Last month, Oura launched its Nasdaq IPO, aiming to raise up to $2.2 billion at a valuation of roughly $15 billion. Eight days later it postponed indefinitely, citing uncertainty in the IPO market despite "strong demand."
It’s perhaps an unsurprising outcome, the inevitable result of a hyped consumer company meeting a nervous market reacting to sky-high AI valuations. Don’t get me wrong — Oura has built a real business. It’s profitable, projects about 90% YoY growth, and has about 5.7M paying members. People clearly want the product.
But what value are they actually getting from it?
Full disclosure: I don't wear an Oura or a Whoop. The only sleep tech in my house is an Eight Sleep, and I use it for only two things: to give my AC a respite during Miami summers, and for my wife who loves to put it on max chill while wearing three layers of clothes.
When I ask Oura and Whoop (and Eight Sleep) owners about their sleep data, they can often recite with incredible precision what their scores have been like the past week, what a given activity does to their score, and what data points they track each day. But when I ask them how they use the data and what changes it drives in their behavior, the apparent value becomes murky.
Most people fall into one of three camps:
1. They don't change anything. This is actually the most common, and often admitted directly. These folks seem to get comfort from seeing their data, in the same way that people like checking the weather in a city they aren't visiting. It’s interesting, but ultimately irrelevant.
2. They make minor changes. Notably, most of the folks were already planning a change or had begun to implement it. The data just gave them “permission” to do it. Nobody needs a ring to know that a third glass of wine will hurt their sleep, that consistent workouts will bring their resting heart rate down, or that a late night of work will leave them tired. The device may quantify the tradeoff (though the accuracy is debatable) but it doesn’t ultimately tell the person anything new.
3. They make major life changes. I’m sure these people exist, but I’ve yet to meet one.
THE PARKER PROBLEM
Wine has a version of this, and the difference is instructive.
A critic's score can be useful, but in theory it should only matter before you taste a bottle. It's directional, and if a critic's palate happens to line up with yours, it's a decent shortcut to navigate a wine list.
Where people go wrong is in trusting the precision. The same critic may score the same wine 94 and a 96 during the same tasting on the same day. Doubly so for different critics, or on different days. And once you've tasted the wine yourself, the score shouldn't matter at all. You've made your own call, and you know yourself better than anyone else.
A sleep score is an even weaker data point, because by the time you check it, you've already been through the experience. Checking your score at 7am is like rechecking the Parker number on a bottle you finished last night.
There's even a clinical name for the extreme version of this phenomena, "orthosomnia”: an intense anxiety about their tracker data made their sleep worse. It’s an example of a growing phenomenon where a device sold to improve an outcome, actually made the outcome worse.
The obvious counter is that 5.7M people are paying for it anyway. I don't think they're necessarily wrong to, but I think they're ultimately paying for comfort, enjoyment, and (in my wife’s case when she destroys my sleep score) bragging rights — ultimately something other than the outcome.
THE TEST
The test I use for data-centric products is simple: they're only valuable if they help you make a decision you couldn't otherwise make.
If you have a heart condition and a wearable flags an irregular rhythm before you'd ever notice it, that's immensely valuable. If it only confirms that you feel like shit because you didn't fall asleep until 3 a.m., it's an expensive dashboard.
It's the same question we ask in deal diligence when a founder claims a data moat. Collecting enormous amounts of information doesn't make it valuable unless that data lets you predict, automate or decide something that wasn't possible before.
Part of the problem with this category is what it does to the user. The default becomes checking the number instead of building a process and tuning into your own body. We tend to over-trust data in general, and we trust it most in exactly the places where it's least reliable.
Jeff Bezos has a story about this, where during his tenure Amazon's metrics showed customers reaching customer service in under a minute, but customer complaints said the wait was much longer. So in the middle of a meeting, he called the support line from the conference room and everyone sat and waited — surprise! it took 10 minutes to get a human on the line. His takeaway: when the data and the anecdotes disagree, the anecdotes are often right. Usually it means you're measuring the wrong thing.
Your body IS the anecdote. If you wake up feeling great and the ring says 71, the ring is the one that should have to explain itself.
WHAT I'M DOING
My wife and I took a few days in Grand Cayman and stayed at Palm Heights, It’s a boutique hotel known for its incredible gym and fitness classes. That’s not really my scene, but the service, food, wine, and rooms were all excellent. I did some scuba diving, and they took us through some pretty large reef canyons (30-50 feet from top to bottom) and the Balboa freighter wreck that is only accessible a few days each month due to shipping traffic.
WHAT I'M SIPPING
Tignanello, on the beach, paired with tenderloin. Yes, it's about as famous as an Italian wine gets, but when I’m abroad in a region not known for their wines, I’m more apt to go with something I know. And Tig never disappoints.
WHAT I'M TRACKING
Ascerta. Timely, given the Oura conversation. Companies are handing AI tools to employees faster than they can tell what it's worth. Ascerta started as an AI cost tracking tool, and will be increasingly focused on measuring which tools and agents actually pay off against the outcomes that matter to the company. The team is ex-Microsoft, and just raised $18M in a round led by Dell.
Etched. I’ve been following this company for the past year, given some of the early hype around the technology (specialized chips designed from the ground up for AI inference) and their uber-talented team (Harvard math whizzes). They raised at a $10B valuation in July, $21B in September, and now have offers at $40B to $50B, seemingly with sufficient commercial contracts to back the valuation. If the tech proves to be as innovative as they claim, $50B may end up being a bargain.
WHAT I'M LEARNING
AI is moving fast enough that even using it every day doesn’t feel like enough to keep up. So I’ve started working through an AI expert program—not because I’m worried that I’m behind (though I am), but because the gap between casually using these tools and actually understanding what they can do is getting wider by the month.
I’ll share more of what I’m learning in future issues. For now, I’ve been focused on one of the simplest ways to get more out of them: asking better questions via prompts that are actually optimized for the systems as they exist today. Here’s a simple framework I’ve been using that you’re welcome to use for your own work and life.
We're entering an era where almost everything that can be measured, will be measured. The scarce resource won't be data itself, it will be determining whether we can use the data to improve our lives — or whether it will simply empty our wallets and make us go crazy. Here’s to hoping it’s the former.
Cheers,
Greg