Matching patients with coordinators

Summary

Recommendation infrastructure for Bookimed’s call centre.

Coordinators handled varied patient requests, while needs, preferences and staff experience affected which pairing made sense.

A real-time recommendation engine using request context, website behaviour and coordinator experience to support assignment and follow-up.

Tech Challenge

• We had to build a scalable recommendation engine based on deep learning model that would increase the conversion of prospects to customers by matching them with the right coordinator at the call center.
• The pairs are formed based on prospects' medical needs, clinic location and treatment preferences, their behavior on the website and the skillset of the coordinator to convert such prospect into a client.
• The recommendation system supports request assignment using patient needs, preferences and coordinator experience.

Solution

• Recommendations are processed in real time to support coordinators’ handling of individual requests.
• Built from scratch, the solutions is an individual customizable recommendation engine, which takes into account hundreds of parameters coming from client’s analytics engine.
• Model was trained on Tensorflow and exposed with TF Serving.

Impact

A recommendation engine integrated into Bookimed’s daily coordinator workflow.