Starting state: lack of insights from product
Our customer is one of the largest American retailers. They collaborate with hundreds of vendors that are uploading content to their website. They realized that they had a lot of product content at their disposal but no insight into the actual quality of data or how this can be used to gain insights into conversion improvement or creation of additional sources of revenue.
They wanted to ensure that their product content was high quality (product images, product descriptions, and product reviews) and create a scorecard for vendors for their submitted product content. They would then give vendors insights about how to optimize their submitted product content to maximize sales and prevent low quality content from being uploaded.
Building a product recommender with super.AI
The customer reached out to us to help them better categorize vendor content and use the product information to build a product recommendation engine.
To remove the need for any customer IT resources, we scraped all product content from their website. Then we:
- Categorized the images
- We scored the products by adherence to customer guidelines, variety, and quality
- We rolled up the product scores to the brand and category level
This information was then combined with transaction data and used to compare products to create a recommendation engine and calculate lift from content improvement.
The customer had granular metadata on their product content which was used to build models for product. The customer was able to start product recommendation via email campaigns and on their website for the first time, and improve their and their vendor’s bottom line.