In December 2024, Oura introduced Symptom Radar, a new feature available to Oura Ring Gen3 and Oura Ring 4 members with an active subscription. Powered by an advanced algorithm years in the making, Symptom Radar helps identify early signs of strain in a member’s biometrics, empowering members to take proactive steps toward rest and recovery.
Oura partnered with external partners, such as the University of California, San Francisco (UCSF), to analyze two years’ worth of anonymized member data, ensuring that Symptom Radar can detect signs of strain accurately and precisely, up to two days before a member selects an illness-related tag using Oura’s Tags feature.
Read on to learn more about the development process and the science behind the feature.
March 2020: TemPredict Study Launches
As the COVID-19 pandemic began, the University of California, San Francisco (UCSF), launched TemPredict, a study that tested whether physiological data collected by Oura Ring, combined with responses to daily symptom surveys, could predict illness symptoms.
TemPredict enrolled more than 63,000 participants, including 3,180 healthcare workers at university hospitals across the country. The aim of the study was to build an algorithm to identify patterns of the onset, progression, and recovery from COVID-19.
Results published in 2022 found that multiple biosignals from Oura Ring (heart rate, HRV, respiratory rate, temperature, and activity) could collectively detect signals associated with the onset of COVID-19 “an average of 2.75 days before participants got diagnostic testing.”
October 2020: Rest Mode Launches
When someone is under physiological strain, we know that rest and recovery should be the main focus, rather than aiming for optimizing sleep or reaching an activity goal. In late 2020, Rest Mode rolled out to all Oura members, allowing members to focus on rest to help their bodies to recover if they’re feeling ill, tired, or overworked.
For as long as Rest Mode is on, a member’s Activity Progress Goal, Activity Score, and all activity-related contributors will be disabled. Readiness and Sleep insights will also be adjusted to help prioritize recovery. Note that Oura continues tracking your biometrics and trends so you can prioritize your health without losing your historical data.
READ MORE: How Sleep Affects Your Immune System
2020-2024: Development of the Advanced Computation Model (V2)
From 2020 to 2024, Oura’s Science and Data Science teams continued to develop V2 of our computational model, designed to be more accurate and responsive than the original model.
“One of the main motivations was leveraging a larger dataset, especially from several years worth of anonymized member-tagged illness data,” says Ketan Patel, Senior Manager, Machine Learning Data Science at Oura. “This allowed us to improve detection performance and incorporate predictive capabilities—alerting members about something straining their body one to two days in advance of symptoms.”
As more members joined Oura and used Oura Tags, the number of tags related to symptoms of illness has reached 3 million. Using this data, Oura’s Science and Data Science teams were able to fine-tune the original algorithm for gains in accuracy and responsiveness.
Compared to the initial computational model, the Science and Data Science teams made several additional improvements to the now-named Symptom Radar V2, including:
- Improved responsiveness: Symptom Radar V2 is designed to be more responsive to signs of strain and recovery, and tends to detect signs of strain earlier than V1. Conversely, V2 will show “no signs” after recovery more quickly than Symptom Radar V1.
- Additional context: If the model flags that a member has signs of strains, Symptom Radar V2 identifies potential reasons, such as elevated heart rate or temperature trends, based on deviations from the member’s baselines over time. This gives insights into why the member may receive a notification of “minor” or “major” signs of strain.
- Cycle awareness: The new model also incorporates phase data for cycle awareness for members who are opted into Cycle Insights. “We’ve calibrated the algorithm to provide similar experiences for different demographic groups, adjusting thresholds accordingly,” Patel explains. “This ensures users have consistent results regardless of their phase or demographic category.”
- Runs locally, not in the cloud: The updated algorithm runs on a member’s device—i.e., smartphone—instead of in the cloud, like the previous iterations of the model. This means results are available faster and without a network connection, so Oura members never miss a day of data.
April 2024: Symptom Radar Available for Testing in Oura Labs
In April 2024, Symptom Radar was introduced in Oura Labs. When opted in, members could test the feature and provide direct feedback to developers.
As members tested the feature, the development team listened to the feedback and iterated on the features. For example, members reported that they were unclear as to whether their Symptom Radar result had been updated, especially in cases where an elevated risk was not detected. To solve this, we added a historical graph so members can see that the feature has been updated each day.
Additionally, members wanted more information on which biometric changes indicated risk. With the new Symptom Radar feature, Oura provides each biometric input that contributes to their risk level, so members can see which markers changed and better understand why the risk increased.
December 2024: Symptom Radar Feature Launches to All Members
Just in time for cold and flu season, Symptom Radar has fully launched to Oura Ring Gen3 or Oura Ring 4 members with an active subscription.
Now, Oura will alert members when their body is showing signs of strain, allowing them to detect potential illness—up to two days in advance of experiencing symptoms. We hope this will help our members identify a potential oncoming illness and take steps to rest and recover, so disruptions to their daily lives can be minimized.
RELATED: 9 Tips to Boost Your Immune System & Stay Well This Winter