About Bayser


Publications, Conferences and Tutorials

Contact Us


Compliance & Persistence
Patient Prescription Behavior
Physician Networks
Hospital-Retail Spillover
Patient-Level Targeting
Other Services


Individual Patient Behavior Tool
Rx Tracker
MC Viewer/Targeter
Deal Finder
Look-See Pipeline
Bayser Aligner


Slot Machine
Rat Race of Healthcareville
Oncology Cards

Bayser Consulting
Skokie, IL 60076


JP gave a talk at the PMSA 2002 (Pharmaceutical Management Science Association) in New Orleans (April 28 - May 1, 2002) on how to build accurate promotion response models by aggregating physician-level responses. Here is a practitioner's account of the PMSA 2002 conference.


PMSA 2002 Conference in New Orleans  (April 28 to May 1, 2002)


A Practitioner’s Account of the Event

By Jean-Patrick Tsang, Bayser

Disclaimer: The account below by no means reflects the views of PMSA, speakers at the PMSA, or attendees of  the PMSA. It is meant to portray the views of JP and JP only.


There were two pre-conference tutorials on forecasting on Sunday. Since I did not attend either, I won’t say much.  There were 8 talks on Monday. I attended all of them and offer my comments below. Tuesday had two tracks. One on sales force effectiveness: four talks plus a panel. This is the one I participated in. The other one was on Direct to Consumer. I would have liked to attend but could not materially do so. Wednesday morning was dedicated to e-business. Since I had to get in and out of the talks, I could not give my undivided attention to the speakers. In the interest of objectivity, I will refrain from making any comments. All told, I’ll give you my two cents on 12 talks: 8 on Monday and 4 on Tuesday morning.


If there is one trend that stands out, it is clearly patient-level data. In addition to talks, no less than four vendor booths out of the 15 or so (that’s a solid 25%) were singing the virtues and praises of patient-level data. Rightly so! I see the future being in the combination of patient-level and traditional physician-level script level data (e.g. IMS Xponent or NDC Source Prescriber).

Monday April 28 Session

  1. Jeanne Scott (NDC Health): Update on Medicare Rx Drug Benefit. Jeanne, a veteran industry lobbyist, delivered a great presentation on a slew of issues ranging from Medicare bankruptcy, 34 million uninsured people, Medicaid, double digit cost increases, aging baby boomers, premium support, etc. She clearly understands the big picture and quoted compelling statistics to bring home the bleak picture she was portraying. Spirited speaker!
  1. Andy Zoltners (ZS): Sales Force Decision Models: Insights from 25 years of implementation. Andy talked about insights on sales force sizing, structure, resource allocation, and territory alignment gained from several industries. If you are new to the industry, you may have learned something.
  1. Dick Anderson (Mattson-Jack ROI): ROI of Promotional Campaigns. Dick kicked off by demystifying the complexity involved with the prediction of promotional impact. He went on to present a four-component model that zeroes in on market environment, promotional environment, product environment, and disease environment. He talks a lot about momentum but refused to give a crisp definition of what that concept really entails. I was expecting some lofty parallels with mass and velocity. Did not happen.
  1. Franklin Carter (St Jospeph’s University): Developing a call attractiveness model: a disaggregate negative binomial model applied to the pharmaceutical industry.  Franklin combines a Poisson with a Gamma distribution to develop a model that predicts the number of prescriptions the physician will write in a given promotional context. The model takes into account promotional dependent as well as promotional independent parameters to forecast the probability with which the physician will write scripts of the drug.  Great presentation. Franklin gave one of the best talks of the conference in my humble opinion.
  1. Avi Shatz (Intercon): Standardizing ROI around compliance and persistency programs – Making pharmacy intervention programs pay off.  Avi took the angle that the whole industry had been oblivious to the most important player in healthcare: the patient. He went on to explain that patient-level data is the way to go and more importantly the behind-the-scenes computer system that makes all that possible. He is right when he talks about the virtues of patient level data. He would have made a more compelling case if he underscored the current difficulties and limitations of the patient-level data.
  1. Kevin Kirby (GSK) and Paul Rabideau (Novartis): Effectiveness of Print Advertising. You probably recall Dr. Scott Neslin’s ROI Analysis, also known as the RAPP study, that sought to measure and compare the ROI of detailing, DTC, JAD, and PME. The results of the study were published a year ago in MedAdNews, Medical Marketing Media, and can be accessed on the web at www.rappstudy.org. Well, it looks like the regression methodology is flawed, prompting a second more detailed study. Kevin and Paul outlined the study and invited other pharmaceutical companies to participate. As of today, there are two participants: GSK and Novartis.
  1. Kamel Jeddi (Columbia): Understanding sources of New Product Demand: Which Market segments/competitors? Kamel described a pre-launch new product model that forecasts sales of new products based on patient simulation studies. The model identifies segments of the physician population and evaluates alternative market strategies called concepts. Interesting approach.
  1. Michel Denarie (Quintiles) and Brian Burk (Pharmacia): CME Event Campaigns: a new way to design and evaluate them using patient-centric data. This presentation had some nice slides that brought home the fact that the same prescription level data snapshot can actually correspond to very different patient compliance and persistency situations.  Implication? You need patient-level data to get a finer-grained view!

Tuesday April 30 Sales Force Effectiveness Track

  1. Terry Overton (ImpactRx): The Ripple Effect – Unintended Experiments in Promotion Response.  Terry focused on the withdrawal of Bayer’s Baycol and looked at those who benefited. First, other players in the lipid market: Lipitor, Zocor, Pravachol, Lescol. Second, other products from Bayer, namely Cipro and Avelox. What is interesting is this does not stop there. Indeed the ripple effect hits the quinolone, oral diabetes, anti-depressant, ARB, and anti-arthritic markets as well. Rationale: the detailing zero-sum game whereby pushing harder on one drug amounts to pulling back on the other drugs.
  1. Jean-Patrick Tsang (Bayser): Building GIGO-Free Promotion Response. I started off by making the case why response models are important for sales force sizing, portfolio management, and product promotion. I introduced the promotion response model and quickly opened it up to incorporate the notion of opportunity cost among other things. The Achilles’ heel of the whole approach is the accuracy of the promotion response model. I described a new way to build response models by aggregating physician-level responses and discussed the implications on adaptive targeting. The talk was extremely well received.
  1. Sandy Balkin (Pfizer) and Edward Bryden (Pfizer): Non-linear Time Series Models for Pharmaceutical Forecasting.  Sandy and Edward did a comparative study where the contenders where Neural Networks, Projection Pursuit, Regression Trees, and Multivariate Adaptive Regression Splines. The conclusion, as one would expect, is that non-linear models outperform their linear counterparts. Very good talk.
  1. Scott Nass (Anabus) and Donald Rubin (Harvard): Estimated Causal Effects – A New Metric for Sales and Marketing Effectiveness. The talk started by explaining that the impact of a promotional campaign should not be measured by comparing the after-promotion situation with the before-promotion situation. Rather, it should be between what happened with promotion and what would have happened without promotion. The catch is we do not know what would have happened. Solution? Good old paired test control where the pairing is more trendily called “cloning”. The contribution lies in an algorithm that summarizes a very large number of physician characteristics into just one: the linear propensity score. Very interesting. Challenges: sell the black-box propensity score to reps, let alone upper management!


Any feedback, comments, suggestions, or questions, feel free to send an email to Jean-Patrick Tsang (JP) at bayser@bayser.com.



“One of the brightest individuals that I've met in my career. JP has an incredible skill regarding simplifying issues, and preparing presentations for senior management." -- Director, Large Pharmaceutical Company

“Extremely brilliant and gets it right away.” --Director, Large Pharmaceutical Company

“Very professional!” --VP, Large Medical Devices Company

“JP and I are a great team. I get all kinds of ideas and he gets them implemented.” --VP, Large Pharmaceutical Company

“Always does quality work.” --Director, Large Medical Devices Company

"The amount of knowledge that they bought, not only about their tools but about the industry & tool applications.” --Market Researcher, Large Pharmaceutical Company

“I am very pleased with Bayser’s work. ” --Director, Large Diagnostics Company

“A real guru at Excel. Taught me everything I know about spreadsheets.” --Account Manager, Large Diagnostics Company

Reaction to a demo of Bayser’s Rx Tracker: “I am having an out-of-body experience right now.” --VP, Large Pharmaceutical Company



All contents copyright 2002 Bayser Consulting, (847) 920-1000