How many TPPs? How about 1000?
The TPP illustrates a failure of the process of prediction
The TPP kills innovation in pharma. Not because of the TPP itself, but because of the quantity. 'The TPP' is set up to fail, because it is the product of a failed/ failing process.
If you generate one, you know it is imperfect, but it is there to represent a good guess about its future profile. If you would even generate two, you'd have alternatives with which to run your simulation. How about 1000? Consider Formula 1... 120 sensors on each car generate 3GB of data during each race - 1500 data points per second, or around 750 million data points during a 2 hour race. That helps a team like Aston Martin Red Bull generate 1000 prototypes between each race, or 30,000 per season.
Why so many? Well, simulation is critical to the way you learn how your prototype interacts with your model of the future. Now consider pharma's miserable solitary TPP - a terrible prediction of how the drug might work in one environment, interacting with terrible models of the future - built at a point when you know almost nothing about either. When it fails, and it does with statistical significance, you should look at this as a failure, not of this specific prediction, but how you're making predictions.
Seeing the TPP as a prototype helps illustrate the essential problem with the TPP process (I've written a lot about the TPP before... The Perennial Problem, An Enemy Of Innovation, Killing The TPP, and more...). With one prototype, you can find out that it breaks, but no more - let's call that 'failure'. As soon as you regard it as a process, rather than a finished product, however, you start to learn...
In Formula One, winning is as much about decision-making as it is about having the best drivers and vehicle on the day.
Aston Martin Red Bull Racing’s ability to optimise and dynamically adapt their decision-making in real time is almost impossible without data.
Strategy is the discipline of making better decisions, and Formula One has embedded learning into its decision approach - 'digitally-driven agility':
“Race teams are increasingly becoming DevOps businesses; organisations that have had to become more digitally-enabled through ever more stringent regulations and a corresponding reduction in the amount of time they had to innovate and test new designs.
“Many of the tasks that would have been laborious and time-consuming or based on the ‘gut feel’ of a senior engineer are now being automated or supported by simulations and analytics. That’s a transition many manufacturers have already been through or are currently working their way through.”
Is pharma easier than Formula 1? Absolutely not. A conversation I had with Red Bull a while back: if they see their models fail in the real world, they assume their models were wrong (which, oddly, wasn't a typical view in F1...). Even in F1, there is a lot to model - it is why they're moving rapidly toward 'digital twins'. However it pales into insignificance next to the complexity of biology and the real lives of humans. That doesn't mean that we should give up, it means we should try a different approach than the one we know doesn't work.