How to structure product data in a PIM data model
Peter Nilsson
The same item has three prices out there right now. The webshop shows one. The reseller portal shows another. The datasheet, untouched for two years, shows a third. The customer asks which one applies, and the answer depends on who opens the file.
The information lives in three systems. Each system has its own version of the truth. Sales fixes the portal. Marketing rewrites the website. Finance lets the business system set the price. The error is back next week, often in a fourth file. A product register that only stores an item number is not enough when the same product has to appear in several channels at once.
Why a product register is no longer enough

PIM is where you manage and enrich product information once. From there it goes on to the website, customer portal and reseller channel. A data model for PIM is the blueprint for what a product consists of. It defines which fields exist, which values are allowed and how products relate to each other.
Think of a library catalogue card. The book is the product. The fields on the card are the attributes. Without fixed fields, two people catalogue the same book in different ways, and the shelf becomes impossible to search.
Three questions guide the rest of the work: which parts the model has, who owns each part and where an overly simple model breaks. A short walkthrough of what PIM is in practice helps if the terms still get mixed up with the business system.
Six building blocks in a PIM data model and who owns them
The model has six building blocks, and each one has an owner who may change its values. Product and variant tie the item number in the business system to the variant relationship in PIM. Attributes, packaging levels, relationships and enrichment are owned by PIM. Price and stock are owned by the business system.
Take an installation cable that an industrial customer sells. The base item holds what is shared, such as brand and conductor area. Black, blue and grey are variants. Each variant gets its own identity, normally a GTIN, so that carton and single unit can be told apart along the chain.
GS1 Sweden describes GTIN as the global identifier for trade items. In a Swedish context, Validoo is the local data pool, and GDSN is the network for synchronised product data. The model should therefore separate product, packaging level, channel and market.
A flat item record puts carton dimensions, consumer dimensions and pallet information in the same field. A measurement that applies to the pallet is then published as the consumer dimension. The error follows into the webshop, portal and datasheet.
Every attribute needs a data type, a unit, a mandatory level, a validity period and a source. Dimensions in millimetres are then read the same way in every channel. Weight in kilograms does not get mixed up with grams. A field without a source cannot be defended when a reseller questions the value.
Relationships come next. The cable points to an accessory, a connector, and to a spare part that replaces an older variant. The link is a structured relationship with its own type. Free text in a comment field cannot be filtered.
Enrichment stays in PIM. Datasheets, installation instructions and sales copy can exist per language and per channel. They should not be written back to the business system, because marketing copy is not a base item.
Price and stock stay in the business system. Customer prices, stock levels and delivery status change through transactions. PIM reads the values, and the business system owns them. The usual boundary is simple. The business system, for example Business Central or Monitor, owns item numbers, prices and stock. PIM owns enrichment, classification and relationships. If two systems can change the same field, you have two truths, so choose an owner before the field is created.
The channels decide which fields have to exist
The same cable needs to appear in four places: your own webshop, the customer portal, the reseller portal and an external data network. The fields are not the same, and a copy per channel creates exactly the three prices you have already seen.
The webshop filters on dimensions, compatibility and area of use. Those values have to be structured attributes. Long sales copy cannot be filtered. Headings and keywords can be channel-specific, but they must not overwrite the name in the core record.
The customer portal needs datasheets, manuals and drawings, often per variant. Some documents should only be visible to certain customer groups. Visibility is a publishing rule linked to the same file. A separate copy per customer group goes out of date as soon as the drawing changes.
The reseller portal needs names, images and specifications per language. The same cable often has different names in different markets. Currency belongs in the pricing logic, together with customer agreements. The product name belongs in PIM.
The external data network requires standardised units and code lists. The recipient should be able to read the record without special rules for your company. If the code is missing, or the weight is in the wrong unit, the record fails the recipient's checks.
Each channel points back to the same source through a publishing rule. The channel gets its own view with its own allowed fields and shares the product record with the other channels.
Run a simple check. Choose three fields per channel and write down who owns each one. If the owner is missing, the field is not ready to publish. The same check applies when you later add a digital product passport, where documents and properties must be traceable to a source.
Packaging unit, minimum order quantity, availability, safety documents and compatibility often matter as much as the sales copy. Those fields belong in the model from the start if your buyers order in packs and need data for their own systems.
How to model attributes and relationships step by step

Start with one product family, for example lighting and cabling, and work through seven steps before you add more items.
First, list the product types and decide which ones need their own attributes. A lamp and a cable each get their own set of fields. The pitfall is a single shared template containing every conceivable field. Most of it stays empty, and people stop filling in what is actually required.
Then define the attributes per product type. Each attribute gets a name, data type, unit, mandatory level and source. Write "weight in kilograms", never just "weight". Otherwise grams and kilograms end up in the same column, and the filter lies. The source is also how you mark ownership field by field: item numbers, core data and pricing logic come from the business system, while copy, specifications, images and documents are enriched in PIM.
Next, gather the shared core attributes for all products. Identity, name, brand and unit of measure go a long way. If the item number is owned by the business system, it still needs to be in the core. The field exists in the model so every channel can show it, but the value flows in through the integration and is never changed by hand. What only applies to the lamp stays on the lamp's type. What only applies to the cable stays there.
The next step is variants and packaging levels as separate records. The colour is a variant with its own identity. The carton of twelve is a packaging level with its own dimensions and quantity. If you put the pallet's height on the single-unit record, you publish the wrong dimensions in every channel.
Then link the relationships. Accessories, replacement products and compatibility are links with a clear type. If compatibility is written as free text, it cannot be filtered, and the spare part cannot be shown for the right variant.
After that, set channel rules per field. Each field knows which channels may publish it, and in which language. Sales copy for the webshop should not leak into the data network, and a safety data sheet should reach the portal even when the sales copy is missing.
Finally, test the model on the product family before adding more. Enter ten real items, including a variant, a carton and a spare part. Adjust fields nobody filled in. Remove fields no channel reads.
The same errors usually show up first in the test: weight without a unit, so grams and kilograms get mixed; compatibility as free text, so the filter cannot be used; and an attribute without an owner, where the value is left over from an old import.
A working model passes a simple test: a new product can be added without anyone having to ask which field applies, who owns it or where it may be published.
Checks that stop incomplete products before publishing

A product in a given GPC class should have the dimensions, weight or material that the class requires. GPC is GS1's way of grouping trade items, so rules can be attached to the type. If the weight is missing, the record is stopped. If a dimension is given in a free-form unit, it is stopped too.
Quality comes from testing the model's rules before publishing. Mandatory attributes, code lists, units and dependencies need to be testable before an item is released, or half an item leaves the building.
GS1 Sweden states that the manual quality assurance in Validoo tests 70 different attributes. That is a benchmark for your own checks. Choose the checks your channels actually read.
The boundary between the systems should be visible in the checks. Item numbers, core data, pricing logic and transaction information come from the business system and should not be rewritten in PIM. Copy, attributes, images and documents are enriched in PIM and can be approved there, but they must never silence a conflicting value in the business system.
Images and technical documents need a language, an area of use and a validity date. A datasheet without a date cannot be trusted after a material change. An image without a language can end up in the wrong market. Documents meant only for partners or logged-in users need a visibility rule.
Traceability belongs in the same layer. Every value should have a data source, a responsible role, a change history, an approval status and a version number. When a rule changes, you can see which items are affected and who needs to review them again.
AI-assisted enrichment needs the same boundary. Keep the source value, the AI suggestion, the approved value and the language version apart. AI may suggest sales copy. Dimensions, weight and item numbers should come from the business system or a standard source. A suggestion becomes a valid value only after human approval, with role and date.
The same gate applies at three levels. Mandatory fields for the product type must be filled in, with the right unit. Code lists should be used, so colour and status are not spelled freely. Documents need a language, a use and a validity period, and dependent fields must be present when a controlling field is set. A field without approved status should not be publishable, however urgent the launch.
When a simple model is enough and when it breaks
A simple model is enough when you have few products, one channel and one language. A spreadsheet with one row per item and fixed columns can carry that load. A PIM project would be a detour there, and the spreadsheet is an honest tool.
The limit is set by how complex the data is. A base item with many variants needs a structure the spreadsheet lacks. Several languages and currencies published in the same week force parallel versions of the same field. When several companies share the range, each one edits "its" copy, and soon nobody knows which one applies. Seasonal peaks expose the rest: the range changes faster than the files can be corrected.
Maintenance has a cost. The model needs a responsible role and a decision every time the attribute list changes. Without that role, the list grows uncontrolled within a few months. New fields are added for a campaign and stay.
PIM does not fix bad core data. If item, price or unit contradict each other in the business system, PIM shows the errors faster, because the same error now reaches every channel at once. Then the business system needs cleaning up first. That order of work is described in our approach to integration with the business system.
Governance also has a running cost. The role, the reviews and the upgrades belong in ongoing management, alongside the licence. Glanser Care is a contracted responsibility at a fixed monthly cost that covers version tracking, compatibility through upgrades, maintenance of the integrations with the business system and bug fixing, so the model does not break at the next version change.
A simpler way of working fits better when a single product family is published to a single channel and nobody else is waiting for the same data. Start there. Bring in PIM once the second channel, the second language or the second company actually exists.
Three questions to ask the supplier about the model
A field list can look complete and still fail in the first channel. Use the questions to test how the model behaves in daily operation.
Start with mandatory attributes per product type: which fields must be filled in before a product may be published, and who approves changes to that list after go-live? A good answer names the type, the fields and the role, and shows a value that must be unambiguous, such as the unit of measure. "We set up the fields in the project" is not enough if nobody owns the list in ongoing management.
Move on to variants and packaging levels. Ask the supplier to separate the GTIN per variant from the item number in the business system, and to show carton and pallet as separate records. Then have them show a validity date on a document. Without the date, you cannot know which version is out with resellers or in the portal.
Put the ownership list next to your channel list and ask the third question from the buyer's side: which rules can we change ourselves when PIM and the business system disagree, without ordering development? Test the answer against a real channel, for example filtering, variant documents and files only partners can see.
Ask the supplier to show an attribute that changed owner within a year. If they can point to the field, the decision and the date, the ongoing management is real. If they answer with a field list without owners, they have delivered a launch, which is different from a model that holds up in ongoing management.
Decide who owns each field before you choose a platform
One field that is often left without an owner is the product sheet. Datasheets, manuals and declarations can come from the supplier, sit in a salesperson's folder or be uploaded by marketing without anyone being responsible for the version. When the filter, the product page and the customer portal all read the same field, every gap in ownership shows up in the channel straight away.
The order is short. List the fields that actually sell, the ones the filter, the portal and the data network read. Choose one product family or market as a pilot. Let the model grow from there, family by family. The same foundation carries the next market once the first one holds. The ownership list is your decision, and it should be settled before procurement.
We introduce PIM as a shared source for product data. The connection runs through the PIM integration and, when needed, the integration with the business system, so items, variants, specifications and documents stay up to date. How that connection works with orders and pricing is covered in e-commerce connected to the business system.
This approach requires your organisation to appoint a responsible role and make the decisions about the attributes. A supplier cannot do that part for you. Without the role, the model becomes a starting point nobody dares to change. Documents meant only for partners or logged-in users also need their own access rule, so ownership covers who may see the content as well as the content itself.
On the PIM page, you can see how we split product data into attributes, channels and owners, so you can bring one product family and see how it should be structured.

About Peter Nilsson
Peter Nilsson writes about integrations and product information, focusing on how business systems, PIM and the web can share the same data without duplicated work.

ERP systems as the hub for your website and portal
Business systems consolidate finance, orders, inventory, and product data into a single source. This is then connected to the web, portals, and EDI.
Peter Nilsson
PIM for wholesalers and who owns the product data
PIM for wholesalers starts with deciding who can change what. How to split ownership between ERP and PIM, choose stable keys and set limits on where AI can enrich product data.
Peter Nilsson
Ordering portals: how they work and what to demand
An ordering portal lets customers see their contract prices and stock levels and place orders themselves. How the order flow reaches the ERP system, when EDI is the better fit and what to demand from your vendor.
Kenneth Lyngshede