The science
We spent eighteen years
making stress measurable.
Then we found out measurement was only a third of the problem. Here is what we learned about the other two, and the research behind all of it.

18years
of NIH funded research
100days
of the MOODS validation study
10%
drop in stress intensity, self initiated
2
NIH funded national centers
The problem
Steps work. We wanted to know why stress never did.
If you are a thousand steps short at 6 PM, you know what to do about it. You trust the number, you already believe activity matters, and the fix is obvious. Three things have to be true before a measurement changes what someone does, and steps has all three.
Stress had none of them. If people do not trust the number, nothing happens. If they trust it but do not see why it matters, nothing happens. And if they trust it and care about it but it arrives when there is nothing they can do, still nothing happens.
We had to solve all three. Measurement took eighteen years. The other two took us longer to even see clearly.
Eighteen years, and we published the failures too.
Nothing else works until the measurement is right, so we started there. We were among the first teams the NIH funded to detect stress from wearables in real life rather than in a lab. Plenty of groups took the problem on. Not many stayed with it long enough to get it right, and every dead end we hit is in the peer reviewed record.
Eighteen years in, the accuracy was finally good enough to change behavior, on the ordinary smartwatches people already own. We proved it in a nationwide study.
One thing about how we ran it. Most stress tracking research trains a model and tests it on the same data afterwards, which tells you the model fits that data and not much else. We deployed ours pre trained, with no fine tuning on the study at all. If it worked, it would work for anyone, unchanged. We called the study MOODS, the Mobile Open Observation of Daily Stressors.
Then something happened we had not designed for. Across a hundred days, the hundred and twenty two participants simply seeing their own data started changing their behavior on their own, in fourteen distinct ways, and their stress came down: ten percent less intense, and ten fewer stressors a month. We did not expect that. Most stress interventions do something in the moment and nothing that lasts past the study.
The model went into the study pre trained, with no fine tuning on study data. Whatever accuracy it showed is accuracy anyone can have.
CuesHub's model matched self reported stress about as closely as cortisol, the stress hormone, does.
The papers underneath
- Momentary stressor logging and reflective visualizations: implications for stress management with wearables
Neupane, S., Saha, M., Ali, N., Hnat, T., Samiei, S. A., Nandugudi, A., et al. (2024). CHI Conference on Human Factors in Computing SystemsCuesHub team
- Feasibility and utility of AI triggered prompts for efficiently capturing when and why people experience stress in natural environments
Neupane, S. (2025). Doctoral dissertation, University of MemphisCuesHub team
- Pulse-PPG: An open source field trained PPG foundation model for wearable applications across lab and field settings
Saha, M., Xu, M. A., Mao, W., Neupane, S., Rehg, J. M., & Kumar, S. (2025). Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 9(3)CuesHub team
- cStress: towards a gold standard for continuous stress assessment in the mobile environment
Hovsepian, K., Al'Absi, M., Ertin, E., Kamarck, T., Nakajima, M., & Kumar, S. (2015). ACM International Joint Conference on Pervasive and Ubiquitous ComputingCuesHub team
- Continuous inference of psychological stress from sensory measurements collected in the natural environment
Plarre, K., Raij, A., Hossain, S. M., Ali, A. A., Nakajima, M., Al'Absi, M., et al. (2011). ACM/IEEE International Conference on Information Processing in Sensor NetworksCuesHub team
- AutoSense: unobtrusively wearable sensor suite for inferring the onset, causality, and consequences of stress in the field
Ertin, E., Stohs, N., Kumar, S., Raij, A., Al'Absi, M., & Shah, S. (2011). ACM Conference on Embedded Networked Sensor SystemsCuesHub team
We had to make the cost of stress something you can count.
Almost everyone treats stress as the price of doing well. It is also a driver of chronic disease, but that damage is slow and invisible, so the connection never feels real. We needed a way to make it real, and we found one: your heartbeats.
Your stress response was built to save your life. It raises your heart rate and pushes energy to your limbs so you can fight or run. It still does that today. But you are in a chair, nothing uses that energy, and it turns into low grade inflammation instead, which is a known path to chronic disease.
That rise in heart rate is what we measure as Workload Heart Rate. Here is the part that changed how we think about it. You get a lifetime budget of heartbeats, and spending it slower means living longer: five beats per minute off your resting heart rate is worth about a year. Spend heartbeats on exercise and you get them back, because your heart runs slower the rest of the day. Spend them on stress and they are gone. Exercise is an asset. Stress is a liability.
It costs you today as well. Under stress your brain is busy calming your nervous system, and it will take the decision that makes the stress stop over the decision that is right for you or your team. You have less patience with people. If you lead anyone, that is not a private problem, and it is why our executive users check they are calm before something that matters.
Every five beat per minute reduction in resting heart rate is associated with about one more year of life.
Heartbeats spent on exercise are recovered later in the day. Heartbeats spent on stress are not.
The papers underneath
- Association between change in heart rate over years and life span in the Paris Prospective 1, the Whitehall 1, and Framingham studies
Gaye, B., Valentin, E., Xanthakis, V., Perier, M. C., Celermajer, D. S., Shipley, M., et al. (2024). Scientific Reports, 14(1), 20052
- Balancing exercise benefits against heartbeat consumption in elite cyclists
Van Puyvelde, T., Janssens, K., Spencer, L., D'Ambrosio, P., Ray, M., Foulkes, S. J., et al. (2025). JACC: Advances, 4(10 Part 2), 102140
- Using minute ventilation for ambulatory estimation of additional heart rate
Wilhelm, F. H., & Roth, W. T. (1998). Biological Psychology, 49(1-2), 137-150
- The neuroendocrinology of stress: the stress related continuum of chronic disease development
Agorastos, A., & Chrousos, G. P. (2022). Molecular Psychiatry, 27(1), 502-513
- Inflammation: the common pathway of stress related diseases
Liu, Y. Z., Wang, Y. X., & Jiang, C. L. (2017). Frontiers in Human Neuroscience, 11, 316
- Impact of changes in heart rate with age on all cause death and cardiovascular events in 50 year old men from the general population
Chen, X. J., Barywani, S. B., Hansson, P. O., Ostgard Thunstrom, E., Rosengren, A., et al. (2019). Open Heart, 6(1), e000856
- Elevated resting heart rate, physical fitness and all cause mortality: a 16 year follow up in the Copenhagen Male Study
Jensen, M. T., Suadicani, P., Hein, H. O., & Gyntelberg, F. (2013). Heart, 99(12), 882-887
The obvious ways to help make it worse.
This part is harder than it looks. Tell someone they had a stressful day and you have added disappointment to a bad day. Interrupt them while it is happening and you have broken their work and probably made the situation worse.
That is why just in time intervention has not worked for stress, and adding large language models to it did not fix anything.
So we built around the thing that did work. Reflection is what moved people in MOODS, so reflection is the base. On top of it goes a predictive nudge carrying your coach's own words, sent before a hard moment while you can still do something about it, rather than in the middle of one when you cannot.
Only one in five detected stress events actually warranted an intervention at all.
Seeing their own data led the 122 MOODS participants to start fourteen kinds of behavior change on their own, and brought stress intensity down ten percent.
The papers underneath
- Wearable meets LLM for stress management: a duoethnographic study integrating wearable triggered stressors and LLM chatbots for personalized interventions
Neupane, S., Dongre, P., Gracanin, D., & Kumar, S. (2025). Extended Abstracts of the CHI Conference on Human Factors in Computing SystemsCuesHub team
- Sense2Stop: a micro randomized trial using wearable sensors to optimize a just in time adaptive stress management intervention
Battalio, S. L., Conroy, D. E., Dempsey, W., Liao, P., Menictas, M., Murphy, S., et al. (2021). Contemporary Clinical Trials, 109, 106534
- Momentary stressor logging and reflective visualizations: implications for stress management with wearables
Neupane, S., Saha, M., Ali, N., Hnat, T., Samiei, S. A., Nandugudi, A., et al. (2024). CHI Conference on Human Factors in Computing SystemsCuesHub team
Model lineage
Five results, fourteen years apart.
Every milestone links to its paper. Read them rather than take our word for it.
AutoSense
The first wearable sensor suite able to infer the onset, causality and consequences of stress in the field rather than the lab.
Read the paperContinuous inference
The companion result: psychological stress inferred continuously from sensor measurements collected in the natural environment.
Read the papercStress
A gold standard for continuous stress assessment in the mobile environment, deployed in longitudinal studies with thousands of participants.
Read the paperMOODS
The hundred day nationwide study that validated the model in the field, and found that seeing the data changed behavior on its own.
Read the paperPulse-PPG
An open source, field trained PPG foundation model for wearables, benchmarked across both lab and field settings. It powers CuesHub's engine on the watch.
Read the paperWhere the work was done
Two national research centers. One director.
CuesHub’s CEO directs both MD2K and mDOT. The research behind CuesHub was done there, in public, funded by the NIH and the NSF.
NIH funded national center advancing biomedical discovery and health through mobile sensor big data.
NIH funded national center for mobile and wearable health technologies and digital biomarkers.
Bibliography
Everything cited, in one place.
Twenty papers: the eighteen the app itself cites in the educational panel behind every number it shows you, plus the two that report the MOODS study directly.
- 1The neuroendocrinology of stress: the stress related continuum of chronic disease development
Agorastos, A., & Chrousos, G. P. (2022). Molecular Psychiatry, 27(1), 502-513
- 2Everyday stress components and physical activity: examining reactivity, recovery and pileup
Almeida, D. M., Marcusson-Clavertz, D., Conroy, D. E., Kim, J., Zawadzki, M. J., et al. (2020). Journal of Behavioral Medicine, 43(1), 108-120
- 3Sense2Stop: a micro randomized trial using wearable sensors to optimize a just in time adaptive stress management intervention
Battalio, S. L., Conroy, D. E., Dempsey, W., Liao, P., Menictas, M., Murphy, S., et al. (2021). Contemporary Clinical Trials, 109, 106534
- 4Impact of changes in heart rate with age on all cause death and cardiovascular events in 50 year old men from the general population
Chen, X. J., Barywani, S. B., Hansson, P. O., Ostgard Thunstrom, E., Rosengren, A., et al. (2019). Open Heart, 6(1), e000856
- 5Seeking positive strengths in buffering athletes' life stress burnout relationship: the moderating roles of athletic mental energy
Chiou, S. S., Hsu, Y., Chiu, Y. H., Chou, C. C., Gill, D. L., & Lu, F. J. (2020). Frontiers in Psychology, 10, 3007
- 6AutoSense: unobtrusively wearable sensor suite for inferring the onset, causality, and consequences of stress in the field
Ertin, E., Stohs, N., Kumar, S., Raij, A., Al'Absi, M., & Shah, S. (2011). ACM Conference on Embedded Networked Sensor SystemsCuesHub team
- 7Can allostatic load cross over? Short term work and nonwork stressor pile up on parent and adolescent diurnal cortisol, physical symptoms, and sleep
French, K. A., Smith, C. E., Lee, S., & Chen, Z. (2025). Journal of Applied Psychology
- 8Feasibility and utility of AI triggered prompts for efficiently capturing when and why people experience stress in natural environments
Neupane, S. (2025). Doctoral dissertation, University of MemphisCuesHub team
- 9Co variation of fatigue and psychobiological stress in couples' everyday life
Doerr, J. M., Nater, U. M., Ehlert, U., & Ditzen, B. (2018). Psychoneuroendocrinology, 92, 135-141
- 10Association between change in heart rate over years and life span in the Paris Prospective 1, the Whitehall 1, and Framingham studies
Gaye, B., Valentin, E., Xanthakis, V., Perier, M. C., Celermajer, D. S., Shipley, M., et al. (2024). Scientific Reports, 14(1), 20052
- 11cStress: towards a gold standard for continuous stress assessment in the mobile environment
Hovsepian, K., Al'Absi, M., Ertin, E., Kamarck, T., Nakajima, M., & Kumar, S. (2015). ACM International Joint Conference on Pervasive and Ubiquitous ComputingCuesHub team
- 12Elevated resting heart rate, physical fitness and all cause mortality: a 16 year follow up in the Copenhagen Male Study
Jensen, M. T., Suadicani, P., Hein, H. O., & Gyntelberg, F. (2013). Heart, 99(12), 882-887
- 13Brief review on physiological and biochemical evaluations of human mental workload
Lean, Y., & Shan, F. (2012). Human Factors and Ergonomics in Manufacturing & Service Industries, 22(3), 177-187
- 14Inflammation: the common pathway of stress related diseases
Liu, Y. Z., Wang, Y. X., & Jiang, C. L. (2017). Frontiers in Human Neuroscience, 11, 316
- 15Momentary stressor logging and reflective visualizations: implications for stress management with wearables
Neupane, S., Saha, M., Ali, N., Hnat, T., Samiei, S. A., Nandugudi, A., et al. (2024). CHI Conference on Human Factors in Computing SystemsCuesHub team
- 16Continuous inference of psychological stress from sensory measurements collected in the natural environment
Plarre, K., Raij, A., Hossain, S. M., Ali, A. A., Nakajima, M., Al'Absi, M., et al. (2011). ACM/IEEE International Conference on Information Processing in Sensor NetworksCuesHub team
- 17Pulse-PPG: An open source field trained PPG foundation model for wearable applications across lab and field settings
Saha, M., Xu, M. A., Mao, W., Neupane, S., Rehg, J. M., & Kumar, S. (2025). Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 9(3)CuesHub team
- 18Wearable meets LLM for stress management: a duoethnographic study integrating wearable triggered stressors and LLM chatbots for personalized interventions
Neupane, S., Dongre, P., Gracanin, D., & Kumar, S. (2025). Extended Abstracts of the CHI Conference on Human Factors in Computing SystemsCuesHub team
- 19Balancing exercise benefits against heartbeat consumption in elite cyclists
Van Puyvelde, T., Janssens, K., Spencer, L., D'Ambrosio, P., Ray, M., Foulkes, S. J., et al. (2025). JACC: Advances, 4(10 Part 2), 102140
- 20Using minute ventilation for ambulatory estimation of additional heart rate
Wilhelm, F. H., & Roth, W. T. (1998). Biological Psychology, 49(1-2), 137-150
Working on biosignals?
We publish, and we collaborate. If you are running a study that could use continuous, field grade strain measurement from consumer wearables, we would like to hear about it.
hello@cueshub.com