Why I wouldn’t invest in consumer health hardware

Yesterday’s post on Apple’s health strategy solidified a suspicion I’ve been harboring for awhile, which is that the future is going to be rough for connected health and fitness devices — Peloton, Eight Sleep, Withings, Mirror, WHOOP, etc.

To be clear, this could take a long time, and connected health devices could be great businesses in the meantime. But I think the long term is bleak.

Here’s why:

Somebody, most likely Apple, is going to build the OS of health data, and once they do, the health device makers will have no choice but to integrate with it. This will modularize them, the same way Facebook and Google modularized publishers. 

How does this happen? The OS will define some sort of standard data format, and the device makers will be forced to conform to that standard if they want to reach end users. For example Apple defines a variety of standard health characteristics (blood type, DOB, “fitzpatrick skin type”), activity types (step count, swimming stroke count, flights climbed), and vital signs (heart rate, oxygen saturation, body temp) in their health API. If you’re creating a device and have a slightly different way of doing things for some reason, well, too bad. Now you’ve got to choose between distribution and optimization.

This is the interesting thing about the “finding power” framework. It posits that for every interface between two layers of the value chain, one side must be suboptimized and conformable (in this case, the health devices) in order to allow the other side to be optimized and integrated (in this case, the health OS that I imagine Apple is building). 

The reason the OS has all the power is because they control the critical choke point that determines the overall system performance for the end user. In simpler terms: I’m betting that people care about the things the OS can do more than the things each individual medical device maker can do.

Currently all these devices operate almost entirely independent of each other, and their data lives in silos. What will most people value more — incremental improvement on the devices we currently have, or all the data from these devices feeding into a unified space where our healthcare providers can use it to offer better care? I think the latter.

This is bad news for the device makers, because once I have a “good enough” smart bed, scale, exercise bike, and so on, then I’m deciding more based on price or convenience than I was before. All the data is standard (and therefore commoditized), and feeds into the same place, so it’s easy to switch. It’s like the TV business — even if technologically there is a lot of complexity needed to achieve performance, the profit gets competed away pretty quickly if the main thing people care about is somewhere else in the value chain.

The same thing goes for primary care medical providers, like One Medial, Forward, and Q.bio. They’ll be kind of like hotel chains — able to charge a premium if they serve wealthy consumers with a premium experience, but with cost increases that are roughly proportional to value increases, and, even worse than hotel chains, utterly dependent on the health OS layer.

There is one big exception: what if one of these providers actually creates a highly differentiated, patentable scientific advance that everyone else in the ecosystem comes to rely on? The prime example of a company like this is the role Intel played in the 90s PC ecosystem.

In theory, anyone could build a CPU chip that integrated with Windows, so it should have been one of those “modular, conformable, suboptimized” businesses. But in practice one of the main problems with computers was how slow they were. And Intel had the biggest R&D budget and best talent to solve the problem. So everyone relied on their chips, and conformed to the standards they specified in order to use them.

In healthcare, I could imagine a similar business cropping up that has an incredible machine learning algorithm to take in a bunch of data and spit out a warning anytime the person should see a doctor or go in for a test, if a certain risk factor for a certain condition was a bit too high. This feels like the kind of problem that an Intel-like company could grow off of. But it would still be dependent on the OS layer.

The weakest part of this whole argument, of course, is that Apple is still a loooooong way off from monopolizing healthcare. Nobody is anywhere close. It’s a wide open opportunity! There’s a $1b dollar bill on the ground!

Is your spidey sense tingling? Yes, it should be.

These types of opportunities are usually captured before they become apparent to most people. So if the healthOS opportunity is real, we need to have a specific theory about why it doesn’t already exist. It can’t be the case that nobody has bothered to even try yet.

In fact others have tried. Google and Microsoft both built short-lived database products for medical records, but they required a lot of manual work for users to upload their information, and there wasn’t much you could do with it once you bothered to upload your stuff. These products were launched in the late 2000’s and killed in the early 2010’s. 

What’s different now is that Apple is integrating directly with healthcare providers to make the upload process automatic, and there’s a whole new category of data from connected devices that only came into existence in the recent past.

So, perhaps now is the time. But is Apple the company to do it? It’s hard to be sure. Google certainly seems like a good candidate to tackle the problem, but Apple’s privacy focus gives them an edge.

Regardless, whoever ends up building the OS for health, the impact will be the same on consumer health devices: near certain commoditization.

That’s why I wouldn’t invest.

Would you? (Leave a comment!)