How to govern product information across every channel
Peter Nilsson
The price sits in the business system. The product copy sits in an Excel file with the product manager. The images sit in a folder shared by marketing. The datasheet is a PDF in an email thread from the supplier, and the compatibility list only exists in the head of a salesperson who leaves in the spring.
That is how it looks in companies that have grown through several channels, acquisitions and a business system once built for invoicing. Each part works on its own. The problems start when a customer needs to compare two variants, when a reseller needs an image in the right format or when a service technician needs to find the right seal for a pump from 2019.
Product information is the data that lets a customer, a salesperson or a system choose, compare and order a product. It consists of several layers owned by different people, and the errors arise at the joins between those layers.
What product information covers
Product information is everything that describes a product well enough for it to be chosen, compared and ordered without anyone having to call and ask. In practice, it comes down to six or seven groups of data, and they are rarely owned by the same person.
Identity comes first: item number, EAN and variant code. Without a stable identity, you cannot link a row in the business system to an image in PIM and a row in a spare parts list. Then comes core data such as name, unit, weight and packaging, followed by price and terms, where the list price often differs from the customer-specific price in a portal. Stock and availability are a group of their own, because the stock level changes every hour while the description changes a few times a year.
Enrichment is the part most people associate with product information: copy, attributes, images and user instructions. It includes documents and certificates such as datasheets, declarations and drawings, and the regulatory data on materials, origin and repair instructions. Often, identity and price sit in the business system while copy and images sit in a separate tool, and then there is nowhere the whole picture is reviewed.
The split below is a reasonable starting point for an inventory.
Identity, with item number, EAN and variants, normally comes from the business system and appears in every channel. Core data such as name, unit, weight and packaging also comes from the business system and appears on the website, in the portal and with resellers. Price and terms, meaning list price, customer-specific price and discount, sit in the business system and appear in the customer portal and on the logged-in website. Stock and availability come from the business system or the warehouse system and appear on the website and in the portal.
Enrichment, with copy, attributes, images and user instructions, sits in PIM and appears on the product website and in the channels. Documents and certificates such as datasheets, declarations and drawings also sit in PIM and appear on the product website and in the spare parts portal. Regulatory data on materials, origin and repair instructions sits in PIM or a separate source and is used in the digital product passport when required.
These groups have become a business issue because of how selling happens. Since 2013, Statistics Sweden has tracked companies' sales through digital channels as an official statistical series, measuring orders via their own website or app, via other websites and apps and via electronic data interchange, EDI (Statistics Sweden's database on e-commerce in enterprises). The series covers 2013 to 2024 and also reports e-commerce sales as a share of companies' total turnover.
For EDI, a well-written description is not enough. Product number, price, unit, stock status and order information have to be structured and machine-readable for a transaction to go through. The same data also has to work in a webshop, a customer portal and a reseller's system, and that is where the inventory of source and channel per group becomes useful in practice.
How to model product data so it can be reused
A field called Material looks harmless. In an export from three suppliers, one row says stainless steel, the next says Inox and the third says SS316. To a person, they mean the same thing. To a filter, a comparison table or an export to a spare parts portal, they are three different values, and none of them can be searched.
The difference lies in whether the data is free text or an attribute with a model. Every critical attribute needs a data type, a unit, a code list, a language variant, a source, an owner and a validation rule. Material, repairability and recycling instructions are among those that should be structured values with controlled vocabularies, while marketing copy can vary by language and channel. Modelling product data semantically, with defined data types and shared concepts, is a core technical requirement when the data has to be exchanged between systems and organisations. Standardisation work points towards globally unique identities and standardised carriers such as 2D codes and GS1 Digital Link, and towards access to data being governed on a need-to-know basis.
The image below shows the same problem in its simplest form.

The model works best in three levels: product family, item and variant. Attributes are inherited downwards where it makes sense. A family might be called Pump PX and carry material, pressure rating and media, while the item carries size and the variant carries connection and voltage. You then maintain the material once for the whole family and avoid correcting the same data in forty items when the supplier changes alloy.
The information can also sit in several data sources and still work as a whole, as long as the concepts are shared. That means you do not need to gather everything in one database, but you do need a shared view of what an attribute is and where it comes from.
A practical first step is to choose one product family and list which attributes must be mandatory for your category and which can be optional. That work is described in more detail in our post on master data management.
How data is gathered, enriched and published in a PIM
A PIM works as a flow in four steps. First, data is gathered from the business system and from suppliers' files. Then it is enriched with copy, images and documents. Next, it is validated against the product model, and finally it is published to the channels that will show it.
Each step needs a named owner. Purchasing and whoever is responsible for the business system make sure item data comes in, the product manager owns enrichment, a data steward owns validation and a publishing owner owns the release. Without a name on each step, the PIM becomes yet another file nobody feels responsible for.
Say a family has a thousand items and a new variant is created in the business system. The variant only appears on the website once the mandatory attributes are filled in and validation passes. That is the whole point of the flow: an item missing an image or a datasheet should not be publishable by mistake.

The channels have different needs, which is why a single export rarely works.
The webshop fetches price and stock continuously, while the product website fetches attributes and documents whenever something changes in PIM. The customer portal needs customer-specific prices and documentation linked to what the customer has bought, the reseller portal needs the same data in a different format and the spare parts portal needs compatibility and spare parts lists. Access is controlled per channel: public product data can be shown openly, while customer prices and some documents require login.
Regulatory requirements reinforce this. Digital product passports mean product information is linked to a unique product, model or batch identity and made accessible through a carrier such as a QR code or RFID. The EU's ESPR regulation covers almost all physical products in principle, with gradual introduction until around 2030, according to the Swedish Environmental Protection Agency's FAQ on digital product passports, with product-group requirements set in delegated acts. That makes version control, data provenance, responsible source and validity date per attribute part of the model.
The errors that checks catch before a product is published
Published does not mean correct. Most errors in product data come from copying and manual steps. Systems rarely make mistakes on their own. Someone pastes a price into the customer portal, someone else enriches an item in a tool that is not connected to publishing, and a third person writes centimetres in a field where a colleague writes millimetres.
Three types of check catch most of it. Mandatory fields make sure the image, datasheet and unit are there before the item can be published. Code lists make sure material, surface finish and connection come from the same vocabulary every time. Comparison against the source makes sure the price in the portal is calculated with the same logic as in the business system, without manual entry.
The check has to happen before publishing, because an incorrect specification in a spare parts portal can lead the customer to order the wrong part, with the error only discovered at installation. A principle that goes a long way: an attribute without an owner and a validation rule should not be published at all.
Error rates vary between ranges and channels, so a percentage from someone else's setup says little about yours. Measure your own level with spot checks, for example fifty randomly selected items a month, and count how many could have been published without any issues.
A wrong price in the customer portal is often caused by the price being copied in by hand, and it is caught by comparing against the business system's pricing logic for the customer group. A missing image or datasheet comes from enrichment done in a different tool and is caught with mandatory attributes in PIM. Mixed units in the same category come from free text without a unit and code list, and are caught with a code list and unit per attribute. An old specification left on the website comes from missing version control and is caught with a publishing log showing date and owner. A variant missing from the filter means the attribute is missing from the model, which is caught by settling the filter model before rollout.
Each error belongs to a channel. The price error shows in the customer portal, the missing datasheet shows on the product website and the mixed unit shows in the spare parts portal's filter. That is why the checks need to be linked to the specific channels where the error has consequences.
Digital product passports change what has to be in place
A digital product passport depends on an identity. A QR code on a valve should point to that exact model or batch and open the data relevant to whoever scans it.

The same code can give different views. A customer sees material, warranty terms and installation instructions. A service technician sees an exploded drawing and recommended spare parts. An authority may need to see origin and recycling instructions. That makes access control part of the product model, where every field needs an audience, a level and a validity period.
The standard SS-EN 18222:2026, approved in 2026, covers shared information models, semantic product descriptions, APIs, searchability and access permissions, including IT security, data protection and business confidentiality (SIS). It points to a model where public product data is combined with role-based access to, for example, repair instructions, supplier data, cost information and commercial terms.
For you, that means the data model has to hold audience, permission level, language, validity period and source, as well as attributes and images. Publishing also differs from an ordinary web page: the passport uses stable identifiers and a controlled API layer, so regulatory attributes are not copied by hand to the website, portal and reseller channels. We cover what separates a passport from a product page in the post on digital product passports.
Responsibility, versions and access after launch
The question rarely asked in a PIM project is who owns the data after launch. It becomes urgent when the business system is replaced, when a supplier changes or when a product is discontinued but is still out with customers.
An example. A price changes in the business system in March, a technical specification changes in PIM in April, and in June a customer disputes an order by referring to what the website said when the order was placed. Without a change log and version history, you cannot show which version applied. It works like a receipt: without it, you cannot prove what was actually agreed.
Retention requirements are getting stricter. The standard SS-EN 18221:2026 sets requirements for storage, archiving, history, backup and continued availability of digital product passports, even if the company that created the information ceases to exist (SIS). Product information may therefore need to outlive both systems and organisations.
What needs to be in place is concrete: an owner per attribute, a change log showing who changed what and when, version history for price and specification, traceability to the source and a plan for recovery after an integration or platform upgrade.
A central source does not create governance by itself. Without ownership, a change log and a retention strategy, the PIM can become one central source of errors, and you have just moved the problem to a place where more people see it.
Ongoing management costs money over time, and that line is often missing from the first calculation. 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.
What this approach requires and when it does not fit
A structured approach to product information almost always succeeds or fails on ownership. Every attribute needs a named owner with the authority to stop a publication that is incomplete. Without that person, the model may well be central, but nobody is responsible for keeping it right when a supplier changes or a system is upgraded.
Before import, data quality in the business system and in supplier files needs cleaning up. Duplicate item numbers, old names and units mixed within the same category will otherwise follow into the new model. Every change also needs a decision on who approves it.
With few items, a single channel and no document requirements, a structured product list in the business system is often enough. A PIM is too heavy for an organisation that has not yet decided which channels to prioritise, and then it is better to wait a quarter than to build a model that will have to be redone.
When more than one channel shows the same items, or regulatory requirements are approaching for your product groups, it is time to review your setup.
Then budget ongoing hours per month for attributes, images and documents. That work rarely shows up in a first quote, but it is what keeps the model alive.
Three questions to ask the supplier before you choose
The first question is which attributes are modelled as standard and how long it takes to add a new attribute to an existing category. The answer shows whether the model is built for change or for a single channel. Can a new field be added without development, and does it show up in filters and exports straight away?
The second question is about history: how are price and specification stored over time, and can the supplier show which version applied to an order placed two years ago? The answer shows whether responsibility and retention are solved today or sit in a future release. An answer that starts with "history can be handled with an export from the business system" is an answer about the business system.
The third question is about access: which roles can see which fields, and is the difference between public and logged-in data a setting or a separate development job? The answer shows whether access control is part of the product or something added on later, and it affects both timeline and cost.
The answers are rarely yes or no. It is the detail in them that decides whether the solution can handle the growth of your range.
Take the next step with your product information
Start with one product family or market. In the first week, list the attributes that drive the filters in the category and name an owner for each, for example the product manager for enrichment and finance for pricing data. It sounds simple and is the step that takes longest.
Then define the attribute model for filtering and variants, and connect the data sources through the integration with the business system. A pilot area lets you see which checks catch errors before rolling out to more channels, so you avoid discovering model errors in five channels at once.
Expand the range step by step, and let each step get an approved publishing log before the next one starts. Then there is evidence of what was published, when and by whom, which is the same material you need the day a customer disputes a specification.
On the product information page, we go through how the data can be structured, connected and published, and you can see which steps belong to which function.
Glanser builds and maintains PIM solutions where product information is managed and enriched once and then distributed to websites, customer portals and reseller channels. In a 72-hour analysis, you set aside two hours, and within 72 hours you get an inventory of functions and integrations, a functional specification, a prototype and a complete cost picture with a timeline, free of charge.

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.

Switch e-commerce agency without losing orders or data
How to switch e-commerce agencies without losing data, orders or control.
Kenneth Lyngshede 
ERP and EDI that keep orders and invoices together
ERP EDI integration explained: which messages are included, how to validate data, and which metrics indicate whether the flow is working.
Peter Nilsson 
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