Predictions and productions

We have a lamplight problem.

The only drugs in our pipelines, certainly by the time they hit phase II, are those with a positive NPV. At many companies now, ‘forecasts’ that don’t exceed $2bn will see a molecule eliminated from a portfolio review.

Given the 5 to 10 years from that forecast to it being proven unreliable in real practice, we need to realise that the pile of molecules we discarded may well contain more gems than the ones we kept.

This physics thought experiment is telling:

(Ted Chiang)

Given how relatively predictable billiard ball movement is, versus pharmaceutical markets, we have to accept that all of the numbers we see for pharma are 15 or 50 collisions into the future. These drugs are already under the light of the lit streetlamp, not the rest of the street that is in the dark. Attrition rates, and phase transition rates alike, we’re only studying those drugs that once had a positive commercial analysis sitting behind them. And we know they are wrong…

Phase I cannot be a time for accuracy, but a time to see the whole street - where a drug might launch versus how much it might make, on average, or in one pre-selected indication. Aiming for accuracy will not give us accuracy, but may lose our opportunity to get it right.