Leveraging ML Capabilities for providing Personalized Real Estate Recommendations for a Proptech Conglomerate

Business Objective

Our client is a Proptech conglomerate with operations in Australia, USA, and India. They used a heuristics-based relevance logic to rank order properties on their search results page. This was heavily focused towards monetization, i.e., generating more leads for sellers (brokers, builders, and more) who have purchased specific packages to prioritize their properties with potential buyers.

Due to this, they were facing challenges that affected the Customer experience to a large extent and also reflected in their revenue generation
ecosystem.

Challenges

The current methodology did not capture user browsing patterns, user behavior, or user likes
Difficulty in providing near real-time recommendations
Difficulty in processing the live clickstream feeds to create a feature store
Cold start problem for new properties and users

Solution Methodology

Set up a data processing pipeline to continuously consume live clickstream feed from the web & app and create a raw data store.
Created property profile vectors by identifying key property attributes based on historical lead submissions and applying top search filters.
Derived implicit ratings for the properties from the users based on the kind of touch points. Weightage for these is based on their magnitude of correlation with lead submission.
Next, derived a user-specific profile vector across key property dimensions by aggregating historical user-property touchpoints and applying search filters.
Built a two-stage recommendation framework
Built a similar property framework to handle cold-start on newly added properties between model re-training windows.
For cold-start users – Created heuristic rules to identify and recommend the most popular properties within customized cohorts and recommend them.

Business Impact

Implemented a cost-optimized near-real-time personalized real estate recommendations framework in 20 weeks
22% increase in coverage of properties shown while also bringing in personalized recommendations as per the user’s requirements
Improved customer experience on both the website & app
Generated more CTR and leads for sellers

Privacy Notice Cookie Policy