Select a Region North America

Account Engagement Excellence: How Advanced Analytics Can Enable Next-Generation Account Segmentation

Account/customer segmentation is a crucial exercise at the beginning of any strategy related to healthcare provider, patient, or payer engagement. Due to the large variety of customers, it is critical to know one’s customers well, from basic information such as demographics, firmographics, and geography, to more nuanced aspects of the customers’ beliefs, attitudes, and the psychology that drives their behaviors. All this information is critical to formulating the right messages and strategies to engage customers most effectively.

Account segmentation allows for accounts to be grouped based on their behaviors, capabilities, strengths, and challenges. By grouping accounts, one can better understand similarities and differences to draw implications for account needs and priorities. These data points should be used during account planning to ensure alignment on needs, the role a pharmaceutical company can play, and how best to message based on an account’s
individual needs.

Traditionally, account segmentation has been a manual process that only considers a few variables. The next generation of segmentation, however, is more powerful, involving a large number of different variables for each customer, and is driven by robust statistical methods and machine learning algorithms to help minimize human biases and errors.

Machine Learning Approaches for Account Segmentation

Machine learning is beneficial when product receptivity is influenced by complex and numerous segmentation dimensions or categories of parameters. The optimal segmentation approach depends on the strategic goals and intended application of the account segmentation. The two main types of machine learning approaches to segmentation are cluster-based approaches and rule-based approaches (also known as tree-based approaches). The approach a business chooses to follow for account segmentation depends on three key things: objective function (a combination of variables that directly influence the desired outcome), data availability, and intended use of the segmentation output.

Cluster-based approaches, such as k-means or latent class, are best for situations with multiple segmentation objective functions and a large number of driving dimensions (Figure 1). This
approach can account for more drivers, such as a complex market or a multi-product segmentation. However, it can get complicated, as segment scoring can be more challenging compared to the rule-based approach.

Rule-based approaches, such as Classification or Regression Trees (CART) or Chi-Squared Automatic Interaction Detection (CHAID), are typically useful for situations with a clear business objective and relatively few driving segmentation dimensions. The advantage of a rule-based approach is that segmentation solutions are usually quite intuitive and driven by hypotheses based on business logic. The downside is that it may not be useful for complex markets that require more variables to drive the segmentation.

Complete the form below to download the full white paper. 

Hai Nguyen
Associate Principal

Hai Nguyen is an Associate Principal at EVERSANA MANAGEMENT CONSULTING with a background spanning life sciences, healthcare administration, and financial/economic consulting. Hai has vast experiences in conducting quantitative and qualitative primary market research, building…

Tanuj Mehra

Tanuj Mehra brings more than a decade of experience working at the intersection of commercial and R&D with both large pharmaceutical and smaller biopharma clients. He has led numerous global engagements with leading Fortune…

Keyla Badillo, PhD

Keyla is a Senior Consultant at EVERSANA™ MANAGEMENT CONSULTING, with a background in pregnancy research, genetics, molecular biology, and diagnostics. Her experience in life sciences and healthcare consulting spans several therapeutic areas, such as…