Forecasting Market Research: Ensure reliable forecasts by using appropriate methodology and realistic discounting

This is the second issue of a two-part series on forecasting market research. It highlights methodology considerations and issues related to discounting assumptions for forecasting models. The June issue provided an overview of product sales forecasting concepts.

As outlined in the previous issue of this series, healthcare companies depend on reliable product forecasts. Yet with so many variables influencing the industry, accurately predicting performance is challenging. A variety of tools, including sophisticated forecasting models, are designed to simulate the real world to accurately predict product sales; but they are only as good as the data that go into them. Accurate inputs, or assumptions, are everything.

Methodology

The first step to achieving valid assumptions for a forecasting model is to conduct market research using appropriate methodology. At the most general level, there are two forecasting approaches: market-based and patient build-up. Picking the right one for your brand depends primarily on its lifecycle and marketplace.

• Market-Based models predict how a new brand will perform based on prescription data and other information for comparable products already in the market. Knowing how to use existing information to accurately forecast your product’s performance requires familiarity with the therapeutic category and marketplace. Consider differences in promotion strategies, demographic shifts, patent expirations, off-label uses, regulatory changes, and reimbursement issues to develop your forecast.
• The Patient Build-Up approach predicts performance for first-to-market products by building a likely market size and scenario. It begins by identifying the number of patients with a particular condition, and it narrows the market from there—step by step—by estimating population fall-off at each point in the treatment process. For instance, a large number of people may have a disease, but only a percent may be aware of it. Less will visit a doctor for the condition, and fewer still will actually fill their prescription. An evaluation of unmet needs in the market and knowledge of events and changes that are expected to occur in the forecast period are important components of this approach.

Physician interviews are used for both models. To minimize respondent fatigue, try to limit the number of product profiles/ scenarios to be assessed. Use the best, most likely, and worst profiles, and ask physicians to determine if the last patient treated for a particular condition would have received each product. This encourages physicians to make use predictions based on actual patients rather than hypotheticals. For this reason and others, this interview format often achieves more accurate information than other methods.

Discounting Data

No matter how you approach your forecasting study or how sophisticated your data is, discounts need to be applied. Discounting creates a more realistic picture about the market by managing exaggerated predictions of use by physicians. For the following reasons, physicians do not typically provide accurate estimates about future use:

• Product profile information may be incomplete; it may not encompass all factors that influence a decision to prescribe.
• All patients who are candidates for “product x” may not have the opportunity to receive it, due to reimbursement policies and other real-world issues. It’s also likely that many patients are stable on a medication and not actual candidates for a change in therapy.
• Physicians may over-estimate their patient population as a result of the survey exercise, during which an unusual amount of attention is focused on a particu