A UNSPSC code tells a procurement system what kind of thing was bought. It tells a customer nothing at all about the product. That distinction is where most UNSPSC work quietly goes wrong. We map UNSPSC for distributors and manufacturers most years, usually alongside ETIM or a merchandising tree. The same question always arrives late. Why can we still not filter anything on the website?
This piece covers the structure, the governance change most teams have missed, the jobs UNSPSC does well, and the jobs it cannot do at all.
What a UNSPSC code is
UNSPSC stands for United Nations Standard Products and Services Code. It is an open, global classification system covering both products and services, owned by the United Nations Development Programme.
The core code is eight digits. Those eight digits carry four levels of hierarchy, two digits per level. An optional fifth level adds two more digits to record a business function, such as rental or retail, rather than the item itself.
The current published codeset is version 26.0801. UNDP makes it available to browse and download at no cost. That is worth knowing if your last exposure to UNSPSC came through a paid subscription feed.
How the UNSPSC hierarchy works
The four levels run from broad to specific: segment, family, class, commodity. Each one narrows the previous one by two digits.
Here is a real worked example from the electrical branch of the codeset:
| Level | Code | Official title |
|---|---|---|
| Segment | 39000000 | Electrical Systems and Lighting and Components and Accessories and Supplies |
| Family | 39120000 | Electrical equipment and components and supplies |
| Class | 39121000 | Power conditioning equipment |
| Commodity | 39121004 | Power supply units |
Read the code left to right and the hierarchy is visible in the digits themselves. That is the part people like. It makes UNSPSC coding easy to roll up. Report at two digits for a board summary, six digits for a category manager, eight digits for a buyer.
The optional business function digits sit on the end. They let a procurement system distinguish buying a power supply unit outright from renting one, using the same commodity code for both. Most organisations we work with never use this level. If you do use it, decide once and document it, because half-populated business function codes are worse than none.
Who owns UNSPSC now, and why the handover matters
UNSPSC was developed jointly by UNDP and Dun and Bradstreet in 1998. Management passed to GS1 US, which ran the codeset for around two decades. GS1 US managed it until 31 December 2024. From 1 January 2025 UNDP took stewardship back, and now handles code change requests, revisions and scheduled updates.
This is not trivia. If your PIM, ERP or spend analytics tool was configured against a GS1 US subscription feed, check where it pulls the codeset from now. We have seen mapping tables sitting on a file share, last refreshed years ago, being treated as current. Codesets that nobody owns internally are the ones that rot fastest.
What UNSPSC is genuinely good at
UNSPSC earns its place in four jobs, and it is very good at all of them.
Spend analysis. This is what it was built for. Roll up spend across divisions, currencies and ERPs. Then answer the question finance actually asks. How much did we spend on cabling last year, across every entity?
Supplier registration and public tendering. Buyers register their capability against UNSPSC codes. Public sector portals across the UK, EU and Australia use it. If you sell into government, you will meet it whether you like it or not.
Punchout and eProcurement catalogues. Ariba, Coupa and similar platforms expect UNSPSC on catalogue lines. Get it wrong and your catalogue fails validation before a buyer ever sees it.
Tail spend visibility. Coding the long, messy tail of low-value purchases to class level is often the fastest route to a rationalisation business case.
Notice what those four have in common. They are all buy-side jobs. None of them is about selling the product.
What UNSPSC cannot express
Here is the part the procurement blogs skip. UNSPSC is a classification of what a thing is. It is not a description of the thing, and it was never designed to be.
It carries no attributes and no units
Commodity code 39121004 says “power supply units”. It does not say 24V DC, 240W output, DIN rail mount, IP20 rated, screw terminal. There is no attribute layer in UNSPSC at all. No properties, no value lists, no units of measure, no tolerances.
This is the single most common surprise on a UNSPSC project. A team codes 80,000 SKUs, ticks the classification box, then discovers the product listing pages still have nothing to filter on. Classification and attribution are two separate pieces of work, and only one of them is solved by a code. If you want faceted search, you need an attribute schema built per node, sitting alongside the code rather than inside it. Our guide to product attributes covers what that schema has to contain.
Standards like ETIM and eCl@ss do carry features and values. GS1’s own Global Product Classification carries brick attributes. UNSPSC carries none of them. That is a design choice, not an oversight.
It has no variant or packaging model
UNSPSC classifies a kind of thing. It does not distinguish a single unit from an inner of ten from an outer of a hundred. There is no pack level, no indicator digit, no case hierarchy.
For a distributor selling the same item as an each, a box and a pallet, that matters. Those three lines share a UNSPSC code and need different GTINs, different weights and different order multiples. The identifier standards handle packaging levels. UNSPSC does not.
The depth is uneven across the codeset
In our mapping work, granularity varies sharply from branch to branch. Some classes carry dozens of distinct commodity codes. Others cover a whole product family in two or three.
When the commodity level does not fit, teams do the obvious thing. They code to the class and leave 00 on the end: 39121000 rather than 39121004. The spend report then looks fully classified when a large slice of it is sitting one level up. If you are running a UNSPSC programme, measure the proportion of lines coded at eight significant digits. That number tells you more about mapping quality than any completeness percentage.
It says nothing about fitment or compatibility
For an automotive parts business, the commercially valuable fact about a brake pad is which vehicles it fits. UNSPSC gives you one commodity code for brake pads and stops. There is no fitment model, no compatibility relationship, no accessory or spare part linkage.
That is what ACES and PIES exist to carry in the automotive aftermarket. We have run this exact conversation with parts distributors who had a clean UNSPSC mapping and still could not answer “will this fit my van”.
The codeset moves and your mapping does not
UNSPSC is versioned, and versions bring additions and revisions. A mapping made against one release drifts against the next.
The fix is cheap if you do it early. Store the codeset version alongside the code, on every record. Then a future migration can tell the difference between a code that is wrong and a code that was right in 2019. Almost nobody does this, and the cost lands years later.
When UNSPSC is the wrong standard
The honest answer is that UNSPSC is usually right for one job and wrong for every other job in the same business. Pick per job, not per company.
| The job | The standard that fits |
|---|---|
| Spend analysis and procurement reporting | UNSPSC |
| Filterable technical attributes for electrical and MRO | ETIM |
| Engineering and process industry property sets | eCl@ss |
| Retail and grocery data pool exchange | GS1 GPC with GDSN |
| Automotive fitment and product content | ACES and PIES |
| Website navigation and merchandising | Your own taxonomy |
The failure mode we see most often is a single standard being asked to do all six rows. It never works. A B2B product classification approach that survives contact with a real catalogue keeps these separate and maps between them deliberately.
How to run a UNSPSC mapping that holds up
Six things, learned the expensive way.
Decide the level you actually need before you start. Class level often satisfies spend analysis entirely. Insisting on commodity level across the whole catalogue can triple the effort for no reporting benefit.
Map from your taxonomy nodes, not from SKUs. Mapping node to code means a few hundred decisions instead of a few hundred thousand. New SKUs inherit the code from the node. This alone is usually the difference between a mapping that finishes and one that stalls.
Handle the exceptions explicitly. Some nodes genuinely contain products that belong to different UNSPSC codes. Flag those nodes, override at SKU level for them only, and leave the rest inheriting.
Record the codeset version on every mapped record. As above. It costs one column.
Keep UNSPSC as a property of the node, never as the tree. The moment UNSPSC becomes your category structure, your merchandising team loses control of the website. We have been called in to unpick that more than once, usually in industrial distribution catalogues where procurement drove the original project.
Sample-audit before you sign it off. Take a hundred SKUs at random, have a category expert check the code, and count the errors. If the error rate is above five per cent, the mapping rules are wrong rather than the mappers.
Key takeaways
- UNSPSC is an eight digit, four level classification: segment, family, class, commodity, with an optional two digit business function level.
- UNDP owns it and resumed stewardship on 1 January 2025 after GS1 US managed it until the end of 2024. The current codeset is version 26.0801 and it is free to download.
- It is strong for spend analysis, supplier registration, punchout catalogues and tail spend visibility. All of those are buy-side jobs.
- It carries no attributes, no units, no variants, no packaging levels and no fitment. Classification and attribution are different problems.
- Map at node level, record the codeset version, and never let UNSPSC become your merchandising taxonomy.
Most businesses need UNSPSC and something else, running side by side against one governed taxonomy and attribution model. Getting that separation right at the start is far cheaper than retrofitting it. Maybe you are mapping a catalogue now. Maybe you have a UNSPSC mapping that is not delivering what procurement expected. Either way, book a thirty minute call and we will tell you plainly whether the approach holds. We do a lot of this work in electrical distribution, where UNSPSC and ETIM usually have to coexist.