What is Asymmetric Learning?

Let's imagine that you and another team are given the same molecule to develop.

In scenario 1, you learn at exactly the same rate what it can do, what the market might want, and what might be approvable.

In scenario 2, you learn faster, and more creatively, about each of those parameters.

Scenario 1 would be symmetrical learning, and you'd expect to see similar performance from your launches. In Scenario 2, you may launch faster, or second-but-better, or you might end up in a different place entirely. That would be asymmetric learning - outperforming by design.

In pharma, the only difference between the molecule that leaves the lab and the one you launch is what you learn in between. If it is all 'learning', it makes sense to assess that discipline, which is why we're pioneering asymmetric learning in pharma.

All companies are learning organisations. In most, the company teaches the way it used to be done, or has always been done. In others, the company learns the ways it could be done. Amazon, Google? Learning organisations, by design - they learn continuously, every second, about what you want. Their product is learning. Using your behaviour to influence their product (recommendations, search results) is the epitome of a learning organisation.

If you did give the same asset to two different teams (and I am still to thoroughly understand why companies don't do this...), you would expect one to do better than the other. Why? Your first thought might be that one team had more 'talent' or 'knowledge' than the other, but the difference in performance will be in what they learn, about what they find out, about those three parameters (what the molecule can do, what the market might want, what might be approvable) - the collection of superior information on which to base decisions. You will also see that these two teams are having more fun than they used to when they were just feeding the PowerPoint beast machine - this growth mindset is what teams actually want, and why they'd stay up later and be more constantly inquisitive.

It is reasonable to try to improve our processes of prediction - it is a helpful exercise, but it rarely produces accurate results. 2020 has shown that the best machine learning, the best social media, the best rapid publishing and the best minds on the plant are unable to tell us which drug might work against a 'simple' virus. We have only learned by doing, and those who have organised their learning have learned the most. If you don't think that is true, ask 10 people whether they think remdesivir 'works', whether hydroxychloroquine 'works' and whether dexamethasone does - chaotic learning vs organised learning. If all of this knowledge cannot do that for a virus, how can it do so against more complex pathologies?

Consider: we do not know how valproic acid works. We know that it does. Key to it being useful to epilepsy was that the market asked questions of the molecule that the molecule could never have suggested.

The molecule is not the product. Finding a way to learn what is will be the key to your competitive advantage