Lessons in Simulation 1: How a Negative Debate Led to a Positive Solution

Prescribing simulation was born 20 years ago in the aftermath of a methods debate in which supporters of two flawed methodologies focused more on attacking the other than improving their own.

Takeaways For Marketing Researchers:

  1. Don’t settle for the “least-bad” method. You may need to get creative to develop a better approach, but the payoff in gaining clearer, more robust insights will be worth the effort.
  2. When evaluating physician prescribing, behavior-based simulation allows marketing researchers to avoid the compromises inherent in attitude-based research.
  3. Methods that work well in CPG or in certain forms of B2B research might not translate well to healthcare.

The Story

Prescribing simulation emerged 20 years ago from the verbal and philosophical equivalent of a food fight over the best way to test a physician’s price sensitivity for medicines. (Irony alert: back in those days, doctors never cared). There were fundamentally two camps: conjoint and monadic. The conjoint camp argued that a basic choice exercise could indicate the relative  importance of price in the offering for a new pharmaceutical as well as thresholds at which price beings to matter. The monadic camp argued that any method in which physicians could manipulate price of a medicine was inherently flawed. The pristine monadic approach involved creating parallel cells with a sufficient number of respondents, asking each cell to review a product profile in which the only difference was price, from which the researcher would determine whether the cells differed in product favorability or intent to prescribe.

Each camp was aware that its method had shortcomings, and responded by proclaiming that “the other approach stinks  worse than mine does.” (if you see a similarity between this line of attack and current political marketing, you are not alone).

The monadic camp argued that for pricing medicines, a choice method was unrealistic. It might work fine for consumer electronics for which the purchaser can custom design a configuration of features (memory, memory speed, monitor size, graphics, etc.) knowing exactly what each might cost, or for toilet paper for which a shopper can consider price while making a selection that might encompass colors, patterns, 1-ply vs. 2-ply, softness and brand name.

Pharmaceuticals are different. A doctor might want a medicine with the efficacy profile of the newest brand, the confidence in safety that comes from being on the market for five years and the price of a generic, a medicine that does not exist.

Insight: just because a method works well in CPG or B2B research does not mean that it will work as well in healthcare.

The conjoint camp could not prove these objections wrong, so adherents pointed to the extensive amount of academic literature “proving” that conjoint-based pricing studies “work” by producing actionable price curves. (It would be years before commercial researchers knew enough about Behavioral Economics to question whether smooth curves reflect actual human behavior or merely our biases.) They argued that while monadic might be a perfectly appropriate sampling design, it did not allow for sophisticated payoff questions. Indeed, the standard monadic payoff question was either a likelihood to prescribe rating for the product or an estimate of the percentage of patients that would receive the target medication. When monadic advocates argued that the absence of significant attitudinal differences between cells proved that price level was irrelevant to product success, they were on very shaky methodological ground.

Insight: A robust approach to sampling cannot overcome challenges inherent in attitude-based research methods.

The Solution

Tuning out the negative back-and-forth, I began to focus instead on each side’s merits. Monadic sampling might be the correct bias-free way to compare value propositions. Choice modeling provided richer, more statistically efficient insides than simple allocation scores and rating scales. What if a method could utilize monadic sampling and choice methodology in the same exercise?

This focus on solutions led to a final realization: the attributes used to characterize patients might not be completely uncorrelated, but we could design patients from these attributes using a modified conjoint design.

The original insight that drove prescribing simulation was that one could create a monadic sample, turn conjoint 90o (from products to patients) and, with the right analytics, develop a method of enduring value.

Now it is over twenty years after I came to this realization.  Prescribing simulation has been the hub of hundreds of studies and thousands of separate simulations. RG+A has run prescribing simulation studies in over 20 countries and just about every non-orphan disease category you could imagine. Simulations now incorporate electronic prescribing platforms and all forms of utilization controls, and are frequently linked to other kinds of simulations (contract simulation and patient decision tree simulation, to name two ).

And it all started with a verbal food fight. As the novelist Joseph Heller wrote, “Go figure.”