Fully automated SEO optimized product categorization based on artificial intelligence for a beauty portal

Our successes

Our AI-based system fully automatically assigns all customer products delivered via data feed to 160 SEO optimized product categories
Product categorization accuracy increased to over 96% in a very short time using artificial intelligence
The duration of a product assignment has been minimized to only 0.000035 seconds - a well-trained person needs 35 seconds for that
Customer-facing personnel expenses for the product categorization area were reduced from three employees to one position

About our client

Our client operates a beauty portal with an enormous product assortment of over one million SKUs, which are provided by various retail partners in terms of a marketplace logic.

Our customer came to us with the following questions:

Produktkategorisierung für ein führendes Beauty (8)
Produktkategorisierung für ein führendes Beauty (6)
Produktkategorisierung für ein führendes Beauty (2)
Produktkategorisierung für ein führendes Beauty (4)

IT services based on artificial intelligence

Challenges

Categorizing products by humans typically requires hundreds of man-hours and the process is not scalable for large amounts of data. In addition, companies that work with multiple customers typically receive a large number of products every hour, and each customer categorizes their products in a different way. Therefore, it is impossible to easily merge the data.

Solutions

We supported the customer in sifting through his data sets and tested various machine learning models together. Based on the data, we selected the best performing system and further optimized it until it was ready for the market.  Particular attention was paid to the speed and accuracy of these models so that they could run live and categorize products in very small time units (usually milliseconds). Human intervention after go-live was reduced to checking random samples.