Customer Acquisition via Location Intelligence and Machine Learning

Business in Society; Computer Sciences; Engineering and Society; Applied Statistics; Statistical Methodology; Information Science; Data Science

The project proposes a new solution to customer acquisition by identifying potential customers through spatio-temporal analytics and trajectory pattern mining using population-scale location data.

Research Interests
  • Business Strategy
  • Machine Learning
  • Data Mining
  • Spatial statistical analysis
  • Privacy

Customer acquisition and customer retention are key to the survival and profitability of any organization. While an organization can rely on the present customers' historical data for retention strategies, it does not observe the potential customers for acquisition. For instance, a Korean restaurant in downtown Charlottesville does not observe who might become its customers -- is it a UVA faculty who passes by it daily on the way home but never stopped in, or a student without a car who cannot reach downtown easily, or a local resident who never likes Korean food? It is imperative to understand who these potential customers are, who are most valuable to the organization, why they are not patronizing, who are more responsive to acquisition strategies, and how to acquire them.

While the literature has relied on either limited competitive intelligence (such as customer purchase data across competing organizations from a centralized source) or geo-fencing technology (targeting those individuals in proximity to the organization’s physical location). The former overlooks customer location intelligence and also a large number of potential customers who never purchased from any competing organizations but may be interested in the future. The latter accounts for location intelligence, yet holds no information about the nearby individuals’ brand preference. 

The recent availability of population-scale, granular, consumer mobile location data with privacy compliance grants unparalleled new opportunities to develop a panoramic view of individual mobility trajectories, 24/7 lifestyle, and offline patronage to any organization. These data address the shortfalls of the above two conventional customer acquisition approaches by combining location intelligence (proximity to the business, trajectory patterns, road map) with brand preference inferred from customer lifestyle and historical patronage to similar or related organizations.

To summarize, we aim to leverage the population-scale consumer location trajectories to address the following key research questions:

  1. How can we quantify each individual's "potential customer value" to an organization of interest?
  2. Why have these potential customers not patronized the organization of interest yet?
  3. Which strategies can we leverage to acquire these potential customers?

To accomplish the above, we will leverage a multi-TB-sized sample of location data of a major metropolitan area over a span of six months. We will leverage the functional statistical analysis and trajectory mining methods to uncover individuals' brand preference, lifestyle and trajectory patterns. Then we will calculate each individual's "potential customer value" as an integrative metric of preference proximity plus geo proximity to each organization of interest (such as the Korean restaurant in downtown Charlottesville). Then, for instance, a person visiting other Korean restaurants (or other Asian restaurants, spicy menu restaurants) in other parts of the city will be of higher potential value to the Korean restaurant in downtown Charlottesville. Similarly, a person who passes by this restaurant each weekday will be more valuable than someone who passes by it once a year. We can further combine this calculation with a trajectory prediction model to predict the optimal targeting timing.

Desired outcomes

The outcome of the project will be a new solution to customer acquisition by targeting potential customers. More specifically, it will include the following:

  1. Constructing a novel index to quantify each individual's "potential customer value" to each organization of interest while leveraging cutting edge statistical and machine learning methods;
  2. Uncovering the underlying mechanisms of why individuals with high potential customer values have remained potential instead of active customers;
  3. Testing optimal targeting strategies to acquire high-value potential customers and accomplishing the most effective and finessed targeting and customer acquisition of immense value to any organization.
  4. Developing novel frameworks to identify customers with high "potential customer value" while preserving customers' privacy.