The Science of Uncertainty
Adapted from “Forecasting for the Pharmaceutical
Industry” by Arthur G. Cook, Principal, ZS
How will sales for a blockbuster drug erode once a comparable (but cheaper) alternative hits the market? To what extent will increased competition impact biosimilar drugs in the pipeline?
Pharmaceutical companies constantly try to predict – or “forecast” – answers to these and similar questions; answers that affect crucial business decisions today. Financial groups make R&D investments based on anticipated sales for a new drug. Marketing teams forecast success ratios of various tactics when planning a go-to-market strategy. Members of the C-suite look to forecasts to provide accurate direction in product portfolio decisions.
Michael Schrage, an MIT Digital Business Research Fellow, wrote that “The most dangerous, hideously misused and thought-annihilating piece of technology invented in the past 15 years [is] the electronic spreadsheet. Every day, millions of managers boot Excel, twiddle a few numbers and diligently sucker themselves into thinking they’re forecasting the future.” He says that forecasting isn’t about predicting the future, but about embracing uncertainty – and that “if you want provocative scenarios…hire a science fiction writer.”
“Prediction is very difficult, especially about the future.” – Niels Bohr
While forecasting has become absolutely essential, the datasets that contribute to it have become immense and unwieldy. Just defining forecasting is itself a task. It is a picture of the future, a framework for interpreting events, an identifier of assumptions and choices, an aid in decision making, and more. But in all these definitions there are two basic concepts: the balance between user friendliness and technical complexity, and educating decision makers about the uncertainty in the forecast numbers.
What’s the job of a forecaster? To create a logical framework in which future events can be quantitatively evaluated – to create stories that paint a picture of the future.
Let’s look at an example. Consider three projects – A, B, and C – that forecast returns to the company, with the expectations shown in the figure below. Project C has the higher risk-adjusted net present value, so it looks like the better option. But what happens when we consider the uncertainty around each forecast, as shown in the next figure?
Project C becomes a high-risk decision, Project B is a low-risk decision, and Project A represents a more balanced risk profile. In this example, the decision maker can now incorporate risk in project selection as well as the expected net present value. The selection decision between projects has become more informed and well considered.
This example illustrates the forecaster’s role. Not simply to create a point forecast, but to communicate the “story” behind each forecast:
• What is leading to the uncertainty?
• Can it be resolved in the planning process?
• How were the uncertainty ranges determined?
• What is causing the value for Project C to tend towards the high side of the uncertainty range?
• What is driving the forecast for Project B towards the lower potential?
What questions do you have to ask?
Here’s a brief look at a representative case study. Let’s say PharmaCo is about to launch a product called Steadertin. Already included in their portfolio are Cytoflux, Chizophrin, Focalitine and Listromycin, among others. PharmaCo’s forecasting director, Sharon Spring, has been tasked with drawing up a forecast for the entire portfolio. Without getting too far down into the weeds, here is just a selection of the 34 questions I indicate that Sharon has to answer to develop a reasonable forecast:
Question 1: Demand sales for Cytoflux have been steadily increasing, while ex-factory sales have had spikes, dips and cyclical variation. What are some of the causes for this, and how can Sharon correct for it in her analysis?
Question 2: To what extent can PharmaCo influence or control these cycles, and how?
The job is not simply to create a forecast, but to communicate the story behind it.
Question 6: In a patient-based forecast, the result of a market-share calculation is the number of patients on each product. Why might we be interested in calculating days of therapy or TRxex by product?
Question 8: Calculate the minimum volume of product that must be produced to satisfy the demand suggested by average days of therapy per patient, overall compliance and resultant forecasted days of therapy.
Question 11: Despite a marked increase in the number of treated patients for Focalitine, the total market TRx has been declining. What are some possible explanations?
Question 14: How do you explain the difference between Focalitine’s TRx data and demand sales? Is this desirable? How might PharmaCo maintain (or reduce) this trend?
Question 18: Can cannibalization be a problem when a new formulation is launched for an existing product? How could it be potentially advantageous?
Question 23: Total drug treated patients for the entire portfolio are growing, but total diagnosed patients are decreasing. What possible explanations are there?
Question 28: Do you think it’s possible to score a product across too many attributes? At what point might increasing the number of attributes detract from the insight of the forecast.
And that isn’t even half of the list!
Where is the industry going?
In the history sidebar, I mention alternative futures. This is also referred to as “game theory” or “war games,” and allows a company to evaluate potential strategies. Your forecaster should be able to:
• Draw insights from historical data
• Capture expert judgment
• Evaluate these effects in structural modeling
• And, ultimately, couple identification of various scenarios (alternative futures) to help in decision-making
Pressure is incredibly high that predictions be accurate. There is no magic eight ball when it comes to making predictions, but today there are more thoughtful methodologies that help various stakeholders strategize efficiently, effectively and insightfully.
What are some of the challenges in the industry right now? Biosimilar versions of blockbuster drugs are about to make a grand entrance as nine of the 10 largest and highest revenue drugs go off patent in the next 10 years. Pharmacos are keen to learn a) what the financial losses will be once blockbusters lose that protection, and b) how they can protect their bottom lines when biosimiliars hit the market. Also, increased demand for orphan drugs (i.e., drugs that treat rare diseases) also spurs new challenges for the industry.
The orphan drug market is estimated to reach $127 billion by 2018, but this is still unknown territory for much of the industry. New competition and changing market dynamics associated with orphan and biosimilar drugs will challenge pharma’s conventional ways of forecasting future market share and adoption.
To prepare your own company for more intelligent forecasting, you need to research the most sophisticated tools, methods and analytics available to forecasters, and devise detailed approaches and algorithms that can be used in new product and in-market forecasting.
Forecasting today is necessary for strategic planning, business development and portfolio optimization. It has moved from a “numbers only” exercise to a decision-driven strategic tool. •
Arthur G. Cook, Principal, ZS
Arthur G. Cook has been involved in pharmaceutical forecasting for over 30 years, working with many major companies on their forecasting processes. He has created forecasts for over 150 therapeutic areas, and has written and presented widely on forecasting, portfolio management, market research and pricing. ZS is a global sales and marketing firm that specializes in a variety of industries, including the pharmaceutical industry. email@example.com
THE HISTORY OF PREDICTING THE FUTURE
What exactly is forecasting, and how can you be sure you’re doing it right?
Way back in the 20th century, pharmaceutical companies used to publish their projections for sales when they launched a product. There were some major blockbusters (Motrin, Naprosyn, Tagamet, Xanax), but also a lot of busts (you wouldn’t recognize the brand names, but they were from major companies like Merck, Sterling and SmithKline). What was most consistent about both categories, though, is that the projections were way off – in both directions – often by factors of 500 –1000%. The upshot was that the industry has become much more circumspect about estimates, largely because of the investors who got burned on the downside.
A lot has changed since then, but we are still in danger of thinking that the data alone can do the forecasting.
Forecasting is about alternative futures. It’s both a science and an art, and neither of those was in much demand with respect to marketing in this industry for a long time. We depended on innovative products and a robust marketplace to support our continuing growth. All that has changed, however. Here’s a snapshot of how and when that happened:
1970s AGE OF UNPLANNING– uncomplicated markets, not much strategic planning
1980S DELUSIONS OF GRANDEUR – burgeoning global growth, few restrictions
1990S AGE OF DISAPPOINTMENT – economic contraction, price and competitive pressures, need for more accurate planning and forecasting
2000S ERA OF REVITALIZATION – more sophisticated data and forecasting, better strategic planning
2010S BIG DATA RULES – explosion in data about patients, clinical trials, product development