Image by John Schnobrich

Customer Segmentation

Personalize Your Customer's Experience with AI




A medical company that delivers medication for chronic conditions at a reduced price wants to maximize patient acquisition through a customer service center. This company owns 20 million patient lab records.

  • 3% Conversion. Although the service created savings and convenience, their customer service representatives (CSRs) had a poor acquisition rate.

  • Poor Data Quality. Although the service had access to millions of records, it could not benefit from it due to its unpreparedness for analysis.

  • No Data Infrastructure. Each CSR had a unique way of inputting data into the system, which caused a lack of unification for further examination.


The entire data inflow and dimensionality had to be modified to be able to process and analyze the information to find non-intuitive correlations in the data

  • Proper Data Ingestion. Data wrangling and cleansing operations had to be done, which caused the clearing of around 2 million records and preparation of the other 18 million while also fixing the options in which the CSRs could choose when inputting the data.

  • Data Sources. Data sources were chosen based on necessary needs, such as financial data (e.g. insured or not), demographic data (e.g. age and number of dependents) and circumstantial data (e.g. number of visits to different labs).


Random Stratified Sampling. Instead of simple random sampling, an approach that showed more promise was Random Stratified Sampling, which is a statistical method that forms subgroups from the population according to similar characteristics and then samples from each subgroup. After sampling, a variety  of Machine Learning algorithms are used to forecast patient acquisition and how to approach the patient.


  • Conversion Growth. Conversions grew from 3% to 18%, which is also higher than the global average by 15% due to proper cohort analysis.

  • No Loss. CSRs have a limited yet diversified set of data inputs for each patient, therefore inducing no loss of data throughout their operational duties.

  • Better Recommendations. CSRs are much better equipped at specific product targeting, script setting and method of reaching out.


The project is successfully running, minimizing inaccuracies and increasing scoring quality as it enters phase 2: The Insight-Driven Transformation phase.