2011 Fall Symposium

Find the useful information hidden in your data! This workshop covers computer-intensive methods for classification and estimation, drawn from Statistics and Data Mining. Dr. Elder will describe the key workings of leading algorithms, compare their merits, and briefly demonstrate their effectiveness on real applications. The course emphasizes practical advice and focuses on the techniques essential to successfully analyze data for business challenges. Along the way, attendees will be able to try out key steps using a leading commercial software product for Data Mining.

Modern modeling techniques, like Decision Trees and Neural Networks, will be explained and compared with conventional methods, such as Linear Regression. The great value of Visualization, Resampling, and Ensembles will also be demonstrated. Concepts will be illustrated through real-world problems drawn from several domains.

Dr. Elder heads a data mining consulting team with offices in Charlottesville, Virginia and Washington DC. Founded in 1995, Elder Research, Inc. focuses on scientific and commercial applications of pattern discovery and optimization, including stock selection, image recognition, text mining, biometrics, drug efficacy, credit scoring, cross-selling, investment timing, and fraud detection.

In the world of Big Pharma centralized, national planning is still king. However, there are clearly sub-national differences and a compelling need to understand and address that variability -- be it in physician responsiveness to managed care influence (or other regulatory differences), effectiveness of promotional channels, access to physicians or demographic factors. It is imperative that brands address these issues in their strategic planning.

In this session, we will talk about our experience in building predictive actionable sub-national plans based on measurable sub-national differences. Specifically, we will walk you through two case studies – "Territory Level Call Capacity" and "Sub National Level Growth Drivers". The intent is to provide a flavor of how business objectives, analytical frameworks, data and techniques are leveraged to provide focused insights to the overall brand strategic planning process.

Historically most pharmaceutical companies have assumed the same call capacity for each rep on a given team despite wide variations in historically delivered call volumes across geographies. Our case study "Territory Level Call Capacity" uses regression methodologies to uncover the key drivers of variations in call delivery within geography. This, in turn, was used by the client to determine future call delivery goals that were appropriate in all geographies. Several factors pertaining to geography, environment, targeting, reps and execution were factored into the study. In examining the model results, we also uncovered the truth behind many myths relating to call delivery. For example, a common myth is that the length of relationship between a rep and a physician impacts the quality, but not the quantity of the calls. Our model showed a significant impact on call volume as well.

Predictive analytics has been used to solve business problems over the last century. In this era of limited marketing budgets, it has never been more important to leverage advanced analytics to roll out smart and efficient marketing programs that deliver impactful results and higher ROI.

In this presentation, we will share a case study utilizing cutting-edge uplift modeling techniques to optimize a client's savings card program. Savings cards or co-pay cards are powerful tools to motivate adoption and retention, yet they can consume a substantial portion of the marketing budget if not used effectively. Uplift models will help you identify the appropriate combination of patient targets and financial incentives that will enable brand teams to meet their revenue objectives and deliver a healthy ROI. We will shed the insight of uplift model development as well as the lessons learned and other various challenges in implementing this innovative predictive analytics solution.

Pharmaceutical marketers have struggled with the challenge of optimizing the overall marketing mix for some time now, and reasonable progress has been made. Most companies have a decent understanding of how to measure ROI on individual programs or campaigns of professional marketing, and the techniques for doing so are well-understood. Data availability issues can sometimes remain a problem, but these are generally declining in the face of increased computing power and data marts. However, all of the above fails to hold with consumer marketing, and particularly consumer digital marketing. A variety of channels exist (banner ads, paid search, branded websites with or without consumer registration, etc.). Some of these currently benefit from predictive analytics (e.g., optimal pricing for "ad words" in paid search). But most of these are still poorly measured (even for basic ROI) and are generally not yet fully integrated into the overall marketing process. In particular, the interaction effects with professional marketing are not yet well understood, and an appropriate budgeting framework (considering those interaction effects) is lacking. Based on our experience both in Pharma and other industries (financial, retail, etc.), this talk will explore these issues and the role that predictive analytics can play in resolving them.

• We will present case examples that illustrate approaches for measuring ROI of digital marketing programs in Pharma (addressing the challenge that most consumer digital marketing is difficult to link to actual Rx volume)
• We will present a framework for modeling consumer promotional channels and their interaction among each other, and with professional promotion. This allows for optimization of overall budgeting (accounting for interaction effects) and lays the groundwork for a more detailed microtargeting on both sides.
• Examples from non-Pharma industries will be reviewed, time permitting.