Implementing PIM at a manufacturer differs from implementing it in an online store – instead of migrating an existing catalog, you have to build a data model from scratch, because product data is scattered across price lists, marketplaces, and merchant spreadsheets. Using the example of implementing Akeneo's PIM for Astra Polska (a manufacturer of school and office supplies), we show you how this process works step by step: from building the attribute structure and product families, through integration with Comarch XL ERP, to automating attribute mapping and generating descriptions using AI. This article explains the approach, not sells the implementation – it's for those wondering what such a project looks like from the inside.
How does implementing PIM at a manufacturer differ from implementing it in an online store?
When implementing PIM for online stores, the starting point is usually an existing catalog – it needs to be organized, categories mapped, and duplicates cleaned up. For manufacturers just building a coherent omnichannel sales system, this starting point is often simply absent. Product data exists, but it's fragmented: different descriptions in the merchant price list, different ones in marketplaces, different ones (or none at all) on the website. There's no single source of truth that can be imported and cleaned up.
When implementing Akeneo PIM for Astra Polska – a leading manufacturer of school, office, and art supplies – we started from this very spot. Instead of a migration, the project began by building a data model from scratch: we had to decide which attributes actually described a given product before anything could be entered into the system.
How to build a data model from scratch for a manufacturer?
Building a data model from scratch is, in practice, three decisions, made in sequence: which product families to separate (because a school notebook has a different set of features than a pen or a folder), what attributes to assign to each family, and what these attributes are to be used for, because the model is built with two recipients in mind at once.
The first recipient is internal systems – in the case of Astra Polska, integration with Comarch XL ERP, where product data must align with sales and warehouse logic. The second are external recipients: salespeople, marketplaces, and websites – each of them requires a different set and different format of the same data. In the Akeneo PIM module , this distinction is made at the level of the attribute and channel structure, not separate databases – hence the name "data template for internal and external use."
How to automate the catalog population with product data?
Building the model is only half the job – the other half is populating it with data for thousands of product lines, typical for a manufacturer of school and stationery supplies. Manually populating such a large number of SKUs, attribute by attribute, would be a project in itself, so the key element of the implementation for Astra Polska was using AI for three tasks: automatic attribute mapping, generating product descriptions, and detecting inconsistencies in already entered data.
This isn't a "set it and forget it" automation. The first rounds of mapping and description generation always require human review—especially with new product families, where the model is still stabilizing. The value of automation is that this review addresses exceptions and corrections, rather than entering every attribute from scratch.
How does PIM connect to ERP and other manufacturer tools?
Data modeling and automation only make sense if they don't create another data island alongside ERP. In the Astra Polska project, PIM was integrated with Comarch XL so that product data and sales and warehouse logic remained consistent – PIM became a central source of product knowledge, not another system requiring manual synchronization.
An additional element, which arose directly from the manufacturer's multi-channel sales approach, was the functionality of custom exports with public image links – allowing salespeople and partners to access current product materials without having to request files from the marketing department each time. More information about the implementation process itself – from analysis to production launch – is provided in the PIM implementation roadmap for manufacturers and distributors.
What changes when you implement PIM built from scratch?
For a manufacturer starting without a single, consistent source of product data, implementing PIM isn't a clean-up project—it's about building a structure that didn't exist before. The biggest difference from "off-the-shelf" catalog implementations lies in the workflow: first, the model and attribute decisions, then automation of data entry, and only then integration with sales systems. Skipping any of these steps—especially jumping straight to automation without an established model—results in AI efficiently replicating inconsistencies instead of eliminating them.
Before you start such a project, it is worth checking:
- whether product data actually does not have a single source of truth today (price lists, marketplaces, websites are counted separately),
- whether the assortment can be divided into product families with a separate set of attributes,
- whether the same data model should support both internal systems (ERP) and external recipients,
- whether the scale of the catalog justifies the automation of mapping and descriptions, or whether manual completion is sufficient.
If you recognize your situation in the first three points, see what this Akeneo PIM implementation looks like in practice or let's talk about your product catalog.
