Why the Basis Peak failed: a four-thread post-mortem

The standard story is that the Basis Peak failed because it burned people’s wrists. That’s true, and it’s the part that made the news. It’s also the easy answer, and it does the actual history a disservice. The Peak was in trouble before the burn reports started arriving in volume, and the recall (formally CPSC notice 16-235, filed August 4, 2016) was the event that ended a product that several other forces were already pushing toward the exit.

Here is what we think actually happened, separated into four threads that ran in parallel from roughly mid-2015 to mid-2016.

1. Product positioning: stranded between two categories

The Basis B1, the company’s pre-Intel device, was a round plastic puck that looked nothing like a watch. It was a research-flavored product for people who cared about heart rate variability and sleep stages, and it didn’t try to sell itself on aesthetics. When the Peak launched in November 2014, Basis (now inside Intel) made a deliberate move toward the watch market: a square OLED display, a steel bezel, plastic-and-rubber straps, smartwatch-style notifications.

That was the right instinct in October 2014. It was the wrong instinct by April 2015.

The Apple Watch shipped on April 24, 2015, at $349 for the Sport model. It immediately defined what a smartwatch was supposed to look and feel like. The Peak, at $199 with a small monochrome OLED and no third-party app store, was not in that conversation. At the other end, the Fitbit Charge HR had launched in January 2015 at $149, with a smaller, lighter design, longer battery life on a single charge, and the Fitbit social layer that already had millions of users. The Peak sat between the two, more expensive than the Charge HR, less capable than the Watch.

Reviewers at the time picked up on this exact problem. The Verge’s 2014 review called the hardware “promising but unfinished.” Engadget liked the sleep tracking but flagged the laggy touchscreen. The Wirecutter, which was then still building its wearable category, did not recommend the Peak in its 2015 fitness tracker guide.

What the Peak was actually good at, continuous multi-sensor physiology with an emphasis on recovery rather than activity, had no obvious customer in 2015. The athletes were on Garmin. The general-fitness audience was on Fitbit. The early Apple Watch buyers were paying for the watch part. We wrote more about the comparison in Basis Peak vs Fitbit Charge; the short version is that Fitbit won the head-to-head on price, on battery, and on app polish, and the Peak’s better sensor stack didn’t have a story that explained why a normal buyer should pay extra for it.

2. Engineering: a thermal envelope with no margin

The Peak ran a Curie-class microcontroller, an optical heart-rate sensor, a galvanic skin response sensor, a thermistor, and a three-axis accelerometer, all inside a sealed stainless-steel case with a polymer back and a non-replaceable lithium-polymer battery. The case was not user-serviceable. There was no vent path. The metal back was the thermal exit.

For most of 2015 this worked. Then, in November 2015, Intel pushed a firmware update.

The update is the proximate cause most often cited by people who worked on or near the product. We have not seen the change log made public, and Intel has never named a specific software regression as the source of the burns. The pattern, though, is consistent across the Basis support forum posts archived on the Wayback Machine: complaints of unusual heat at the back of the case rose sharply in December 2015, peaked in February 2016, and continued through the corrective update in March.

What the firmware appears to have changed, and this is consistent with the design goal of fixing the Peak’s weak performance during high-intensity workouts, is how aggressively the optical sensor’s LEDs were driven. Driving them harder gives you a better signal-to-noise ratio when the wearer is moving and sweating. It also dissipates more power as heat, in a part of the device that is in direct contact with skin.

In a watch with a thicker case, a heat-spreading plate, or a replaceable battery you would have margin. The Peak had none of those. The same design choices that made it slim and water-resistant (sealed case, glued back, thin profile) removed every option the engineering team would have wanted when the heat problem appeared. Once the firmware was in the field, fixing it without bricking devices was hard; rolling back partially is what the March 2016 update tried to do, and that did not solve the problem.

The Peak’s industrial design and its thermal failure are the same story. The decisions that made it a credible-looking watch in 2014 made it impossible to fix in 2016.

3. Organizational: a startup absorbed into an unhappy division

Intel acquired Basis Science in March 2014 for a price widely reported as $100M–$150M. Intel does not confirm the figure. The team was roughly 50 people, mostly engineers and product staff based in San Francisco, and it was folded into the New Devices Group, an Intel division created in 2013 under Mike Bell to push Intel into wearables, smart eyewear, and other non-PC consumer hardware.

NDG was at a structural disadvantage inside Intel from the start. It was a hardware-launch organization in a company organized around silicon roadmaps. It had its own factory partnerships, its own retail channels, its own consumer marketing function, and a tempo measured in product cycles of 12 to 18 months rather than the multi-year fab cycles that drove Intel’s main business. Bell reported to Brian Krzanich directly, which gave the group political cover but didn’t give it the manufacturing or supply-chain integration that Intel’s silicon teams took for granted.

The Basis team kept most of its identity post-acquisition. The product roadmap was largely the one Basis had been building before the deal. The company’s San Francisco office stayed open. The engineering culture (fast iteration, web-style releases, a small ops team) was meaningfully different from Intel’s, and the seams showed in the Peak’s software update cadence and in the way the support organization handled the burn complaints. Basis support was not Intel support. The first wave of complaints went to a small team that had been a startup customer-success function eighteen months earlier.

Then NDG started losing political ground. In 2016, with the Peak in the middle of its recall and Intel’s smart-glasses project (the Vaunt) still unannounced, Krzanich pulled the plug on the wider NDG ambition. The group was wound down through 2016 and 2017. The Peak’s cloud shutdown on December 31, 2016 was the visible piece. The less visible piece was that the team capable of fixing the Peak was being dispersed across other parts of Intel, or out the door, at exactly the moment the recall needed them most.

We do not think the Peak would have survived even if NDG had been on stable footing. We do think a more integrated post-acquisition would have caught the thermal margin problem earlier, and would have had the operational depth to handle the recall without losing the brand entirely.

4. Market: the wearables category bifurcated in 18 months

The wearables market in 2014 was still a single category. By the end of 2016 it was three.

Apple captured the high end. Apple Watch unit sales in 2015 are still disputed (Apple does not break the number out), but most analysts put them between 8 and 12 million units, which was already larger than the entire dedicated fitness-tracker market the year before. The Watch defined a price ceiling, and a feature set, that made every “smartwatch-shaped fitness tracker” look underpowered overnight.

Fitbit captured the middle. By Q4 2015 Fitbit was shipping over 8 million units a quarter, with the Charge HR as its volume product. Its IPO in June 2015 valued it at over $4B. It had the social graph, the corporate-wellness sales channel, and a hardware cadence that produced a new device every six to nine months. The Peak’s slower release schedule meant it never got a second swing.

Garmin captured the athletes. The Forerunner 235, also released in 2015, brought continuous wrist-based HR to the running-watch crowd with GPS, battery life measured in weeks on watch mode, and the data-export options that competitive runners and cyclists wanted. The Peak could not match the GPS or the battery, and Garmin’s training-load math was a category of its own.

What the Peak was actually built to be, a multi-sensor recovery-focused device with a long charge cycle and a relatively passive interface, did not have a buyer in 2015. The shape of that buyer didn’t materialize until Oura found it five years later, with a ring that sidestepped the watch competition entirely and a subscription that paid for continuous software work. The Whoop 2.0, launched in 2016, made a similar bet on continuous physiology with a different form factor and a no-display design.

The Peak had the right idea about five years too early, in the wrong form factor, with no time to iterate. Reading the retrospective review now, what stands out is how many of its weaknesses were fixable with another generation, and how few of them mattered once the recall started.

What the cloud shutdown made worse

Once Intel decided to exit, the cloud shutdown was the cleanup step that turned the Peak from a discontinued product into an unusable one. On December 31, 2016, the Basis cloud went dark. The companion apps disappeared from the App Store and Google Play within weeks. Even users who had not received a recall refund and who still had a working device lost the ability to read their own data.

This is the part of the Peak story that we think matters most to current wearable buyers. Every modern wrist device is some version of this same arrangement: the hardware is sold, the value is on a server that the manufacturer can switch off. The Peak made this concrete in a way that hadn’t happened before in the category, and the lessons are still being relearned, most recently with the shutdown patterns we tracked in cloud-dependent wearables. If you cannot export your data on demand, you do not own it.

The one lesson the modern market still hasn’t learned

If we had to pick a single thread from the Peak that the current wearable category has not internalized, it would not be the hardware lesson. The thermal-envelope mistake is well understood now; modern wrist devices have more margin, better instrumentation, and faster over-the-air rollback. The recall mechanics are also better, with a more accessible CPSC database and matured processes for connected devices.

The lesson that is still open is this: a wearable’s data lifespan is shorter than the device’s physical lifespan, and the buyer almost never knows that at purchase time.

The Peak was sold in 2014 as a watch. It became a brick in 2016. That is two years of useful data life on a product that was implicitly sold as a multi-year purchase. Every current wearable, including the ones we recommend, is on the same arrangement: an Apple Watch is good for as long as Apple supports its watchOS version; an Oura ring works for as long as Oura keeps the cloud running and the subscription valid; a Fitbit works for as long as Google decides to keep Fitbit a product line. None of those companies prints the expected service life on the box.

The right response is not to stop buying wearables. It is to treat them as subscription products that happen to come with hardware, to keep your raw data exported somewhere you control, and to be skeptical of any device whose value disappears when the company changes its mind. The Peak is the worked example. Most of the industry has moved on without quite admitting that the same thing can happen again.