97% of home searches take place online. Identifying buyer intent is a classic data science riddle, the size of which is forty three trillion dollars in residential real estate. Vow, a whopping 97% of buyers begin their house search online, imagine the revenue potential of online services offerings possible.
A typical online buyer journey begins with identifying the neighborhoods for buying the home, browsing homes for sale online, setting up auto alerts, shortlisting the homes to visit, working with an agent and closing on their dream home. The online and offline data is powerful to determine the buyer intent. Though it is possible to track the offline actions of the buyers, for privacy reasons, will stick to click stream data which users consent for the websites to analyze.
Real estate portals identify the intent of users by Propensity modeling based on thousands of click stream events. The input to the model is the behavior of the user on the website which leads to feature selection to train the model. Multiple features (data points) are evaluated such as browsing homes in certain areas of the city, setting up auto alerts, saving homes, campaigns engaged, changes of preference over time, interacting with preferred agents online etc. Target labels to train the propensity models are based on desired outcome. Feature selection as described above helps to score the model. The propensity model assigns a score to all users, which is used to serve relevant customer focused experience such as surfacing relevant homes or suggesting agents with expertise in the specific neighborhoods.
The real challenge is to close the loop between online and offline searches to narrow down the serious buyers. As the data science and AI space evolves further, the consumer will be the real beneficiary.