Download the Visual
Five-year sales forecasts for new drugs rarely land close to actual sales
Sell-side analysts’ five-year sales forecasts for new drugs rarely come close. In a November 2023 Sleuth retrospective of select new drugs in the classes of 2013 through 2018, only seven of 60 forecasts had actual sales within 20% of the forecast, and several were off by more than a factor of four in either direction. Deals priced on out-year forecasts carry more risk than their models suggest, so in Sleuth’s view the better approach is to size bets to absorb large forecast errors.
Misses ran in both directions, and class totals hid them
Several of the biggest upside surprises were blockbusters. Merck’s Keytruda, in the class of 2013, was forecast at $1.0 billion for 2018 and sold $7.2 billion, 617% above forecast. AbbVie’s Skyrizi, class of 2018, was forecast at $1.4 billion for 2023 and was running at about $7.2 billion, and Sanofi’s Dupixent, class of 2016, sold $6.2 billion in 2021 against a $1.9 billion forecast.
On the downside, Gilead’s Sovaldi was forecast at $7.25 billion for 2018 and sold $250 million. Twelve forecasts show no sales at all, for drugs that never reached the market, including Merck’s anacetrapib, which appears in four classes, J&J’s sirukumab, in three, and Lilly’s solanezumab.
Class totals looked more reasonable, because a few big winners offset the failures. Actual sales beat the combined forecast in five of the six classes, by 21.5% (class of 2016) to 70.7% (class of 2014), and fell 12.6% short for the class of 2015.
An aggregate forecast can look accurate while the individual forecasts behind it, the level at which a single-asset deal is priced, are far off.
Deals built on out-year forecasts carry hidden risk
M&A that relies heavily on outer-year forecasts deserves caution. Bank of America data cited in Sleuth’s analysis showed that overly optimistic forecasts for several targeted oncology drugs pushed the implied out-year M&A multiple from about 6x to about 11x, and that even current-year forecasts for those assets missed by about 20% to 90%. Misses are likely larger for smaller assets with less analyst attention.
Acquirers burned by poor post-deal performance may deploy capital more cautiously, though companies facing large LOEs have little choice but to stay aggressive; in November 2023, Sleuth pointed to Merck and its Keytruda loss of exclusivity starting around 2028.
The flip side is that deep market knowledge can find bargains. In 2017, Horizon paid $145 million upfront for River Vision and teprotumumab, then forecast at $250 million in peak sales. The drug, Tepezza, later carried peak sales forecasts above $4 billion and was the centerpiece of Amgen’s $27.8 billion acquisition of Horizon, completed in October 2023. That $4 billion already looked optimistic in November 2023, and Amgen reported Tepezza sales of $1.85 billion in 2024 and $1.90 billion in 2025.
Forecast error cuts both ways, inflating some acquisitions and hiding value in others, so the edge lies in understanding a market better than consensus does.
Sleuth’s view: size bets for forecast error rather than predicting
In Sleuth’s view, single-number forecasts are often false precision. A better approach is to hold probability distributions across a wide set of outcomes and position to limit downside while participating in upside. The best deals are appropriately sized bets that can absorb a lot of forecast error, such as high-risk, high-upside bets or doubling down on proven winners: prepare rather than predict, an idea Sleuth credits to Nassim Taleb.
Sleuth applied the lesson to the class of 2023. Sell-side forecasts put 2028 sales for 10 of its drugs at $20.5 billion combined, led by Lilly’s lebrikizumab at $4.0 billion. In November 2023, Sleuth expected most to miss; the question was in which direction.
Forecasts are sell-side consensus estimates, not buy-side expectations built into stock prices; the class of 2018 uses a 2023 run rate as of November 2023. This page is a dated retrospective, with the Tepezza sales update current as of September 2026.
Frequently asked questions about drug sales forecast accuracy
How accurate are five-year sales forecasts for new drugs?
Not very. In Sleuth’s November 2023 comparison of sell-side five-year forecasts for select new drugs in the classes of 2013 to 2018, only seven of 60 forecasts had actual sales within 20% of the forecast.
Which drugs beat their sales forecasts by the most?
In Sleuth’s 2023 retrospective, Merck’s Keytruda sold $7.2 billion in 2018 against a $1.0 billion forecast, and AbbVie’s Skyrizi was running at about $7.2 billion in 2023 against a $1.4 billion forecast.
Which drugs fell furthest short of their forecasts?
Gilead’s Sovaldi sold $250 million in 2018 against a $7.25 billion forecast. Twelve of the 60 forecasts in Sleuth’s 2013 to 2018 comparison were for drugs that never reached the market, such as Merck’s anacetrapib and Lilly’s solanezumab.
What were the 2028 sales forecasts for the class of 2023?
As of November 2023, sell-side forecasts put 2028 sales for 10 class-of-2023 drugs at $20.5 billion combined, led by Lilly’s lebrikizumab at $4.0 billion and Biogen and Eisai’s lecanemab at $3.0 billion.
