Distributed by Design
The Future Of Wearable Biometric Sensors
Liked this piece? If so, you’ll want to check out other articles in Decoding Biology’s Wearable Technology & Biometrics collection. Additionally, you can show your support by tapping the 🖤 in the header above. It’s a small gesture that goes a long way in helping me understand what you value and in growing this newsletter. Thanks!
The largest and most influential wearable technology companies have sold us a delightfully palatable fiction — that a single device worn on the wrist can provide comprehensive insights into human physiology. For instance, Whoop claims they can provide users with personalized insights to "improve performance, build healthier habits, and extend healthspan with continuous health monitoring." A bold claim for a device with two sensing modalities1.
The implausibility of the all-in-one wearable device becomes immediately apparent when you consider how our bodies generate and broadcast information, and how these signals vary in their electrical, chemical, and mechanical properties. No single sensing technology can capture this diversity of signals, nor can they all be measured in the same spot on the body.
Evolution —the relentless optimizer— distributed our bodies sensory apparatus and organ systems across our anatomy for a reason, creating a complex and diverse web of communication. The brain, heart, and muscles generate electrical potentials, chemical signals propagate through the blood and interstitial fluid, and tissues transmit and sense mechanical forces. Yet, we still have companies like Whoop stating that its wrist band— measuring heart rate, HRV, and blood oxygenation— is the "most personalized and powerful health monitoring wearable” available.
Consider by contrast how we monitor complex engineering systems like rockets. NASA’s Artemis II space launch system and Orion spacecraft collectively house ~6,100 distributed sensors— accelerometers measuring vibration, thermocouples monitoring temperature gradients across engine components, pressure transducers tracking propellant flow, strain gauges detecting structural deformation, gyroscopes monitoring orientation, and more. These sensors aren't arbitrarily scattered. They’re strategically positioned exactly where the physical phenomena of interest occur.
More importantly, these sensors don't operate in isolation. Their data streams are integrated through control systems that establish causal relationships between measurements2—distinguishing when an engine temperature spike causes a pressure change versus when both are responding to a third variable. This distributed, integrated approach enables not just monitoring but prediction, identifying cascade failures before they lead to catastrophic events.
The irony is striking — we apply sophisticated distributed sensing networks to monitor machines, but expect simplistic single-point measurement tools to reduce the infinitely more complex human body down to a few key metrics. An aerospace engineer would never accept monitoring a launch vehicle's performance using only sensors attached to the nose cone, yet we've collectively embraced the notion that a wrist-worn device can comprehensively monitor our physiology.
Reconstructing the Physiome — A Multi-Sensor Imperative
The physiochemical constraints imposed by different biological signals make the distributed multi-sensor approach not just preferable but necessary for understanding human physiology and providing personalized health insights.
For instance, measuring muscle oxygenation requires mNIRS sensors to be placed directly on large muscle bellies— like the outer quadriceps— where the optical pathway traveled by photons can interact with the tissue of interest. Electrocardiography on the other hand necessitates electrodes placed on the chest where they can capture the heart’s electrical vectors, while electroencephalography requires that electrodes are placed on the scalp in alignment with the various cortical regions.
Attempting to measure muscle oxygenation proximal or distal to a working muscle is like trying to photograph the night sky through a neutral density filter. The signal exists, but the perspective fundamentally limits what can be captured. Similarly, measuring the hearts electrical activity too far at the periphery sacrifices signal quality and information content compared to chest placement3.
The pioneers of modern genomics didn’t revolutionize DNA sequencing by trying to read the entire genome from a single point. Instead, they developed distributed approaches that read fragments from multiple locations and then computationally reconstructed the whole genome. Similarly, comprehensive biological monitoring requires distributed sensors whose data can be computationally integrated to reconstruct the whole physiological state (ie, the physiome).
Centralized Sensing — Convenience Over Truth
Companies like Whoop are built on a compromise, placing all their sensors on the wrist not because it offers superior insights but because it’s the most easily accessible form factor4. This single design decision comes with significant tradeoffs — wrist-worn devices are limited in the sensing modalities they can employ and come with strict size, shape, and power constraints. This comprise would be acceptable if the product was marketed ethically5. Instead, these companies make expansive claims about comprehensive health monitoring that simply cannot be fulfilled through single-point sensing technologies.
Designing a wearable that can be worn on the wrist does however lend itself to a number of convenient form factors, which in part explains why most wearable devices occupy this niche. Wrist worn wearables also accommodate a number of tried and true sensing modalities— including PPG, EDA, and IMUs — for measuring heart rate and blood oxygenation, skin conductance, and motion. However, they miss important signals from major muscle groups, visceral organs, and metabolic pathways. As a result, trying to assess holistic health with these devices is like attempting to predict global weather patterns using only temperature readings from Manhattan —you'll capture some useful data, but will miss most of the picture.
The irony is that consumers often purchase devices like Whoop seeking comprehensive health insights while being unaware of the fundamental sensing limitations. Marketing claims about "complete health monitoring" from a single wrist-worn device represent a well-intentioned simplification at best and a biological impossibility at worst. The most sophisticated algorithm cannot extract information from signals that aren't being captured in the first place.
Network Biomarkers — Relationships Matter More Than Components
A distributed sensor architecture doesn't just increase measurement diversity—it enables a conceptual leap from individual biomarkers to network biomarkers, from isolated metrics to relationship-focused models that better reflect biological reality.
Consider the history of genomics. The initial human genome project delivered a parts list—a catalog of genes that, while valuable, failed to explain most diseases or traits. The real revelations came later through network approaches examining how genes interact through regulatory networks, how expression patterns change across tissues, and how environmental factors modulate gene expression. Similarly, in physiology, isolated metrics from single-point wearables— like heart rate, muscle oxygenation, HRV, etcetera— represent a parts list without the crucial relationship information.
The history of medicine itself reveals the limitations of singular biomarkers. Imagine measuring a diabetic patient's blood glucose at 127 mg/dL on three separate days. Despite having three clinically identical readings, the patient experiences wildly different subjective states—sluggishness on Monday, mental clarity on Wednesday, and jitteriness on Friday. The glucose metric, our clinical North Star, suddenly seems like a compass that points in different directions depending on the weather.
What explains these contradictions? The missing element is relationship data—how glucose interacts with cortisol, how inflammation affects insulin sensitivity, how sleep quality modulates metabolic response. These relationships cannot be captured with single-point measurements because they involve multiple biological systems distributed throughout the body. Only a coordinated sensor network can capture the inputs necessary to construct these network biomarkers.
This shift from singular biomarkers to networks mirrors a broader intellectual transition happening across disciplines from genomics to ecology. In each field, we've discovered that emergent properties arise from interactions rather than individual components. Just as the wetness of water exists nowhere in the individual H₂O molecule, overtraining syndrome in athletes, for example, emerges not from any single physiological metric but from the disrupted relationships between autonomic, endocrine, and neuromuscular systems.
From Rocket Science to Human Physiology
Distributed sensor systems generate orders of magnitude more data than single-point measurements, creating both new challenges and opportunities. The challenge is in making sense of multiple data streams. The opportunity lies in extracting relationship information invisible to single-point measurements and their associated analytical techniques. To analyze biomarker data from distributed sensor networks we can borrow three mathematical frameworks— derivative analysis, information theory, and causal inference— developed for applications ranging quantitative finance to telecommunications and space flight.
Derivative analysis examines how numerical values change over time rather than just their absolute magnitudes. In rocket telemetry, engineers monitor not just temperature but its rate of change— acceleration— and higher-order derivatives to identify anomalies before they become catastrophic. When applied to physiological data, this approach can reveal phase transitions invisible in raw measurements—the early stages of physiological stress before heart rate increases or the subtle changes in recovery patterns that precede overtraining.
Information theory provides tools to quantify how systems influence each other. Measures like transfer entropy can determine which physiological system is driving changes in another—an important question for understanding causal hierarchies in biological networks. When respiratory rate, oxygen consumption, cerebral oxygenation, and heart rate increase during exercise, transfer entropy can reveal which is driving the other, information that correlation alone cannot provide.
Causal inference methods go beyond correlation to establish directed relationships between variables. Approaches like Granger causality test whether past values of one time series help predict future values of another beyond what its own past values predict. Applied to physiological data from distributed sensors, these methods can map causal networks—which measurements drive others under specific conditions—just as they map which sensor readings predict impending failures in rocket launches.
Learning from Integrated Systems
Transitioning from single-device to distributed sensor systems presents formidable but solvable challenges. Again, other fields offer instructive parallels. Modern aircraft incorporate over 1,000 distributed sensors monitored through integrated systems that balance measurement precision with operational reliability. Space suits used by astronauts include multiple distributed sensors monitoring everything from pressure to temperature to carbon dioxide levels, all integrated into cohesive systems that flag anomalies without overwhelming the wearer with data.
The practical implementation of distributed biological monitoring will require innovations in three areas: sensor miniaturization, wireless communication, and integrated analysis. Sensor miniaturization continues to advance rapidly, with devices like continuous glucose monitors shrinking from cumbersome external devices to subcutaneous implants smaller than a grain of rice. Similar miniaturization trends are emerging for electromyography sensors, blood analyte detectors, and muscle oximeters. These advancements will enable placement of specialized sensors precisely where biology demands without creating unacceptable user burden.
Wireless communication protocols have matured significantly, with low-energy Bluetooth and similar technologies enabling seamless data transfer between distributed devices. The challenge isn't technology but standardization—creating open protocols that allow sensors from different manufacturers to communicate through centralized hubs, similar to how smart home devices from different companies can interact through platforms like Amazon’s Alexa.
Integrated analysis represents perhaps the greatest opportunity. Just as genomics moved from sequencing to functional analysis, biological monitoring must progress from data collection to relationship extraction. This transition will require algorithms capable of extracting patterns from multiple data streams combined with physiological knowledge to interpret those patterns in biologically meaningful ways.
The Future of Distributed Sensing Is Here — It’s Just Not Evenly Distributed
The future of biological monitoring lies not in better versions of the current smartwatches and wristbands on the market, but in distributed systems that match our biology's inherent architecture—specialized sensors placed precisely where they can capture the highest signal-to-noise ratio for their target phenomena, communicating through integrated systems that extract relationship information invisible to single-point measurements.
This future of distributed network physiological sensing isn’t far in the distance6. The sensor technologies exist today, the wireless communication protocols have matured, and the computational power is readily available. What's missing is the conceptual integration—the recognition that the future of biological monitoring lies not in better individual wearables but in smarter networks of specialized sensors interpreted through sophisticated analytical frameworks.
Companies like Whoop have played an important role in familiarizing consumers with biological monitoring, but their single-sensor approach represents a starting point rather than the destination. The truly revolutionary advances will come from systems that embrace biology's distributed nature rather than fighting against it—systems that monitor muscle oxygenation where muscles are located, brain activity where the brain is located, and cardiac function where the heart is located, all integrated through sophisticated mathematical frameworks that extract the relationship information containing the most valuable biological insights.
The human body isn't a collection of isolated measurements that can be captured from a single location—it's a complex, interconnected network of systems distributed throughout our anatomy. Our approach to biological monitoring should reflect this reality. By embracing distributed sensor systems, network biomarkers, and advanced mathematical analysis, we can finally begin to read the flood of signals our bodies broadcast every moment of every day.
Interested in working together? I advise small companies, startups, and VC firms on topics ranging from biosensor development, multiomics and biometric data analysis, network modeling, and product strategy. Contact eva♦peiko♦@gmail.com (replace the ♦ with n) for consulting inquiries or newsletter sponsorship opportunities.
Healthspan refers to the portion of life spent in good health, free from chronic diseases and disabilities that reduce quality of life. By definition, proving an intervention extends healthspan requires longitudinal studies spanning at least 10+ years to detect meaningful differences in health trajectories. Given that the Whoop 1.0 was release in 2015 and the concept of healthspan has only recently gained mainstream attention through advocates like Peter Attia, I find the claim that Whoop helps extend healthspan highly suspect, even before we consider the technological limitation of single-point sensing (as discussed in this piece).
The lack of casual-linkage in time-series physiological data from wearables is one of the biggest opportunities in my opinion. It’s also a problem I’ve been working on with PhysioNexus, which uses Granger causality (open-source version) and transfer entropy (limited-access) to map both linear and non-linear cause-and-effect relationships between physiological variables like heart rate, tissue oxygenation, blood lactate, etcetera.
This is not to say that you cannot get an accurate heart rate measurement with a smart watch or ring. However, these devices use a different sensing modality entirely— photoplethysmography— to measure changes in blood volume and pulse pressure, which could then be used to calculate heart rate.
I’ve written about the respective tradeoffs associated with form-first and function-first approaches to wearable device design in a previous piece, The Garden Of Technological Possibilities.
Ethical marketing, in this context, would mean promoting a product in a socially responsible way, with transparency, and avoiding misleading claims. Telling consumers your wearable device can quantify their biological and track their pace of aging is unethical in my opinion.
Companies like Oura have already started moving in this direction by itegrating data from constant glucose monitors into their platform. This transition from single to multi-point sensing represents a schism opening up between companies like Oura, which are embracing distributed network sensing, and companies like Whoop who are doubling down on their single-point sensing technologies, making progressively bolder claims about its capabilities.