Product Data Management Platform: How to Choose One That Fits Your Catalog

Every company that sells physical products has product data. Fewer have a place where it lives. The ERP holds article numbers and prices, a shared drive holds images, and the real specifications sit in spreadsheets that one person fully understands. A product data management platform replaces that arrangement with one governed source for attributes, variants, relations, assets and translations, plus the rules for publishing them to every channel.

This guide covers what such a platform does, what to test before buying, how the EU Digital Product Passport changes the requirements, and where seven established products fit. The short version: start from your most complex products and your most demanding channel, price five years instead of one, and treat the data model as the actual project.

What A Product Data Management Platform Does

One clarification first. In engineering, product data management (PDM) means CAD files, drawing revisions, and bills of materials, usually inside a PLM system. This article covers the commercial side: the product information that sales, marketing, distributors, and marketplaces need. Vendors sell it as PIM, PXM, product MDM, or catalog management. The labels overlap, and the buying questions stay the same.

A platform in this category does five jobs:

• Modeling: attributes with data types and units, classifications that decide which attributes apply to which products, variants with inheritance, and relations such as accessories, spare parts, and successor models.
• Collecting: imports from ERP, supplier files and legacy spreadsheets, with mapping rules that run again next month without rework.
• Enriching: descriptions, translations, images and documents, often with AI drafts, plus completeness scores per channel and language.
• Governing: roles, field-level permissions, approval workflows and change history, so you can see who changed a torque value and when.
• Distributing: channel-specific output to webshops, marketplaces, distributor formats such as BMEcat, GS1 data pools, print catalogs and APIs.

The ERP usually stays the master for article numbers, prices, and stock. The product data management platform becomes the master for everything that describes the product. Drawing that line early saves a surprising number of meetings.

Where Product Data Breaks Without One

The failure rarely looks dramatic. A distributor rejects a catalog file because some products lack an ETIM class. A marketplace keeps showing an old datasheet. A sales rep sends last year's dimensions to a customer. Each incident costs an afternoon, and nobody adds up the afternoons.

Online buyers forgive less than colleagues do. In usability testing by the Baymard Institute, shoppers who could not confirm compatibility for accessories, spare parts, or consumables either left to research elsewhere or abandoned the product. When details were missing, some testers made wrong assumptions about a product, which ended in frustration and avoidable returns. Professional buyers ordering replacement parts face the same doubt, with larger orders at stake.

For compatibility-dependent products, a missing model number is a lost sale. The buyer doesn't guess. The buyer leaves.

Our customers turn to us with a familiar setup. The ERP holds article numbers, prices, and logistics data. Technical attributes live in Excel files maintained by product management, images sit on a shared drive, and every distributor request for an ETIM-classified BMEcat file starts a manual project.

In projects we implemented for manufacturers in this situation, the split of responsibilities came first. The ERP stayed the master for commercial and logistics data. The PIM took over technical attributes, marketing texts, assets, and product relations. ETIM classes were mapped once per product group, and the BMEcat export moved to a schedule. The catalog file stopped being a project and became a job that runs at night.

The Digital Product Passport Raises The Stakes In The EU

For manufacturers selling in the EU, part of the product record is turning into regulated data. The European Commission's Digital Product Passport page sets out the timeline. The DPP Registry has been operational since 20 July 2026. Passports become mandatory on 18 February 2027 for certain batteries, including those for electric vehicles, light means of transport and industrial use. Under the Ecodesign Regulation, delegated acts are planned for iron and steel in 2026, for textiles, tyres and aluminium in 2027, for furniture in 2028 and for mattresses and ICT products in 2029. Each act comes with a transition period of at least 18 months. Construction products, toys and detergents get passports through their own regulations.

The mechanics look simple on paper. The manufacturer gathers the required information, registers the passport in the EU registry, and links the physical product to it through a data carrier such as a QR code. The full data stays with the manufacturer or a DPP service provider. Depending on the product group, it covers safety, origin, materials, repairability, environmental performance, and recycling.

Each of those fields has to come from somewhere structured. A platform that will feed passports needs attributes for materials, substances, and origin, versioning tied to specific product versions, and access to supplier and component data, because much of a passport describes what went into the product. It also needs an export that a DPP service provider can consume without someone copying values by hand.

A Digital Product Passport is a product data project with a legal deadline attached.

How To Evaluate A Product Data Management Platform

Vendor feature lists look nearly identical. Every PIM has attributes, workflows, imports, and an AI button. The differences show up in six areas.

Data Model Depth

Start with your most complex product. For a manufacturer, that might be a pump with dozens of technical attributes, an ETIM class, three voltage variants, a list of compatible spare parts and documentation in five languages. Ask each vendor to model it and watch for specific things.

Variants should inherit values across more than one level, so a change on the parent reaches every child. Category-specific attributes should attach through a classification without changes to the global schema. Relations should carry their own properties, such as the quantity of a spare part in a kit. Units should convert automatically. And the platform should support the classification standard your customers use: ETIM for electrical, HVAC, and plumbing products, ECLASS in industrial supply, GS1 attributes in retail and grocery.

In projects we implemented with manufacturers of configurable products, multi-level inheritance removed most of the duplicate data entry. A change to a housing material happens once, on the parent, and the variants follow.

Channels And Formats

The channels you feed decide which features matter. Consumer brands selling through large retailers deal with hundreds of retailer templates that change without notice. A vendor that maintains those requirements as a service takes that work off your team. B2B manufacturers deal with a different set: BMEcat files for distributors, e-procurement catalogs, print price lists and datasheets. The deciding question there is whether the platform produces those formats natively or through a partner add-on. For your own webshop and apps, check the API for rate limits, webhooks on change events, and bulk endpoints.

Hosting, License And Ownership

SaaS removes upgrade work and ties you to the vendor's roadmap and release schedule. Self-hosting keeps data and release timing in your hands and adds operating work. The "open source" label hides real differences. AtroCore uses GPLv3. Akeneo's Community Edition uses the Open Software License 3.0. Pimcore moved to a source-available license with a revenue threshold in 2025. Read the license with the same care as a SaaS contract, especially if you plan to extend the core or offer the software to your own customers.

Governance And Plan Tiers

Approval workflows, field-level permissions, and change history decide whether the data stays clean after go-live. Many vendors gate these by tier. In Akeneo's packages, collaboration workflows start in Advanced. In Plytix, workflows and user roles start in Pro, and SSO is Enterprise-only. Price the tier you will need in year two, when more departments and suppliers work in the system, rather than the tier that covers the pilot.

Total Cost Over Five Years

License fees are the visible part. The other lines are implementation, integrations, data cleanup, training, and the internal owner's time. Some vendors price integrations as separate items: AtroCore lists ERP integrations from €1,900 and e-commerce integrations from €3,900. Akeneo packages include one activation channel. Plytix add-ons scale with SKU count, so catalog growth raises the bill while the team stays the same size.

In projects we implemented, internal time for data cleanup and attribute definitions was the cost line most often missing from the budget.

AI And Agent Access

AI text generation is now standard. Akeneo, AtroPIM, Inriver, Plytix and Stibo Systems all offer it in some form. The bigger change is agent access through the Model Context Protocol (MCP). Akeneo lists an MCP server in its packages, Inriver extended its MCP server in its Summer 2026 release, and Stibo Systems offers one for governed master data. MCP lets AI assistants and agents query the catalog directly. The data quality rules stay the same: an agent reading an incomplete catalog produces wrong answers faster and with better grammar.

Why These Seven Platforms

There are dozens of PIM vendors. The seven below were chosen because each represents a different answer to the questions above. Akeneo and Inriver represent enterprise SaaS PIM. Salsify represents brand-to-retailer syndication. Stibo Systems represents multidomain master data management that includes product data. Pimcore and AtroPIM represent self-hostable platforms with configurable data models, under very different licenses. Plytix represents small-business SaaS with public prices. All seven are actively developed in 2026, document their pricing or licensing model, and serve manufacturers or brands.

Analyst coverage helped with the shortlist but did not decide it. Gartner published its first Magic Quadrant for Product Information Management Solutions on 12 October 2026, and Akeneo, Inriver, Stibo Systems and Syndigo announced Leader placements. A Leader placement says a vendor executes at enterprise scale. It says little about fit for a manufacturer with a few thousand SKUs and a handful of distributors.

Syndigo was left out because its strengths overlap with Salsify's in consumer goods syndication. Informatica, now part of Salesforce, approaches product data from the MDM side, much like Stibo Systems. Mid-market SaaS tools such as Bluestone PIM, Pimberly, Quable and Sales Layer overlap with Akeneo's Growth package and with Plytix. Any of them can belong on a longlist.

Seven Platforms And Where Each Fits

The platforms appear in alphabetical order. Prices are list prices the vendors published in autumn 2026.

Akeneo

Akeneo sells Akeneo Product Cloud as SaaS in three packages: Growth, Advanced and Premium. The package comparison shows the logic. Collaboration workflows and advanced rights management start in Advanced. PX Insights and 24/7 support come with Premium. Supplier Data Manager, Shared Catalogs, and commerce apps for Salesforce Commerce Cloud, Adobe Commerce, and Shopify are add-ons. Advanced and Premium are quote-based. The Activation module syndicates to more than 500 retailers and marketplaces. A free Community Edition still exists for self-hosting, but Akeneo's feature development and package comparison center on the SaaS product.

Akeneo fits mid-market and enterprise brands, retailers and distributors with many channels and languages, particularly in fashion, sporting goods and consumer products. Implementations usually run through its solution partners. Companies with strict on-premises requirements are left with the self-supported Community Edition, and small teams will find the paid packages heavy for a catalog of a few hundred products.

AtroPIM (AtroCore)

AtroPIM is an open-source PIM under GPLv3, built on the AtroCore data platform. The data model is configured in the interface: new entities, fields and relations without code, more than 20 attribute data types, classifications on any entity, hierarchical variants with inheritance, and channel-specific values on a single product record. Because AtroCore is a general data platform, the same system can hold suppliers, components, documents, or reference data next to products. Exports cover BMEcat, ETIM, ECLASS and GS1 formats, and a PDF generator and an InDesign adapter produce datasheets and catalogs from the same records. AI connectors for Gemini, ChatGPT and Claude draft product texts in place.

The full price list sits on the AtroCore pricing page. The Open-Source Edition is free to self-host. The SaaS Edition starts at €400 per month for hosting and support in a private cloud, and SaaS Premium, which includes every premium module, starts at €1,930 per month. Modules such as ETIM Classification cost €50 per month in SaaS or €1,250 as a one-off on-premise license. Support packages carry usage limits: Bronze covers up to five standard users and 10,000 records, so the "unlimited users" of the license and the users covered by support are two different numbers.

AtroPIM fits manufacturers, wholesalers, and distributors with technical catalogs, especially those that send ETIM-classified BMEcat files to distributors, need on-premise hosting, or want product data managed together with supplier and component data. It fits less well for consumer brands whose main task is syndication to hundreds of retailers with changing requirements, because AtroCore builds channel exports as configured feeds and does not run a retailer network comparable to Salsify's or Akeneo's Activation. A configurable model also needs an owner on the customer side. Without one, it drifts like any other system.

Inriver

Inriver is a Swedish PIM vendor focused on enterprise manufacturers and brands, delivered as SaaS on Microsoft Azure. Its entity model handles relationship-heavy catalogs with components, kits, accessories, and spare parts. The Summer 2026 release extends its MCP server, adds an AI assistant inside enrichment workflows, introduces Signals and Projects for tracking and assigning content gaps, and automates print and digital publishing. Inriver reports more than 1,600 brands as users. Pricing is quote-based.

Inriver fits enterprise manufacturers and brands with large catalogs, many markets and a need for structured content operations across teams. Smaller manufacturers may find the enterprise orientation larger than their problem, and teams that must host on their own infrastructure should clarify deployment options at the start.

Pimcore

Pimcore combines PIM, MDM, DAM and digital experience management in one open-core platform built on PHP and Symfony. Product data lives in data objects defined in a visual class editor, with inheritance, variants, and a classification store for category-specific attributes. Licensing changed with version 2025.1, when the Community Edition moved from GPLv3 to the Pimcore Open Core License (POCL). POCL is source-available and free until an organization's revenue reaches €5 million, and offering Pimcore as SaaS requires a commercial OEM license. Most mid-size manufacturers therefore need a paid Professional or Enterprise edition.

Pimcore fits companies that want product data and website content in one platform and have PHP developers or an implementation partner. Modeling is very flexible, and integrations and custom logic in real projects usually involve PHP development. Teams without technical capacity, or companies above the revenue threshold expecting a free product, should look elsewhere.

Plytix

Plytix is a SaaS PIM for small and mid-size brands, and one of the few vendors with prices on its website. According to Plytix's pricing documentation, Standard starts at €0 per month for 100 SKUs, Pro at €149 and Enterprise at €699, and every plan scales with SKU count. All plans include unlimited users and 500 AI credits a month. Feed management, brand portals, product data sheets, and connectors for Shopify, BigCommerce, and Wayfair are add-ons priced by SKU tier. Workflows, user roles and automatic inheritance start in Pro, API limits range from 2,000 to 25,000 calls per hour by plan, and SSO and a sandbox environment are Enterprise-only. Billing can be monthly.

Plytix fits ecommerce brands, Shopify merchants, and smaller manufacturers that sell through dealers and need brand portals or datasheets more than deep classification. Companies with large technical catalogs, multi-level variant structures, or distributor formats such as ETIM and BMEcat should test those requirements in detail before committing.

Salsify

Salsify sells a product experience management platform to brands and retailers. Its center of gravity is the PXM Network: syndication to retailer endpoints with content validated against each endpoint's attribute requirements, requirement updates maintained by Salsify, enhanced content for rich product pages, and a GS1-certified GDSN data pool. Retailers use the separate supplier experience product to onboard content from their vendors. Industry focus covers consumer goods, food and beverage, fashion and beauty, home goods, and industrials. Pricing is quote-based.

Salsify fits consumer brands that sell through large retailers and marketplaces, where rejected listings and outdated retailer specifications cost real revenue. B2B manufacturers that mainly supply distributors with ETIM-classified data, companies that need self-hosting, and small teams with a few channels would pay for network capacity they don't use.

Stibo Systems STEP

Stibo Systems is a multidomain master data management vendor. Its STEP platform manages product, customer, supplier, and location data in one system, and the company is a privately held subsidiary of the foundation-owned Stibo Software Group. In 2026, Gartner named it a Leader in both the Magic Quadrant for Master Data Management Solutions and the new PIM Magic Quadrant. Its AI work centers on governed access: an MCP server for AI agents and AgentWorkx, a framework for agents that onboard data and generate product content from governed attributes. Pricing is quote-based.

STEP fits large enterprises that need one governance model across product, supplier, and customer data, often in retail, building materials, or regulated industries. Mid-market companies that want a PIM live within a few months are better served by the lighter options above. Plan for an enterprise-scale project with a systems integrator and a data governance function on your side.

Side-By-Side Summary

Running A Selection Without Losing Six Months

Most selections stall in demo loops. Vendors show their best product. You need to see your worst one.

In projects we implemented, the data model workshop took longer than the technical setup almost every time. Agreeing whether "width" means the housing or the mounting plate is slow work, and no software does it for you.

Installing a product data management platform takes weeks. Agreeing on what each attribute means takes longer.

A short, structured proof of concept compresses the decision:

• Pick 50 to 100 products from your most complex category, including variants, accessories, and at least one product with incomplete data.
• Ask each shortlisted vendor to model them in a workshop or a short paid pilot, using your attribute list instead of their demo catalog.
• Run one real export in the format your most demanding channel requires, whether that is a retailer template, a BMEcat file, or a webshop API.
• Calculate five-year cost with real numbers for users, SKUs, languages, channels, environments, integrations, and support level.
• Name the internal owner of the data model before signing.

Then compare the pilots on two measures: how long the model took to build and how many workarounds it needed. Workarounds made in a pilot tend to survive into production, and each one becomes a small permanent cost.