Froomle beats Google AI

We benchmarked 3 different in-house recommender systems against Google AI

Froomle beats Google AI

We benchmarked 3 different in-house recommender systems against Google AI

CHALLENGE

What has been their challenge?

Could Froomle beat the Google Recommendation AI (GRAI) in CTR and conversion? Based on the possibilities of the GRAI, this comparison could only be based on the Dreamland homepage carrousel on the Dutch language website.

Solution

How did we approach this case?

We continuously benchmarked 3 different in-house recommender systems against GRAI, keeping the one with the best results after each iteration and testing new variations. Our fundamental techniques were collaborative filtering, content-based recommendations (using product metadata), and a deep learning technique.

Process

How did we solve their problem?

This is the final result

We outperformed Google with both our collaborative filtering and deep learning techniques, improving CTR for known users by 20%. During the benchmark, we gathered learnings to increase homepage relevance, including showing products clicked before, diversifying recommendations based on brand, and using a short popularity window to catch trends.

Download case studyDownload case study
Graphic CTR and average attributed revenue per visitor
Next step

Start your 40-day Trial today!

Get in contact with our personalization experts -
start the 40-day trial to improve your user’s experience.

  • Recommended the most relevant products or content
  • Increase brand loyalty
  • Improve your CTR

By submitting your information, you agree to Froomle's Cookie policy and Privacy policy. You can opt out anytime.

Your information has been saved.
Looks like we're having trouble