MRO distributor catalogues break because they combine hundreds of suppliers, very large numbers of technical SKUs and buyers who search by exact specification. When supplier data arrives in different structures, units and levels of completeness, filters fail, search misses and buyers pick up the phone or go elsewhere. The fix is rarely a new search tool. It is MRO product data built on a model.
Why are MRO catalogues harder than most?
Most retail catalogues are wide but shallow. An MRO catalogue is wide and deep. Take an example distributor of maintenance, repair and operations supplies with 80,000 SKUs from 400 suppliers. The range runs across bearings, fasteners, cutting tools, PPE, pneumatics, electrical consumables and cleaning chemicals. Each of those families is described by a different set of specifications. That is the first reason MRO product data is hard: no single template fits it.
The second reason is the long tail. In the catalogues we work on, a minority of lines earns most of the revenue. Buyers still expect the rest to be there and findable. Those slow lines are usually where the data is thinnest, because nobody has looked at them since they were loaded.
Third, MRO buyers search by specification, not by brand story. A maintenance engineer replacing a bearing knows the bore, outside diameter, width and seal type. A site buyer reordering safety gloves knows the cut rating and the size. If those values are not held as structured attributes, the product exists in your catalogue but cannot be found.
Fourth, cross-references and equivalents matter more here than almost anywhere else. Engineers arrive with an old part number, a competitor code or a manufacturer number read off a worn label. Your data has to resolve that to the right product, or to an approved alternative.
Finally, documents travel with the product. Safety data sheets, declarations of conformity, certificates and installation instructions are part of the purchase decision. In many MRO categories, a product without its documents is a product site buyers will not order.
None of this is unique to MRO. It runs through B2B distributor product data generally. MRO just concentrates all of it in one catalogue.
What patterns do we see in distributor catalogues?
We have worked inside a lot of distributor catalogues. The same five patterns show up almost every time, whatever the platform underneath.
Supplier feeds loaded as received
A new supplier sends a spreadsheet. Someone maps the obvious columns (SKU, description, price, pack quantity), loads it and moves on. The other 60 columns sit in a staging table or get dropped. A few years later you have 300 suppliers, each loaded to its own shape. “Thread size” lives in four fields under four different names.
Each load was a rational decision at the time. The catalogue as a whole is not. The root cause is the lack of an intake standard, so every supplier ends up defining your data model for you. Our guide to the supplier onboarding process covers how to set one.
Specifications trapped in PDFs and description text
The values buyers filter on often exist, just not where your site can use them. They sit in a manufacturer datasheet PDF, or inside a description written as one string. Take “M10 x 30 hex head bolt, A2 stainless, fully threaded, pack of 100”. Every value a buyer needs is in that line. None of them is in a field. Your filters show the product under “Bolts” and nowhere else.
Extraction is the fix. Pull the values out of the text and documents into typed attributes, and record where each one came from. This is where AI extraction now earns its keep, as long as a person reviews what it produces before it goes live.
Inconsistent units and value lists
One supplier gives voltage as “24V”. Another writes “24 V DC”. A third writes “24 volts” and puts AC or DC in a separate column. Lengths arrive in millimetres, metres and inches. Colours arrive as “Blk”, “Black” and “BK”.
Each looks harmless. Together they split a filter into fifteen options that mean three things. A buyer who ticks “24 V DC” misses most of the range. The fix is one defined unit per numeric attribute and a controlled value list for every text attribute that drives a filter. Convert on the way in, not on the way out.
Duplicates and broken cross-references
The same product arrives from the manufacturer and from two wholesalers, under three supplier codes. It becomes three SKUs, each with a different description and a different share of the data. Search returns all three, or only the weakest one. Meanwhile the table that maps manufacturer part numbers and superseded codes has not been maintained since the last ERP migration.
Duplicates cost more in MRO because they also split purchasing volume and stock. Matching on manufacturer plus manufacturer part number, then on key attributes, catches most of them. The rest need a person who knows the category.
Categories that grew without a data model
The category tree started life as the chapter list of the print catalogue. New ranges were bolted on wherever they seemed to fit. Now “Workshop” contains spill kits and “Fixings” has a subcategory called “Misc”. Worse, the categories carry no attribute sets. Nobody has defined what a complete product in “Hydraulic hose fittings” looks like. Without that definition you cannot measure completeness, so you cannot improve it.
What do MRO buyers need from your data?
Start from the buyer’s job. They need to find the right part quickly, confirm it fits, check it is in stock and order it through whatever system their employer mandates. Your MRO product data has to support each step.
- Filterable attributes. The specifications they select on: dimensions, thread, material, pressure rating, voltage, IP rating, size. These differ by class, which is why one attribute set for the whole catalogue never works.
- Equivalents and alternatives. Manufacturer part numbers, superseded codes, competitor cross-references and approved substitutes for when the first choice is out of stock.
- Availability and pack data. Stock, lead time, pack quantity, minimum order and unit of sale. Confusing “each” with “box of 100” is a classic source of MRO returns.
- Documents. Datasheets, safety data sheets, certificates and instructions, attached to the right SKU and kept current.
Many MRO customers never browse your website at all. They order from inside their own procurement system. Punchout lets their buyers open your catalogue from that system, pick items and send the basket back for approval. Two protocols come up most in our work.
cXML is a protocol for exchanging business documents between procurement applications and suppliers. It covers purchase orders, catalogues and invoices, and includes a PunchOut feature. SAP’s Open Catalog Interface (OCI) lets a user browse an external catalogue from within SAP and transfer the selected items back.
Either way, the data your customer receives comes from your catalogue. Missing units of sale or wrong pack quantities turn into wrong orders inside their system, not yours.
Larger customers also classify what they buy. Some ask for UNSPSC codes. UNSPSC is an open, global, multi-sector standard for classifying products and services, used for spend analysis and procurement. Others ask for ECLASS (often still written eCl@ss). ECLASS describes itself as the global reference data standard for classifying and unambiguously describing products and services. If those codes are missing or wrong, your products land in the wrong spend category, or get rejected when the customer loads your catalogue.
How should an MRO catalogue be structured?
Use two layers and keep them separate. The first is a customer-facing taxonomy: the navigation tree a buyer clicks through. It should reflect how your customers think about the range, not how your suppliers ship it. The second is the attribute model. For each class it sets the attributes that define it, their data type, unit, allowed values and which are mandatory.
The attribute model is where most catalogues fall short. A class such as “Deep groove ball bearings” needs bore diameter, outside diameter, width and seal or shield type as distinct attributes. Each numeric attribute has one unit. Each text attribute that drives a filter has a controlled list. Completeness then becomes measurable per class, which is what makes improvement possible. Our taxonomy and attribution work starts exactly here.
Where customers and suppliers already use a standard, use it too. ETIM covers technical products used across the electrical, mechanical, heating and installation industries. Each ETIM product class defines a fixed set of features and possible values. ECLASS covers products and services more broadly and supports exchanging product master data between companies. If your suppliers already send ETIM data, mapping your attribute model to ETIM classes saves you rebuilding what they give you.
Many distributors use a standard for data exchange and keep their own taxonomy for navigation. That is usually the right call. Standard classes are built for unambiguous description. They are not always how a buyer expects to browse. The same applies to manufacturers selling MRO lines direct, a pattern we see across industrial and manufacturing catalogues.
Finally, give every attribute one owner. Somebody decides what “Thread size” means, which unit it uses and which values are allowed. Without an owner, the model drifts back to supplier shape within a year. In practice the model and its governance live in a PIM. The PIM is the governed foundation and enrichment is what fills it. Our piece on product catalogue management for distributors shows how those parts fit together.
How do you fix it without stopping the business?
Do not try to fix 80,000 SKUs at once. You will spend a year on a model nobody sees. Work in three moves.
- Prioritise categories by revenue and data quality. Score each category on revenue (or margin) and on current data quality. Start where revenue is high and data is poor. That is where better filters and search pay back fastest.
- Set a supplier onboarding standard. Before you clean what you have, stop new mess arriving. Publish a template per class with mandatory attributes, units and value lists. Validate supplier files against it on arrival and send back what fails. Where a supplier cannot meet it, your team maps their format to the standard once.
- Enrich at scale, with people reviewing the output. For the backlog, use AI extraction to pull attributes from descriptions and datasheets, classify products into the new model and normalise units. Category specialists then review the output before it publishes, starting with the attributes that drive filters. Automation handles the volume. People make the judgement calls.
Then move to the next category and repeat. Each one you finish improves search and filters straight away, so the programme pays its way as it goes.
We followed this pattern for RS Group, a global industrial distributor managing 1.5 million SKUs across 80+ countries. We mapped and enriched product information from thousands of suppliers. We also built workflows for attribute enrichment and classification. The same sequence applies to a catalogue a fraction of that size. Our team delivers it from our centre of excellence in Hyderabad, which has delivered over 50 PIM projects across numerous PIM platforms.
Frequently asked questions
It is the product information a distributor holds for maintenance, repair and operations supplies. That covers technical attributes, identifiers, documents, cross-references and availability.
Buyers search and filter by exact specification. Without consistent attributes and units, the right product exists in the catalogue but cannot be found.
Use what your customers and suppliers exchange. Many distributors use one for data exchange and keep their own taxonomy for navigation.
Once you manage thousands of SKUs from many suppliers across the website, punchout and print, a PIM is the practical way to govern the data.
With the categories that earn the most and have the worst data. Fix the model and the supplier intake there first, then extend.
Get a free Product Data Performance Snapshot, which shows you which data gaps are holding back search and filters. We score 50 of your SKUs and send the results back within five working days.