Simulating success

Editor’s note: S. Kent Stephan is CEO of Princeton Brand Econometrics, a Princeton, N.J., marketing consultancy.

A recent article in the Wall Street Journal reported that the pharmaceutical industry spends $1.7 billion in research and development for every drug that is ultimately approved by the FDA for marketing in the United States. With such astronomical numbers, the stakes are very high for a drug to even recoup its investment.

Accurately forecasting whether a new drug will capture the loyalty of physicians is almost as difficult as developing it in the first place. The odds of even bringing it to market are extremely long. Forecasts of so-called experts are often wildly inaccurate, sometimes as much as several times more than the drug’s actual sales.

Today, most pharmaceutical companies conduct traditional market research before launching a new drug. But there is a better way. It’s called computer simulation or mathematical modeling. And, if done correctly, it can plot a roadmap to the greatest profit with the lowest risk - before the product is even launched!

New products have always been difficult to forecast, whether they are prescription pharmaceuticals, laundry detergents, toothpastes or candy bars. In industries such as consumer packaged goods, companies have used test marketing to reach go/no-go decisions and forecast national sales for over 100 years. They have sometimes used multiple test markets to help determine the combination of strategies and tactics that will produce the best results.

For obvious reasons, prescription pharmaceuticals can’t be test marketed. Test marketing requires that controls be placed on the retail distribution of the product, which is almost impossible for prescription brands. In addition, compared to consumer brands, prescription brands have very short life cycles. After spending more than $1 billion to discover and develop a new drug, a pharmaceutical company needs to start recouping its investment as soon as possible rather than delay the national launch to spend up to a year running tests in small, isolated pockets of the country.

A few decades ago, some packaged goods companies began to realize the value of mathematical modeling and the shortcomings of test marketing. Today, complex mathematical simulations have largely replaced test marketing in many of the leading consumer goods companies, and are becoming more popular with pharmaceutical companies who realize that accurate simulations may obtain for them what test marketing may not.

Instead of being able to test only one or two variables per market, practitioners can now test literally hundreds, even thousands, of different scenarios with computer simulations to ascertain which will be the optimal choices on all counts. Advantages include being able to learn:

  • what combination of marketing tactics will maximize profits for a given brand;
  • which advertising copy points will generate the largest market share;
  • what price will maximize profits;
  • how to optimize sales force productivity for greatest profits (i.e., the number of calls a sales rep must make to each doctor in his/her territory based on each doctor’s responsiveness);
  • what the growth potential or projected market share of a product will be over several years (this is important in determining how much money will be budgeted against a product).

How simulators work

Simulators are developed by measuring consumers’ responses to a description of the new product. A handful of highly-trained experts are then able to translate these responses into forecasts of the general audience’s reaction to the new product as it is promoted, not just hypothetically but in the real world. As such, simulators not only can forecast the impact that the brand message will have on people, but how this impact will play out at different levels of advertising and promotion, thereby enabling the marketer to choose the best mix. Interestingly, while both packaged goods and pharmaceuticals traditionally treat marketing and forecasting separately, an accurate forecast can - and should - determine what the marketing plan should be, since it can show which tactics are best to use and to what extent they should be used.

Simulators are used in the packaged goods industry for several reasons. First, they are much faster and cheaper than test markets. Second, they avoid exposing one’s new product ideas to the competition. Third, as previously mentioned, simulation makes it possible to generate forecasts for a limitless number of promotional plans for any tested message. Finally, packaged goods simulators usually produce forecasts that do a good job of estimating the launch year.

Simulation in the pharmaceutical industry actually has certain distinct advantages over its counterparts in consumer goods. First, a worthwhile prescription brand simulator produces national forecasts that are more accurate than test marketing.

Second, the forecasts change in response to very small changes in the tactical plan. Packaged goods simulators are far less sensitive to promotional plans. Therefore, a good prescription brand simulator is a more valuable planning tool than its consumer goods counterpart.

Finally, a good prescription simulator can create optimized solutions. It can be used to optimize the marketing message and the tactical (promotional) plan. For example, the simulator can be made to specify the least-cost tactical plan needed to achieve a feasible market share. Simulating the launch of a new prescription brand requires that two key elements be quantified:

1. The maximum market share which the brand could ever achieve if launched with the tested message. This is called the brand’s maximum potential share. It is specific to the tested message and may differ if the brand is described in another way.

A maximum potential share is the highest share the brand could achieve in the current competitive environment if it were promoted using every resource that could be effectively expended against the brand. It is a number that has been approached but never actually reached. In a world of diminishing returns, gaining the last few share points or fractions of share points costs far more than the potential gain.

A brand’s maximum potential share, unfortunately, is not a constant, even if the competitive environment doesn’t change. If a new brand is lightly promoted in a large, competitive category, its maximum potential share will be reduced.

2. The “uptake curve.” The uptake curve answers the question “How much of the maximum potential share will be achieved at any point in time, assuming a given amount of promotional effort?”

In the pharmaceutical industry, the primary vehicles for promoting to physicians are sales calls, product samples, journal ads, direct mail and special events. Special events include activities like seminars and group discussions. A high-quality prescription drug simulator accounts for each of these elements in fine detail.

The uptake curve enables a company to see how many prescriptions will be written month-by-month during the first year, if it implements a given tactical plan. Most importantly, it also enables marketers to work through scores of tactical scenarios before committing to an irreversible course of action in the real world.

The simulators that are used to forecast consumer package goods assume a universal uptake curve. Most of the time, this is a valid assumption. In the pharmaceutical industry, uptake curves can vary considerably. For example, one prescription brand might achieve two-thirds of its maximum potential as a result of only one sales call per doctor. Another brand might require a dozen or so sales calls to reach two-thirds of its potential.

Lost in translation

Since a product launch simulator is based on how doctors will respond to the brand’s message, this message must first be tested. This is usually done by exposing a carefully selected sample of doctors to the brand message via the mail or the Internet. If different messages are to be tested, different groups of doctors are exposed to each message.

After the doctors have read about the new brand, they then answer a battery of scaled questions. These questions include issues such as: How disappointed they would be if the product were not introduced? When would they most likely write the first prescription? How satisfied are they with current brands? What portion of their prescriptions for the medical condition would go to the new brand?

These questions are used when the brand is entering an existing category of products. If the new brand is entirely unique and will start a new category, a different set of questions may be used.

Real-world behavior

The key to building a simulator that will produce excellent forecasts is not as much the questions that are asked, but how the answers are translated into real-world behavior. The inability to translate how people say they will behave into how they will actually behave is a major cause of inaccurate forecasts.

Human beings have a strong tendency to overestimate. When it comes to neutral or positive behavior, not only do we overestimate our future behavior, we also overestimate our current behavior. The degree of overestimation of how frequently a product will be selected depends on how frequently its category is considered.

Consider this example. Assume that doctors say they will give a new brand 20 percent of their prescriptions. Empirical evidence from prescription prescribing data establishes that if they prescribe brands in the category every day, their 20 percent estimate will translate into a maximum potential market share of slightly less than 5 percent. If they only prescribe the category once a month, the new brand cannot achieve even a 4 percent share.

A good simulator will forecast filled prescriptions for a new brand during the first year with an average error of less than +3 percent for the strategic and tactical plan that was implemented. Forecasting errors of even 10 percent should be extremely rare, as the simulators help translate what people say they’ll do into what they’ll really do.

However, the primary value of a launch simulator may not be the accuracy of the forecast it produces but the marketing efficiencies it reveals. A good simulator will always be able to show a company how to produce better financial results (more sales or less spending, or both) than it would have on its own. After all, if you are in an industry that requires an average of $1.7 billion for each new product, you cannot afford to waste even a dollar because of bad forecasting.