Effective sales processes are impossible without product data. To ensure that all relevant information for the corresponding product can be found at any time and is available in the correct granularity, a product data model (PDM) is required. Since a product data model grows rapidly with increasing product range size and rising customer requirements, this topic represents a major challenge.
Professionalization and the associated enrichment of product data information in the stationary Commerce are becoming increasingly extensive. For retailers, it is therefore clear that increasing customer requirements cannot be met without investments in this area. Conversely, this means that the advantages of an end-to-end product data model must be exploited today.
In principle, all product-specific information is mapped in the product data model (PDM) according to requirements. This initially facilitates management and automation. The product data shows which products are best described by which attributes. For this purpose, data types and then value lists are used for consistent search filters. This additionally defines what information suppliers should provide. The available data situation also makes it easier to set up comprehensive product data governance and thus automated product data preparation.
The creation and maintenance of a product data model can basically be achieved with two approaches. These two types of creation are determined by two different perspectives: top-down or bottom-up. The top-up approach starts at the top category level and manually refines the primarily coarse model into substructures up to the value list per attribute. However, this approach is associated with a high level of manual effort on the part of competent category or product managers, takes a correspondingly long time, and is naturally often based on gut feeling.
The other approach, namely Bottom-Up takes an opposite starting point for the creation of the product data model. Based on the available product data and categories, product groups and subsequently product families are formed. This involves profiling the attribution of incoming as well as existing product information with examination of, among other things, category-specific frequency, value distribution and fill level of all attributes. The bundling of products with similar attributes from different categories into product families is helpful here to simplify the maintenance of the data model and to prevent inconsistencies. For each category, relevant attributes are also selected for search filters and value lists are stored.
After the definition of data types, value lists and units per attribute, validation and updating is carried out together with experts. The extraction, merging and validation of product data required for this enables a broadly supported, fact-based and rapidly created PDM, which, however, is only feasible with the help of the right technologies.
For the most successful introduction of the new product data model, the continuous migration approach is recommended. This allows validations of the new data model to be performed easily and efficiently. In this way, the practicality of the model can be tested quickly, agilely and iteratively.
