Copyright 2017 IQVIA. All rights reserved. Achieve Superior Segmentation and Targeting through Integrated Analytics of Real-World Behavioral Data and Market Research Attitudinal Data Emily Zhao, VP, Advanced Analytics, IQVIA Anjani Tripathi, Director, Advanced Analytics, IQVIA Elizabeth Wallace, Sr. Manager, Advanced Analytics, IQVIA Clare Tong, Sr. Principal, US Primary Intelligence, IQVIA
Background Understanding customers from both attitudinal and behavioral aspects is essential for developing an effective segmentation and targeting in order to deliver the right message to the right audience at the right time However, leveraging attitudinal and behavioral data in a truly integrated manner to achieve an optimal segmentation solution is still a significant challenge, in addition to the costs associated with such segmentation exercises Here we present an innovative approach, which applies Machine Learning techniques to analyze attitudinal and behavioral data cohesively to obtain superior segmentation solution by uncovering insightful behavior patterns and their attitudinal drivers 1
Objective Identify patterns as well as intensity of relationship between physicians attitudes and behaviors Develop segmentation solution that will optimally differentiate physicians on both dimensions Provide recommendation on differential messaging, compelling and specific to each segment, to impact physician behavior through sales force promotion and non-personal promotion Integrate Attitudinal And Behavioral Data Determine Target Behaviors Identify Attitudinal Drivers Of Target Behaviors Segment Physicians Based On Attitudinal Drivers Develop profiles and insights to guide targeting and messaging decisions Predict Segment Membership For Entire Universe Target list generation for promotion execution 2
Data Sources Attitudinal Data Behavioral Data + Awareness and Attitudes Toward Treatment Intent to prescribe Attribute Importance and Brand Perception Barriers to Prescribing Sales Force quality rating Cost-sensitivity / willingness to work through access restrictions, etc. Prescribing volume Prescribing patterns Branded vs Generics Early Adoption Payer Mix, etc. Attitudinal attributes were sourced from existing PMR/Tracking studies (ATU, Demand studies, BrandImpact), eliminating research costs, while enhancing the comprehensiveness of the data. 3
Determine Target Behaviors A comprehensive correlation analysis was carried out to develop insights into relationship patterns of attitudinal and behavioral attributes. Identify Attitudinal Drivers Of Target Behaviors Random Forest machine learning technique was used to understand how physician attitude/beliefs relate to target behaviors. Attitudinal Attribute Treatment Class A TRx Treatment Class B TRx Correlation Results * X = High Correlation Behavioral Attributes Treatment Class C TRx Brand X Share Within Class A Attribute 1 related to perception * * Attribute 2 related to perception * Attribute 1 related to awareness * Attribute 2 related to awareness * Highly correlated with attitudes Ease of driving change Important for brand s continued growth Tied to brand performance KPIs Top attitudinal drivers of each target behavior were selected for segmentation Ranked in descending order of importance Attitudinal Variable Importance In Predicting Target Behavior 1 Attitudinal Attribute Description Importance Attitudinal Attribute5 W5 Attitudinal Attribute2 W2 Attitudinal Attribute1 W1 Attitudinal Attribute7 W7 : : 4
Segmentation Results & Insights Attitudinal drivers of selected target behaviors were fed into Latent Class Clustering, resulting in a segmentation solution that is highly differentiated on both attitudinal and behavioral measures High High High Behavior Behavior Behavior Low Low Low Negative Attitude Positive Negative Attitude Positive Negative Attitude Positive Strongly Agree on all attitudinal attributes. High prescribing behavior driven by favorable attitudes Continue with the current messaging strategy Mixed responses. "Disagree on certain attitudinal attributes. Med. prescribing behavior driven by mix of favorable and unfavorable attitudes Tailor messages to reverse attitude associated with attributes 2 and 5 Disagree on all attitudinal attributes. Low prescribing behavior driven by unfavorable attitudes Current messages are not working Prioritize attitudes based on ease of influencing them via messaging, Develop messages to target those attitudes 5
Predict Segment Membership For Entire Universe A predictive model was built utilizing secondary data derived predictors to assign segment membership to the entire physician universe Segmentation Universe Build multiple predictive models that differentiate among segments Use the best model to score entire physician universe Final Target List The final model showed great accuracy in predicting the segment membership using secondary data. This is due to the fact that segments were not only highly differentiated on attitudinal measures, but behavioral measures as well. 6
Conclusion The approach serves as an useful framework for systematically identifying and linking physicians attitudes/beliefs with their actions and has several advantages over traditional approaches Cost Effective Utilizes attitudinal measures from all available existing primary market studies (ATU, Demand etc.) to reduce overall research costs, while raising the comprehensiveness of the data at the same time Integrative A truly integrated approach which considers both types of measures cohesively and simultaneously to develop segments that are highly actionable for an effective messaging and targeting Insightful Provides rich insights into how physicians attitude influence their prescribing behavior, and uncovers complex patterns that drive brand adoption and loyalty Actionable highly actionable for segment membership assignment for the entire customer universe, target prioritization, resource allocation and message development, laying solid foundation for a successful promotion strategy 7