We recently made a £500,000 follow-on investment in Skin Analytics, who are world leaders in using AI to help clinicians diagnose skin cancer, and so I sat down a few days ago to write about the deal. Despite my massive and genuine excitement about this investment, I must admit that I’ve been putting off actually writing about it for a while. In these circumstances I find myself producing a sort of gushing boilerplate: Outstanding Management! Huge Market Opportunity! Game Changing Product! I do really think Neil and his team are world class, that they have the best evidenced and closest-to-market solution for AI dermatology, and that this could do a huge amount of good. But I always sound like just another VC shilling for his latest deal.
However I woke up to see that Google had released a skin condition AI diagnostics app. To be honest I’ve had better starts to the day, but at least it gives me a chance to write something less predictable – what it feels like when Google comes to eat your lunch.
As the day went on the board began to chat with management, and emails flew back and forth. The picture that emerged was, in the end, far more positive than we initially thought. Indeed if anything it left me with a higher level of conviction in the investment thesis than I’d had at the beginning. I wanted to expose the reasoning behind this to check if this is confirmation bias on my part; Am I being foolish for being encouraged by the entry of the world’s largest tech company, with the richest body of AI expertise, into the area that our small startup is working on?
Firstly, let’s look at what Google has actually produced: a free app that is aimed at consumers and for a broad range of 288 skin conditions. This is significantly different to Skin Analytics’ goal, which is to produce a clinical grade device that sells into hospitals and health systems with, for the moment, a focus on skin cancers and pre-cancerous lesions.
For Melanoma, which is far and away the most clinically important condition, Google’s published performance is pretty poor. It only catches 15% of melanoma as a primary diagnosis, and 60% as one of the top 3 diagnoses that their AI comes up with. This is a long way from being acceptable for clinical use. Skin Analytics achieves 95%.
One of the things that we worked hard to establish at initial investment was that what Skin Analytics (SA) had managed to do was hard to do, and hard to replicate. Although Google had a different goal in building an app for a broad range of 288 conditions, I can’t believe Google will have been thrilled with an app that put out numbers on this level. Although SA will have to continue to develop and improve if they are to maintain their edge, it does give one some comfort to know that they are ahead of the curve.
Secondly, there was no sign that Google is going after the clinical market for the time being. Developing what they have into a clinical product would take years, no matter how much money was invested into it. It would require a lengthy and fairly onerous prospective clinical study, which would take a year or two to complete and publish, as well as a host of other regulatory qualifications that are rightly associated with a medical device. It also requires software that integrates with clinical workflows, works in all clinical settings, on a range of devices, and offers the full feature set needed; a product as much as a technology.
Thirdly, an app like Google’s will likely generate a lot of demand for dermatology appointments, going into a service that is already considerably overstretched in most countries. There will be an increased need for a clinically reliable triage tool in primary care, which is exactly what Skin Analytics have.
So, if Google is going to enter your market area, it’s great if they do so by publishing evidence that they are not great at what you are very good at, and don’t seem to target the same customers you do, and are likely to increase demand for your product.
Of course there is risk that Google (or another competitor) catches up in performance and feature terms. SA will have to work hard to maintain their differentiation over coming years, improving the range and accuracy of their offer, and their product market fit. Further, there are still plenty of challenges for the company to overcome. But I’m somewhat more confident than I was two weeks ago that they can manage it, and a little nervous about being so!