Schoeller Allibert is a large manufacturer of Returnable Transit Packaging (RTP), such as plastic crates and pallets. These products can be ordered as-is, or made to order.
In this business-to-business market, it is essential to provide the sales team with reliable information about competing products.
Challenge
To provide insight into the competing products, merely providing the results of a web-scraper is not enough. The information must be made accessible by structuring it in the format of Schoeller Allibert’s Product Information (PIM) system and competing products must be identified.
Scraping a website yields a great deal of raw data so it is important to minimize the amount of required handwork.
Nonetheless, the matches between competing products are often ambiguous because important features (such as “food-safe”) are often missing on the competitor’s website.
Solution
This project was carried out in three phases. Firstly, a webscraper was developed that is suitable for most webshops with minor changes. This scraper was able to determine 20 features of each product, including an image.
In phase 2, the data is converted to the format of Schoeller Allibert’s own PIM system: the feature names are converted, units are removed and values in the form length x width x height are converted to three separate features.
Phase 3 consists of developing an artificial intelligence to identify potential competing products. A similarity measure is determined based on the features of the competitor’s product on one side and Schoeller Allibert’s product on the othe such that products that are similar receive a high score and products that are dissimilar receive a low score.
A product expert determines for each product which of the competitor’s products are similar. A graphical tool has been developed to find these products easily.
Result
The figure above shows the interface to select competing products. Products can be compared and “verified matches” can be highlighted for later use. The features of the products can be compared in more detail by clicking on a match.
All product information of the “verified matches” can be downloaded in Excel format.
Want to know more?
The scraping and matching software is available as a Software as a Service (SAAS) solution that can easily be integrated to an existing PIM system.
For more information or a demonstration, please contact Guus van de Mond: gvdmond@machine-learning.company / +31 (0)629 502018.