The ins and outs of new product forecasting
Even for seasoned forecasters, when forecasting for new products, its worth it to take a step back and remind ourselves of the basics, says Miriam Stache, senior European demand forecasting manager, Lilly. Our approach, she says, is essentially determined by whether the market is underdeveloped or developed and whether our product is new or established. But everyones definitions of these terms are different.
Stache says the development stage of a market is highly objective and varies greatly even among the most seasoned marketers, while the difference between a new and established product is a little more cut and dry. Established products, she says, generally have at least 18 months of historical data behind them but not always.
Its not a science, Stache says. You might ask yourself questions like what physicians and patients really know about the disease, whether diagnostic tools/criteria are readily defined or available, if there is a lack of medical literature, sparse advocacy groups and uncertainty and new challenges with regulatory agencies to determine the state of the market and your product.
But when forecasting for new products, Stache says several factors make the situation special and must be taken into consideration:
You are new to the market and lack adequate knowledge about the therapeutic regimen
Your company needs to build new competency in a new therapeutic area
A significant level of uncertainty about expected sales exists
High levels of investment already lead to high expectations and a strong bias for forecasting toward targeting/planning (especially if the pipeline is thin!)
But high expectations she says, while a good thing to aspire too, must be weighed carefully against past performance of newly launched products and the realities of increasingly competitive markets.
Statistical versus judgmental methodologies
Statistical approaches to forecasting work well with established products in developed markets, Stache says, but a combination of statistical and judgmental approaches must be combined for established products in underdeveloped markets or new products in developed markets. And when forecasting for new products in underdeveloped markets, the only option is judgmental approaches, she says.
We must take our position into account when developing expectations for new products, Stache says. But the bottom line is that all approaches involve judgment and none of the situations is mutually exclusive. In fact, a forecast should integrate more than one approach.
But new product forecasting definitely relies more heavily on judgmental approaches. Some general guidelines, she says, include:
Use subjective assessment, qualitative data and rating schemes to devise quantitative estimates
Gather, in a systematic and unbiased fashion, all information and judgments relating to a forecast
Use when qualitative data and/or time is short
Depends on experience, intuition and imagination
Use when there is statistical uncertainty about the market or product or the forecast situation is very complex
Beware when intuition is used instead of analysis (when analysis is possible), data is based only on perception or memory and those whose judgment: is being used lack experience or appropriate intuition or imagination
Judgmental techniques
A variety of judgmental techniques are used by forecasters, Stache says. But two of the most common for new product forecasting are patient flow modeling and analogues or benchmarking. Patient modeling, she says, relies on going from market size into sales by considering relevant nodes, setting up the logical disease flow and modeling future market dynamics. Analogues or benchmarking can be used if a product is similar to an existing product on the market and is based on current market dynamics, Stache says.
There is basically a three-step process for patient flow modeling, she says. First, you must estimate market size. This begins by defining the relevant market and class by looking at factors such as incidence, prevalence and morbidity. And although assessing market size is not a difficult concept, success at this stage is very data dependent.
Its not an easy science, Stache says. It depends on where you get the data from and how reliable it is, whether it really reflects the realistic situation for your new product launch.
The second step is defining market share for your new product. Three key questions must be considered, Stache says:
What will be the share of market at peak?
When will peak share of market occur?
How will share of market reach peak?
Share modeling techniques include profile models referring to relevant key attributes, order of entry models, promotional models and elasticity models, Stache says but all focus on peak, time to peak and uptake. But market share has long been (and continues to be) driven by several key factors, including perceived efficacy, perceived safety, order of entry, total share of voice (or today, with channel mix share of noise) and perceived price level.
In fact, the third step in defining market share is assessing price in relation to major competition and other countries and considering the price perception among key customer groups and relevant discounting schemes.
Pitfalls of patient models for new product forecasts
Like all forecasting techniques, patient models have their pitfalls, Stache says. There are both demand and forecasting issues associated with the various stages of drug uptake and use that must be considered, she says.
Demand issues include considerations such as whether patients will perceive their condition as a problem that should be addressed, if they will seek treatment in the right place, how new treatments might be perceived by patients and payers compared to existing treatments, whether patients will properly comply with a new treatment and if they will attribute success or failure to the drug selection, Stache says. And forecasting issues (that are coupled with such demand issues) and that can throw off forecasts include:
Creating a new market is difficult and acceptance of new treatments is often overestimated
Its not just access to doctors that counts payment position is relevant (new drugs must pass muster with payers
How innovative is your new product really? Does it offer something existing products dont?
Compliance is often overestimated actual average length of therapy is usually less than estimated length of therapy
New product forecasting is a team effort, Stache says, that requires a multidisciplinary approach that includes testing sets of assumptions with experienced key stakeholders. For the most success, Stache suggests keeping it simple to get everyone on board, identifying promoters to support you over time, and proving your reliability by showing the accuracy of past new product forecasts.
The impact of new product forecasts on long-term strategy is being magnified, she says, by more volatile healthcare environments, more complex and unpredictable pharma supply chains, the difficulties in revising significant investments and the influence of new product forecasts on decisions such as merger and acquisition activity and head-count strategies.
With every new molecule that fails, the pressure for remaining molecules in the pipeline increases, Stache says.
The stakes are high. So Stache says you should: rely on experts in your organization, integrate valuable sources and use multi-method approaches, stand up against target-driven high level plans, and track/monitor new product forecasts during the launch phase. At the same time, she suggest you should avoid: sitting in an ivory tower to build one-person forecasts from scratch, choosing complex models in an attempt to cover all possible dynamics and expecting the same forecast accuracy as for established products.
The bottom line: forecasters must realistically set expectations for their organizations, and not fall victim to forecasting to try to meet those set by others, Stache says.