Until recently, I didn’t know a thing about how my roughly 25-minute bike commute across San Francisco—or any other part of my day, really—affects my body, other than that I inevitably arrive at work sweaty and a bit out of breath when I’m in a big rush. How high is my heart rate? Do my sleep habits affect it? How many calories do I burn?

These questions have been on my mind as a number of activity trackers and smart watches have hit store shelves over the past couple of years, promising to track information like steps, sleep, heart rate, sun exposure, and calories. With one of these sensor-filled gadgets on my wrist, surely I could get accurate information about my body.

That’s the idea, at least. These devices could give you more control over your health by making it easier to collect data previously left unmonitored or, as in the case of heart rate, typically gathered only at a doctor’s office (and even then infrequently). And these devices aren’t just tracking data; companies like Apple, Jawbone, and Microsoft offer advice based on what the sensors in their wrist-worn wearables detect. The Microsoft Health app should soon have the ability to compare calendar or contact information with the Microsoft Band’s assessment of, say, your heart rate or skin conductance level—a measure of your skin’s ability to conduct electricity, which tends to climb with stress.

The Apple Watch and Microsoft Band use optical sensors to measure heart rate. The Jawbone Up3, which instead tracks your resting heart rate, uses bioimpedance sensors and several electrodes to measure your skin’s resistance to a small amount of electrical current. These sensors and others in the bands are adequate for measuring routine activity levels, but is the technology really accurate enough to turn wearable devices into digital medical tools?

“We’re at an inflection point, or transition, from lifestyle health stuff to medical metrics,” says cardiologist Eric Topol, a genomics professor at the Scripps Research Institute and a fan of digital health technology. To Topol, the objective is clear: devices that accurately measure vital body signs and even monitor serious health problems like diabetes and heart disease. “It’s the medical metrics where accuracy becomes fundamental,” he says.

The test

How far away are we from such wearables? I tested the accuracy of a few wrist-measured metrics, including heart rate. For several days, I wore an Apple Watch and a Microsoft Band while biking to and from work. I also wore a Polar H7 Bluetooth chest strap, which is one of the most accurate consumer devices for measuring heart rate. Results varied, and sometimes they varied a lot. The Band’s average heart-rate measurements were consistently closer to the results of the Polar chest strap—sometimes within a beat or two per minute, but they could be as many as 13 beats off. The Apple Watch, meanwhile, gave readings as many as 77 beats per minute different from the Polar device.

Measurements of calories burned (something all three bands, including the Up3, track) were also somewhat inconsistent; on one morning commute, for instance, they ranged from 143 to 187. Altogether, the experience was a far cry from the vision of these devices as digital sages drawing deep, accurate insights from the data they collect, helping doctors diagnose ailments, and eventually, perhaps, even predicting health problems or detecting them before they become serious. These are hard goals to achieve, for several reasons. While the wrist seems like a great place to start with sensing on the body, and we’re used to adorning it with watches and jewelry, it’s tricky to make a comfortable, good-­looking device that can stand up to all kinds of daily abuse.

And since everyone’s body is different, the wrist is not always a great spot to take accurate measurements. “You can make millions of smart watches that are identical, but you have millions of people who are not identical. It’s really hard to find something that’s robust across all these people,” says Chris Harrison, an assistant professor of human-computer interaction who leads the Future Interfaces Group at Carnegie Mellon University.

Harrison and other experts say arms that are too hairy, sweaty, fat, or thin can make it hard to get a good reading from today’s optical heart-rate sensors, which read blood flow in the wrist. Tattoos can pose a problem, too—as Apple points out on a support page for the Apple Watch, noting that the ink can block light from reaching the sensor. “All of a sudden that translates to thousands of users out there, all of whom are going to be unhappy and say it doesn’t work because it doesn’t work for them,” says Christian Holz, a researcher on human-computer interaction at Yahoo Labs who focuses on the miniaturization of mobile devices.

Beyond workout tracking

There’s hope for wearable devices that actually take the types of measurements that would be helpful for health monitoring. But realizing that hope will probably mean moving on to radically new technologies. And it will certainly mean developing devices that are able to take a wider variety of measurements.

At Quanttus, a startup in Cambridge, Massachusetts, researchers are building a wrist-worn device to track heart rate, respiration, and blood pressure by way of ballistocardiogram, which uses a sensor to measure the itty-bitty movements of your body every time your heart pumps blood. At a conference in late April, cofounder and CEO Shahid Azim said the company is interested in releasing “some number” of wristbands by the end of the year. Cofounder and chief scientific officer David He says it is still “refining the technology.”

Once we can pin down heart-rate and blood-pressure measurements, He believes, we may well be able to monitor most cardiovascular vital signs with wearables. This could be a boon, not just for fitness applications and those wanting to keep an eye on their own health but also for doctors who want noninvasive ways to keep tabs on patients at a level currently possible only in a hospital.

Another Cambridge-based startup, Empatica, is creating a wristband that measures jumps in skin conductance to figure out when the wearer is having a seizure, so it can alert someone to check on the person. Empatica isn’t able to predict seizures yet, however, and it hasn’t released its product either.

Building these products takes lots of time. Testing, simulations, modeling, prototyping, and problem-solving are all more extensive when you need to make sure the devices can stand up to the requirements of daily wear, such as frequent exposure to sweat and water. That’s a lot more than you’d normally expect from your electronics. But if companies clear these obstacles, being able to sense things like blood pressure and skin conductivity continuously can also open the door to quantifying stress and mood, since they will make it possible to collect data about your body in all kinds of situations.

And we’re just at the early stages of understanding how much we may be able to do with sensors on the skin. In the next several years, noninvasive sensors may become useful for other biometrics that can currently be tracked only with invasive processes. It might be possible to monitor blood glucose with skin readings rather than a needle prick—something that would be helpful to people with diabetes.

In fact, researchers are working on this particular problem at the University of California, San Diego. They’ve developed a temporary tattoo, printed with electrodes and coated with an enzyme solution, that can measure glucose levels. Joseph Wang, director of UCSD’s Center for Wearable Sensors, has been working on the technology for the past five years. He says it will be at least another two years before it is commercialized—initially, he expects, in the form of a single-use temporary tattoo, and then with tattoos that can measure the wearer’s glucose every 20 or 40 minutes for a day or a week. Topol believes that all kinds of accurate data are coming; it’s just a matter of time. “We have a ways to go, but ultimately, that is something machines are really very good for,” he says. “And the algorithms can be developed where for each person it could be a virtual medical assistant.” Given that today’s wristbands still stutter when measuring heart rate during a workout, such applications seem far out of reach. But the research at Quanttus, Empatica, and UCSD suggests that new approaches based on technologies far beyond conventional optical sensors could finally turn wrist-worn devices into tools for monitoring general health.