Hidden Value in the R&D Pipeline: Solving Pharma's $500 Billion Problem with Machine Learning

The $500 Billion Problem

In 2020, the pharma industry invested over $500 billion in R&D, M&A and licensing in search of the next life-changing and market-leading products¹. But a recent Evaluate analysis found that just 20% of marketed drugs generate 90% of commercial returns. This suggests that pharma's investment in novel molecules may not always deliver the returns expected, and raises questions about what can be expected from this latest round of R&D and portfolio investments.

Blockbuster drugs, defined as those generating over $1 billion per year in sales, require the right combination of transformative science, unmet need, competitive positioning and pricing. However, building accurate forecast and valuation models based on these complex factors is both time intensive and difficult to replicate at scale, particularly for strategy, commercial and business development teams operating on tight timelines.

To overcome these challenges, pharma often relies on third-party valuations from equity analysts or consensus forecast providers. However, these sources do not always offer coverage for early-stage assets. A recent analysis of equity analyst forecast availability, shown in the graph below, revealed that less than half of blockbuster products launched in the past decade had forecasts available in Phase II. For non-blockbuster products, just 16% were covered in Phase II.

How the Early-Stage Information Gap Limits R&D Productivity

Initiating equity analyst coverage in late-stage development makes sense; there is little point in creating detailed forecast models for products that may never reach the market, or that are not likely to contribute to company revenues within the forecast horizon. But this tendency also means that pharma is often without independent, readily-available valuation data to inform or gut-check their pipeline decisions.

With the bulk of R&D spending required in Phase III, plus the higher price tags that later-stage assets command in licensing or M&A transactions, earlier insight into product value could dramatically increase the productivity of R&D investments. A separate analysis by Evaluate, shown in the graph below, found that growth in R&D spending from 2012-2020 outpaced growth in sales revenue, even as the number of novel drugs approved steadily increased over this same period of time. This suggests that pharma may be investing their R&D budgets in products with limited commercial potential.

What's needed is a new approach to forecasting the commercial potential of early-stage assets that reduces the need for complex manual forecasting, allows products to be valued at scale and delivers the speed and accuracy needed to make time-sensitive licensing or R&D portfolio decisions with confidence. 

Filling the Gap with Machine Learning

Machine learning algorithms are already used across the pharma industry to generate new insights from complex datasets². By applying similar technologies to commercial forecasting, data providers can analyse millions of product-level data points to identify the product characteristics that most directly correlate to commercial performance, as well as forecast each product's potential based on its unique combination of attributes. 

By removing the time and capacity limitations of human forecasters, machine learning can deliver commercial forecasts at a scale not previously available, providing a more complete view of commercial potential across all phases of clinical development, including the early-stage or privately-owned drugs that are not covered by equity analysts or consensus models.

This technological breakthrough allows pharma teams to identify drugs with blockbuster potential as early as Phase I, so they can better direct R&D investments towards the pipeline products most likely to generate significant returns, or make more informed licensing and acquisition decisions. In particular, machine learning allows pharma teams to:

• Reduce the likelihood of over-investing in products with limited commercial futures
• Better prioritise portfolio drugs with market-changing potential
• Quickly and accurately quantify the value of external assets to make rapid go/no-go decisions
• Identify potential portfolio gaps for where additional R&D spending, licensing or acquisition is needed

The return from leveraging machine learning correctly could be substantial. Further analysis from Evaluate suggests that better focusing R&D investments on products with the greatest commercial potential could add as much as $50 billion in market value in the US alone. Companies eager to capture their share of the potential value created would be wise to take advantage of these new tools.