CLEAN PRODUCT DATA FOR M-WAY WEBSHOP

The Migros subsidiary m-way has a wide range of products in the e-bike sector with 28 shops and a top-quality advice and repair service. The company, which focuses on e-bikes, would like to extend its reach with a revised web shop.

To achieve this goal, the current webshop software was to be replaced and the product data model was to be refined and expanded with additional filters. Different filters lead the end user to his product faster. Filters are based on existing attributes, which are filled with consistent value lists. The new filters were not sufficiently covered by existing data.

Therefore it was useful to enrich the product data with raw supplier data, to standardize and clean up the different attribute values. This allowed the facet search to be expanded, resulting in a better user experience and higher conversion for the newly set up multilingual web shop.

An m-way internal manual data preparation would have been very extensive due to the partially disordered product data received from various sources. Since the new webshop software was to go online in a timely manner, m-way turned to the data experts at Onedot.

HOW ONEDOT HELPED

In the first step, the existing product data from the ERP and the external supplier data were automatically checked for data quality to identify which attributes are available from the supplier, how structured the product data is and what can be extracted from texts. This is necessary because, for example, a helmet model offers different ventilation holes, visor sizes or closure system adjustment options and these attributes must be classified and correctly assigned.

Based on existing products and raw supplier data, the product data model was defined and product families were formed. Products with the same attributes were combined into product families to simplify the complexity of the product data model. Additionally, attributes per category were defined and value lists, units and data types were defined. The extension of the categorization to further levels, for example from two categories to now four, increased the findability of the products.

In the next step, the product data model had to be filled with product data, so the existing products were enriched with the additional supplier data. The enriched data was then normalized and standardized by the artificial intelligence (AI). As a result, more information is now available in the web shop for the purchase decision, which increases conversion in the m-way web shop.

THE RESULT

Onedot was able to refine the product data model by means of specially developed artificial intelligence (AI) and machine learning (ML) algorithms. On average, up to 20 additional attributes per category were included and over 100 value lists were defined. Subsequently, an efficient product matching, a precise attribute mapping and a comprehensive cleansing were performed. The supplier products were matched with the existing products supplier-specific on the basis of different identification possibilities. Afterwards, over 14,000 supplier attributes from over 20 different suppliers were mapped to the m-way category-specific attributes. The almost 100,000 SKUs in more than 175 categories were adjusted in several iterations according to category. The cleansing included various data types, such as figures, units or value lists. For example, 600 supplier colors were standardized to 16 filter colors. This higher data quality forms the basis for increased conversion and higher sales in m-way webshop.