1365 N. Scottsdale Road PMSA Symposium 2020 Speaker Presentation Suite 100 Title: Dynamic Targeting Solution for Field Force Scottsdale Authors: Karthik Somadri, Ankit Chhabra and Robin Varghese, CustomerInsights.AI AZPresenters: 85257 Ankit Chhabra, Sr Director, CustomerInsights.AI Abstract Even with the emergence of new technology-based marketing tactics, “Field Force Detailing” continues to dominate the promotional landscape for Pharmaceutical & Biotechnology companies, accounting for the largest bucket for promotional spend. Analytics around Field Force Targeting appears to have been beaten to death over the years, but it still remains as the primary challenge for Life Sciences companies to ensure they maximize their ROI on their largest promotional spend. Through our two decades of experience in this space, we have come to learn that there is value in moving away from a Static Once-A-Year Targeting solution (majority of Pharmaceutical & Biotechnology companies follow this today) to a more Dynamic Targeting solution, that is flexible to emphasize or de-emphasize targets based on changing conditions on the ground. Why should a Pharmaceutical or Biotechnology company consider moving towards a Dynamic Targeting solution? • Improve ROI by calling on the “right targets” at the “right time” • Minimize wasted efforts, especially with trial-error process on Medium & Bottom Tiers which typically accounts for 50-60% of the overall call volume • Create tailored messaging based on ground reality for each physician • Identify and engage new potential HCPs in an ever-dynamic market Our Data Science team has developed a three phased solution to enable Dynamic Targeting for Field Force: • Integration of New Data Sources • Real-time Insights • Predictive Insights using ML/AI Phase 1: Integration of New Data Sources Traditional approaches rely heavily on prescription sales data, prioritizing physician based on brand and market sales performance. But this never tells the complete story. Understanding the disease and journey of a patient in its entirety is very critical in determining the true potential of a physician. This can be achieved by leveraging newer data sources like Claims data, Biosimilars data, Lab data, Payer-plan data, Clinical data, Formulary data, Medical devices data, Social/Digital data, Hospital data etc. We work with our clients to integrate as many data sources as possible to build a rich history of activity against every physician. This is then followed by a detailed exploratory data analysis and a complex index scoring methodology using the integrated information to evaluate the relative importance of each physician. Finally, this entire process is “operationalized” through a Big Data platform to auto-update as frequently as data is made available. HCPs may get tagged with a new priority index over time based on real-time factors like changing payer landscape, competitive changes, changes in HCO ownership/integration, shift in HCP behavior, changes in patient volume, etc. PHYSICIAN DEMOGRAPHICS • • • • • • Territory, Zip and State related information for HCPs and Patients • PATIENT PROFILE • • Age • Gender • Primary payment type (Commercial, Medicaid, Medicare, Assistance Programs, Cash) Physician Specialty # Total MDs and Visits # Drug X MDs and Visits # MDs w/ Drug X Call Difference in visits/MDs in last 6M vs. Prior 6M Primary Physician Gender and Experience Avg. Physician Decile CLINICAL TRAITS • • • • All Diagnosis (Dx) codes All Procedure (Px) codes All Prescription (Rx) codes Lab test results Page 1 of 3 1365 N. Scottsdale Road PMSA Symposium 2020 Speaker Presentation Suite 100 Title: Dynamic Targeting Solution for Field Force Scottsdale Authors: Karthik Somadri, Ankit Chhabra and Robin Varghese, CustomerInsights.AI AZPresenters: 85257 Ankit Chhabra, Sr Director, CustomerInsights.AI Phase 2: Real-time Insights This phase focuses on tracking multiple events of interest on real time basis that can directly impact the priority of a physician in the already established targeting schema (usually derived from Phase 1). These events are set up to be agile to change over time based on current business priorities. For example, an event could be to track activities around a competitive launch situation and track patient profiles of interest and their associated physicians to keep a check on competition. Some examples of patient related events include: a. Patients switching to competitive drug b. Confirmation of diagnosis c. Specific lab results d. Rejection of your product’s claim or competitor product’s claim e. Undergoing certain procedures f. Experiencing side effects Some examples of physician related events include: a. HCPs writing Biosimilars b. HCPs with an undesirable Source of Business Mix c. Important HCPs with lower than expected call plan adherence d. High writers with declining NRxs e. HCPs/Practices facing higher payer rejections f. HCPs New To Market or New To Brand g. HCPs highly affected by recent Managed Care win for Pull Through As done in Phase 1, this phase is operationalized through a Big Data platform to auto-update as frequently as data is made available. The Big Data platform, which also has “specific” field level views, is used to inform and empower the Field Force with ground intel. The views are also made interactive to easily get actionable call lists for each type of event which helps the Field Force customize their messaging for each HCP. Page 2 of 3 1365 N. Scottsdale Road PMSA Symposium 2020 Speaker Presentation Suite 100 Title: Dynamic Targeting Solution for Field Force Scottsdale Authors: Karthik Somadri, Ankit Chhabra and Robin Varghese, CustomerInsights.AI AZPresenters: 85257 Ankit Chhabra, Sr Director, CustomerInsights.AI Phase 3: Predictive Insights using ML/AI Majority of today’s targeting solutions are all centered around available data. With the emergence of new technologies, we can now leverage advanced ML/AI techniques to confidently predict future events in the data. In this case, ML/AI techniques can be used to predict key events in the journey of a patient even before they have happened by learning from past behavior of other similar patients in the universe. Through this predictive process, one can better understand the “true potential” of a physician which is not completely understood through any of the other phases. We have a comprehensive approach for model building & validation against real time data. We run hundreds of models across a wide variety of ML algorithms like Linear models, Tree based models, Support Vector Machine based models, Ensemble models (Random Forest, XG Boost, etc.) and customized blending/stacking methods and evaluate pros & cons against each other. Our Data Scientists have built models with a prediction accuracy of 98% in some client instances. Finally, insights from this phase get integrated with Phases 1 & 2 to empower the Field Force to take smart decisions on the fly. PU Classification Engine Dataset Creation Client dataset Drug X patients Analyzing PreDrug X pathways Open Target (patient pool) Working dataset for classification model Patients to HCP mapping Drug X Patients are assigned positive label, while others from patient pool are labeled negative Assigning patients to HCPs Unlabeled dataset Stratify overall patient pool Create Positive Unlabeled dataset using selected features Feature Engineering Training set (80%) Testing set (20%) Model Training Model Testing Model Evaluation FINAL Model Prediction set from patient universe Identify Drug X markers List of current & potential Drug X patients & their HCP mapping Drug X patients Potential Drug X patients Aggregate at HCP level & bucket into HML High Potential Medium Potential Validate Predicted Drug X patients from the pool Low Potential To conclude, competitive & constantly changing marketplace calls for advanced strategies for the Field Force. Traditional Once-A-Year targeting solutions (irrespective of the complexity of the model) are just not effective in today’s environment. Real benefits will be realized by clients who move to a Dynamic Targeting solution that can stay in tune with reality on the ground. To realize the benefits of this approach, a client’s entire system must be able to • Adjust any target up or down the entire system based on market conditions & its impact • Adjust messaging mix for each individual target physician based on his or her specific ground reality. Don’t be surprised if this type of thinking and engagement gets a lot of pushback from the Field Force who is used to following a set plan in their respective geographies. Changes of any kind takes patience and collaboration with the Field Force. We are doing exactly this for one of our clients. During the actual presentation at the Symposium, we will walk through an actual client case study sharing insights on all three phases to help event attendees gain some valuable knowledge on the approach & challenges that they can bring back to their organizations for immediate reference. CustomerInsights.AI Founded by Life Sciences & Technology leaders with 25 yrs of experience, CustomerInsights.AI is an Artificial Intelligence company that builds innovative ML/AI solutions with strong emphasis on self-service, stateless execution through automation, speed to insight, and lower costs. We believe that true value of AI will be realized by Pharmaceutical & Biotechnology companies when AI becomes part of one’s everyday plumbing and not a one of event at the individual or project level. Ankit Chhabra, Sr Director With over 12 years of Predictive Modeling & Data Sciences experience, Ankit focuses on challenging status quo by leveraging advanced ML/AI techniques to build innovative solutions to solve Sales Force Effectiveness problems for Life Sciences and Health Care clients. Page 3 of 3