Situation: A large real estate company in South India wanted to understand their 3000-odd customers in their database better to tailor their promotions and offerings appropriately.
Approach: We worked with ~60 factors of information on each customer and used an AI technique, Self-Organizing Maps (SOM) to cluster customers with similar attributes together. We also used multi-variate regression to identify the factors that affect the buying behavior of customers the most.
Results: We identified 5 distinct segments of customers and characterized their behavior to help the client better understand their customers. The model also highlighted which customers were profitable and which were less, that helped the client identify key attributes to predict the future margins for a customer, thus enhancing their customer profitability.
|Client:||A $300 M Real Estate Company in South India|
Situation: The client required a prediction model to determine the probability of re-purchase for new customers.
Approach: After evaluating various machine learning techniques with different permutations of sales data and customer attributes, we identified Gradient Boosted Trees as the algorithm best suited for the client's data. We then developed a sustainable software solution with a user-friendly interface to generate re-purchase probabilities for new customers.
Results: More accurate predictions (72% AUC) allowed the client to run more effective marketing campaigns, thereby resulting in significant top-line gains.
|Client:||A $2B US multi-brand specialty retailer|
|Tags:||Customer Insights, Predictive Analytics|