A PIM data model is the structure that defines which product types (families) exist, which attributes each needs, how variants inherit values, how products relate, and how values vary by channel and locale. A sound model starts from channel and regulatory requirements, uses typed attributes with units, and maps to standards such as ETIM, ECLASS or GS1 GPC.
A PIM data model defines how product information is structured: which product families exist, which attributes each family needs, how variants inherit values, how products relate to each other, and how values differ by channel and locale. A well-designed model makes data entry predictable, validation automatic and syndication simple. Most PIM problems that appear after go-live trace back to model decisions taken in the first weeks.
Which building blocks does a PIM data model contain?
Most PIM platforms use similar concepts, although names differ by vendor:
| Building block | Purpose | Example |
|---|---|---|
| Identifier | Unique key per product | SKU, GTIN (GTIN-8, -12, -13 or -14), manufacturer part number |
| Family (template, product type) | Groups products with the same attribute set | Cordless drills |
| Attribute | One property with a defined type | Battery voltage (number, unit V) |
| Attribute group | Organises attributes for editors and ownership | Technical, Marketing, Logistics, Compliance |
| Variant model | Parent and child levels with inheritance | Model, then colour, then size |
| Category tree | Navigation or classification hierarchy | Web categories, ETIM class, GS1 GPC brick |
| Reference entity | Shared, reusable records | Brands, materials, certifications |
| Relation | Link between products | Accessory, spare part, replacement, bundle |
| Scope | Value variation by channel or locale | Short title for marketplace, German description |
How should attributes be typed?
Attribute typing determines data quality. Practical rules:
- Use number with unit for measurable values (weight 1.8 kg, voltage 18 V), never text such as "approx. 1.8kg".
- Use single or multi-select lists for values used in filters (colour, material, connection type).
- Use boolean only for genuine yes or no facts (contains battery: yes).
- Use reference entities for values with their own data, such as a material with recycled content percentage and certificate.
- Reserve rich text for descriptive copy, and files for documents such as safety data sheets.
- Mark each attribute as localisable, channel-scoped, both or neither.
A worked example for a cordless drill family: identifier SKU and GTIN-13; global attributes for brand, voltage (V), battery capacity (Ah), maximum torque (Nm), weight (kg); localised attributes for title and description; channel-scoped short title (marketplace, 80 characters); compliance attributes for battery chemistry and WEEE category; ETIM class mapping.
How do you design variants and inheritance?
Variant design follows how products differ and how channels display them. A three-step method:
- Identify the axes on which products vary (colour, size, voltage, pack size).
- Decide the levels: for apparel typically model, colour, size; for tools often model and kit configuration.
- Assign every attribute to exactly one level. Images and colour names belong at colour level, GTIN and stock-keeping data at size level, and care instructions at model level.
Limit the model to two variant levels wherever channels allow, since each extra level increases editing and mapping complexity.
How do classification standards fit into the model?
Standard classifications reduce the effort of trading partner exchange. ETIM is widely used in electrical, HVAC and installation sectors, ECLASS (Release 16.0 is current) in industrial and engineering supply, and GS1 GPC in consumer goods data pools. A practical approach maps each internal family to the relevant standard class and links internal attributes to standard features with unit conversion. The details are in ETIM and ECLASS classification, and supplier intake benefits directly, as covered in supplier data onboarding.
How do you build the model step by step?
- Collect output requirements: channel templates, retailer specifications, filters on the webshop, regulatory fields (PPWR packaging, GPSR safety information, DPP attributes).
- Inventory current data: ERP fields, spreadsheets, supplier files.
- Define families from the combined requirements, typically 20 to 200 for a mid-sized assortment (estimate).
- Define shared attributes first, then family-specific ones, with types, units and validation rules.
- Set completeness rules per family and channel.
- Assign ownership per attribute group to a role or source system.
- Test with real data: load 50 products per key family and publish to one channel before freezing the model.
Validation and completeness rules should align with the product data quality framework.
Practical next steps
Before freezing the model, review it with each consuming channel owner: export the test products in the channel's own template and count rejected or empty fields. Zero mandatory-field gaps for the pilot families is a practical exit criterion for the design phase. Document every attribute in a data dictionary with code, label, type, unit, owner, validation rule and example value.
Common mistakes to check against: one universal family for all products, duplicated attributes per family, units in text fields, variant levels copied from ERP without channel logic, and no owner for list values. Model work belongs in the design phase of the PIM implementation roadmap, and its foundations are summarised in what PIM is.
Frequently asked questions
What is a product family in PIM?
A product family, sometimes called a template or product type, is a group of products sharing the same attribute set and completeness rules, such as 'cordless drills' or 'dining chairs'. Families determine which attributes appear and which are mandatory per channel.
How many attributes should a PIM have?
There is no fixed number. Mid-sized B2B models often contain several hundred to a few thousand attributes in total, with 20 to 80 per family. Reuse of shared attributes across families keeps the model maintainable.
Should I use ETIM or ECLASS in my PIM?
The choice depends on the sector and trading partners. ETIM is widely used in electrical, HVAC and plumbing wholesale, while ECLASS is common in industrial and engineering sectors. Many PIMs hold both mappings when customers require them.