Ultrahuman reduces return rate by 10x with the help of Memfault

Highlights:

• Premium user expectations across 100+ countries and 30 retail markets.
• 98% wear time means near-zero tolerance for device failures or downtime.
• Need to ship firmware fast while handling unpredictable real‑world conditions.

Ultrahuman’s audience expects top-tier reliability and responsiveness with the average wear time for the Ultrahuman rings hitting 98%. This level of expectation and reliance on the performance of the device raises the stakes. As Vatsal Singhal, Co-Founder of Ultrahuman puts it, “Our users expect a great experience. That means the app has to be 100%, the hardware has to be performant, responsive, and the battery life should be great.”

Ultrahuman’s devices are also deployed at a global scale with products sold in more than 100 countries, further complicating the monitoring and management of the device fleet. At this scale, things get unpredictable and shipping a bad update can have significant negative consequences.

At this scale, the Ultrahuan team realized that testing devices in the lab was never going to be sufficient to ensure the reliability of devices across such a diverse deployment. Vatsal notes the challenges of this type of deployment, “We have no idea what kind of Wi‑Fi people use. What kind of voltage fluctuations might happen. All of those scenarios that happen in the wild… If you can’t remotely understand it, there’s no way of solving the issue for them because we don’t have service centers in every location.”

Despite the complexity, the connected healthtech market is competitive, and to maintain an edge Ultrahuman knew they had to move faster than typical firmware teams.

“We believe in shipping fast… historically, firmware teams were shipping maybe once a month—we couldn’t run at this pace. And the tolerance for misses was minimal. We didn’t want even one crash to go unnoticed. The cost of failure was so high.”

To allow them to move fast, whilst maintain reliability and delivering a product people love Ultrahuman knew from early on that the status quo for firmware development wasn’t going to cut it. They needed near-real time visibility of their devices in the field and they had to be able to detect issues quickly, debug remotely, and keep releasing at speed without compromising user trust.

Highlights:

• Integrated early across Ring and Home devices to accelerate development and improve launch quality.
• Use Memfault’s observability as a core part of their ongoing development process, supporting rapid release cycles
• Make data driven product and engineering decisions based on data collected from hundreds of thousands of devices in the field

Memfault became Ultrahuman’s device observability layer enabling safe, fast releases and remote fixes the moment an update makes it onto end user devices. Ultrahuman adopted Memfault early in their product lifecycle, realizing that the data collected from devices would help them accelerate their pre-launch testing, as well as make life easier once devices hit the field. “We started chatting with Memfault when we were just doing our first version of the Ring and we started integrating and working with it right away.” They also wired it into new product lines from day one: “We just launched the home device and even on day one, we integrated Memfault into it.”

The concept of observability is well established across Ultrahumans engineering organization, and the embedded team is no different. As Vatsal put it, “I don’t know a single engineer who hasn’t implemented Datadog, or New Relic in their tech stack at some point. No one would deploy an application without observability. So why would you want hardware running in the dark? You need a Datadog for hardware, and that’s why you need Memfault.”

Memfault now forms a core part of the Ultrahuman engineering team’s daily routine and is deeply embedded in their release cycle, “The devs come in every day, they open the Memfault dashboard and check for issues or the Memfault Slack notifications might tell them something new has come up.” Even better, with the data Memfault provides, the team are able to very quickly solve the issue and ship a fix. “We have solved issues within a few hours, sent a custom patch and users didn’t even realize there was a problem.”

Memfault is a key tool for the team when deploying a new release. They ship fast, but when they roll out updates they push to only small subsections of the user base and use Memfault to monitor not just crash rates, critical metrics like battery dissipation, and performance deltas across cohorts and releases “as soon as the code hits production.” This allows the team to move fast and ship game changing features, without increasing risk.

With Memfault, Ultrahuman turned releases into measured, observable events supported by instant telemetry, rapid triage, and the flexibility to ship fixes to exactly the users who need them.

Highlights:

• Device replacement rates reduced by 10x.
• Firmware release cadence accelerated from monthly to weekly or faster.
• Resolved an issue that was increasing battery drain rate by 10%
• Crashes identified and resolved before users even notice

The combination of proactive monitoring, real-time crash analytics, and remote remediation translated into faster iteration and fewer disruptions for users. With Memfault in place, the team from Ultrahuman reduced the recall rate by over 10x during the lifecycle of their product, down to a fraction of a percent. Although hardware improvements contribute, Vatsal attributed a lot of this improvement to the ability to so easily identify and resolve software issues in the field before they cause issues for users.

“We were able to reduce our replacement rate dramatically, just by having deeper visibility into each issue. They have come down by 10x to a fraction of a percent. Of course, it’s not all software, but having visibility also helps us improve our hardware, so with every new batch the reliability gets better.” – Vatsal Singhal

In several cases the team used Memfault to identify and resolve problems including a buffer overflow during workout sessions, and an unexpected crash during the power saving “chill mode” that they identified and resolved without disrupting users. The team were also able to use the fleet wide data to identify a battery regression introduced in a release that was increasing battery drain by around 10%, and again, they resolved the issue remotely without disrupting users. 

Beyond hard metrics, the user experience improved and users can feel it, Vatsal notes, “I think it’s the kind of experience that users feel. They feel they are being listened to, and that we are actively resolving issues. Despite being a young company without the resources of others we are able to be highly responsive.”