Recommending Online Courses

Recommendation systemMachine learningCollaborative filteringK-means clustering


Descola is a well-established Brazilian startup offering online education. Despite having over 30,000 customers already, their conversion rate was underwhelming and as a result, cost of acquisition too high to enable rapid growth. It seemed like it was difficult for customers to choose from the large number of courses and we were hired to change that.


In less than one month, we developed a real-time, plug&play recommendation system that reacts to user behaviour and promptly presents courses that the user is most likely to buy, based on immediate browsing history.

System Architecture

System architecture


  • To enable real-time performance, we used a combination of K-means clustering and collaborative filtering.
  • We also weighted the algorithm to favour newer courses.


1 week

time to market


conversion rate increase

22 ms

API response time

Read More Case Studies