products arrive
Supplier category "Plumbing Misc" for 340 products, including push-fit elbows, radiator valves, PTFE tape and a pipe cutter.
After: classified into your taxonomy
- Push-fit elbows: Plumbing > Pipe fittings > Push-fit fittings
- Radiator valves: Heating > Radiator accessories > Radiator valves
- PTFE tape: Plumbing > Sealants and tapes
- Pipe cutter: Tools > Plumbing tools > Pipe cutters
- Placements checked in the grid before approval
How it works in SKULaunch
- Load your taxonomy. Import your category tree, or build one in SKULaunch. Classification always works to your structure.
- Import products, classified or not. Use an unclassified import for mixed files, or bring products in from your PIM or ERP.
- Auto-classify. Each product is matched to the most specific end node using its title, description, supplier category and attributes.
- Review and approve. Check placements in the grid, filtered by category, and correct any that need it before the products move on to enrichment.
Why categorisation comes first
In a well-built catalogue, attributes belong to categories. A product in the wrong node gets asked for the wrong attributes, appears under the wrong filters and confuses search. That is why classification should happen before enrichment, not after: once a product sits in the right node, SKULaunch knows exactly which attributes to fill and which values are allowed. Getting the category right is the cheapest quality win in the whole process. See how to give each category its own attributes.
What makes product categorisation accurate
Automatic classification is only as good as the two things it works from: the product data and the taxonomy. Most placement errors trace back to one of them.
On the product side, the title and description carry most of the signal. The supplier's category helps as a hint but should never decide on its own, because "Plumbing Misc" tells you nothing about which of your nodes a product belongs in. Attributes already captured, such as a thread size or a voltage, help separate products that share similar names.
On the taxonomy side, clear node names and clean boundaries matter more than depth. If two nodes overlap, products will split between them and neither a person nor a model can tell you which is right.
Get your taxonomy ready for automatic classification
- Name nodes for what is in them. "Radiator valves" classifies well. "Heating extras" does not.
- Remove overlaps. If "Sealants" and "Tapes and sealants" both exist, merge them or define the line between them.
- Classify to end nodes. A product placed in a parent node gets no specific attribute set.
- Separate products from their spares. A pipe cutter and a replacement cutting wheel belong in different nodes.
- Limit the catch-alls. An "Other" node attracts everything the classifier is unsure of, and so does a person in a hurry.
- Keep channel categories separate. Your master taxonomy and a marketplace's category list are different things.
By hand versus at scale
By hand, a person reads each product, opens the category tree and picks a node. That works for a few dozen new lines a week. It does not work for a 340-line supplier file, a migration or a taxonomy redesign. At scale the job changes from choosing to checking: products are placed automatically, and the team reviews placements in the grid, category by category, correcting the few that need it. Reviewing by category makes odd ones out easy to spot, because a pipe cutter stands out in a list of fittings.
What changes
Mixed supplier files stop being a manual sorting job. Products land in the right node on import, the right attributes are requested, and the "Uncategorised" bucket stops growing.
Common questions
How does automatic product categorisation work?
Each product is matched to the most specific node in your taxonomy using its title, description, supplier category and attributes. Placements are then reviewed and approved in the grid.
Does it work with our own taxonomy?
Yes. Classification is always against the taxonomy you load, at whatever depth you use.
Can it reclassify an existing catalogue?
Yes. If you redesign your category tree, existing products can be reclassified into the new structure in bulk.
What about products that could sit in two categories?
You decide the primary node, which drives the attribute set. Channel-specific categories can be handled separately on export.
What is the difference between product categorisation and ETIM or UNSPSC classification?
Categorisation places products in your own taxonomy for browsing and attributes. ETIM and UNSPSC are external standards your customers ask for. SKULaunch holds them as separate classifications on the same product.
Can we categorise products before we have any attributes?
Yes. Title and description are usually enough to place a product. Attributes come afterwards, once the node tells you which ones are needed.
More on product taxonomy, the taxonomy and schema builder, and how to import products before you know their category.

.avif)