Skin temperature on wearables: what the number actually means

The temperature graph in your wearable’s app is one of the most misread numbers in consumer health tech. People see a degree or two of variation and either panic about a fever or assume the sensor is broken. Neither reaction is right, because the number on the screen is not what most people think it is.

Your wrist tracker is not a thermometer. It is a peripheral skin-temperature sensor doing trend analysis, and the difference matters every time you look at the chart.

What the sensor is actually doing

Almost every wearable that markets a “temperature” feature uses a small thermistor (a thermally sensitive resistor) pressed against the skin on the underside of the device. The Apple Watch Series 8 and 9 use two of them, one on the case back and one inside the device, so the algorithm can subtract ambient drift. Oura uses a single thermistor on the inside of the ring band, in contact with the finger. Fitbit’s Charge 5 and 6 and the Sense 2 use a wrist thermistor. Whoop 4.0 uses a wrist thermistor. Withings’s ScanWatch 2 uses what the company calls TempTech24/7.

These sensors are sampling skin temperature, not core body temperature. Skin temperature is what your skin feels like. Core body temperature is what a clinical thermometer reads under the tongue, in the armpit, or rectally. The two are correlated. They are not the same.

A few uncomfortable facts about peripheral skin temperature:

• It is usually two to four degrees Celsius lower than core temperature.
• It tracks ambient temperature heavily. A 20°C bedroom and a 26°C bedroom produce visibly different wrist readings.
• It tracks peripheral blood flow. When your blood vessels dilate (after a hot shower, a glass of wine, an evening workout, or just because the room is warm), warm blood moves outward and the wrist reads higher. When they constrict (cold room, anxiety, caffeine), the wrist reads lower.
• It varies by site. A finger reading at night runs a degree or so warmer than a wrist reading on the same night, because the hand under a blanket is in a different microclimate from the back of a watch case.
• It varies by time of day. Skin temperature rises in the evening as you wind down and peaks in the first few hours of sleep.

If a wearable showed you a raw, absolute number, the reading on the same person on the same night could swing by more than the actual fevers most people get. That would be useless at best and alarming at worst. So manufacturers don’t show absolute numbers. They show a deviation.

Why deviation, not degrees

Open the temperature tab on any current wearable app and you’ll see the same convention: a value reported as “+0.3°C from baseline” or “−0.5°C from baseline.” The baseline is your personal average overnight skin temperature across a recent window, typically the last 14 to 30 nights of wear. Apple uses about five nights to establish an initial baseline and rolls it from there. Oura builds a baseline over the first month of wear, then runs a rolling comparison.

Reporting as deviation does two things at once. It normalizes for the fact that every person sleeps in a different room with different blankets, and it forces the user to think in trends. The data gets useful around week three or four, when the baseline has settled.

This is the part most people miss. If you bought a temperature-tracking wearable last week and you’re already trying to read meaning into night-by-night numbers, the chart is mostly noise. Real signal shows up over weeks, especially around two patterns: illness onset and the menstrual cycle.

Illness onset: the 24 to 48 hours before you feel sick

The most repeatable real-world finding in consumer temperature data is that skin temperature often rises 24 to 48 hours before symptoms of an acute illness, most consistently for respiratory infections. The Scripps Research DETECT study, led by Jennifer Radin and published in The Lancet Digital Health in early 2021, looked at Fitbit data from tens of thousands of users and found that combining resting heart rate, sleep, and skin-temperature deviation outperformed self-reported symptoms alone for predicting a positive COVID test. A parallel project at Stanford, in Michael Snyder’s lab, reached a similar conclusion: an unusual elevation in overnight wrist temperature of roughly 0.5 to 1.0°C above the rolling baseline frequently preceded symptom onset.

This is the pattern wearable temperature is genuinely useful for. Not “is my temperature 38.5°C right now” (your watch can’t reliably tell you that), but “my overnight baseline has drifted up by three-quarters of a degree for two nights in a row, my resting heart rate is up by five beats, and my HRV is suppressed.” That cluster of signals shows up before a person feels unwell, and a few extra hours of warning can be the difference between rescheduling a Tuesday and infecting a meeting.

It is not diagnostic. None of the consumer skin-temperature features are FDA-cleared as fever detectors or as a screen for any specific infection. The signal is real; the regulatory claim is not.

Cycle tracking: the post-ovulation rise

The other clean signal in overnight skin temperature is the menstrual cycle. After ovulation, progesterone rises, and one of its effects is to raise basal body temperature by roughly 0.3 to 0.5°C until the next period. Fertility-awareness methods have used this shift for decades with a basal body thermometer at the mouth or vagina.

The wearable data is a noisier proxy for that shift, but in practice the good devices can pick it up. Oura’s cycle tracking, built around overnight finger temperature, has been the most-validated of the consumer implementations. Oura published a study in Nature Scientific Reports in 2024 (Maijala et al.) showing that finger skin temperature trends correlated well with cycle phase in the company’s user base. Apple introduced Cycle Tracking with retrospective ovulation estimates on the Series 8 in September 2022. Apple’s wrist-temperature-based cycle history feature has obtained regulatory clearance in some markets, for the purpose of estimating past cycle phases, not for contraception. Fitbit’s cycle tools and Whoop’s “menstrual cycle coaching” feature follow similar logic but lean less heavily on temperature alone.

If you want a much deeper look at how these devices map nightly physiology, we wrote up the underlying signals in our sleep stages explainer and our wearable sensor accuracy guide.

How wrist sensors stack up against the alternatives

The case against wrist temperature is mostly that it’s a noisier source than a thermometer or a dedicated basal body temperature device, both of which read closer to core. The case for it is that you wear the watch anyway, every night, with no extra ritual.

The takeaway in the table is the bottom row. A wearable is the right tool for “is something drifting in my body over the last few nights” and the wrong tool for “do I have a fever right now.” A drugstore thermometer costs less than a dinner and gives you a real answer for the second question.

Where the wrist data falls down

Even if you’re using a wearable for the right thing (trend detection, not absolute fever) there are limits to take seriously.

Night-by-night variance is real. Alcohol pushes the curve up. A bedroom that drops from 21°C to 17°C overnight pushes it down. A heavier comforter or a partner who runs warm changes the microclimate the sensor sits in. Most platforms will show a band around the baseline to communicate how much normal variance there is for you; learn what that band looks like before you read meaning into a single point.

Consistency of wear matters more than people think. If you wear the watch four nights this week and one next week, the rolling baseline is reset enough that the signal is gone. Whoop’s coach text will tell you this directly. Oura’s monthly summary becomes useless if you drop below about 80% nightly wear.

Sensor location changes the number. A finger ring sits inside a closed hand at night and runs warmer than a wrist with a watch on it. You cannot meaningfully compare the absolute deviation on an Oura to the absolute deviation on an Apple Watch on the same night. Each platform’s baseline is internally consistent and not cross-comparable.

Battery life on temperature-heavy devices. Continuous sensing has a cost. The Withings ScanWatch 2 advertises 30 days of battery and we measured 25 to 27 in real use with continuous skin temperature on; turning the temperature feature off pushes that closer to the marketing number. The Apple Watch and Fitbit Charge 6 sample temperature only at night, which is why the impact on their daily battery is small.

No wearable is a fever screen. This is worth saying twice. A reading of “+1.0°C from baseline” on your wrist does not mean your core temperature is 1.0°C above normal. It means your sleeping wrist was a degree warmer than usual, which often correlates with something interesting but is not a number a doctor will accept.

Practical advice if you actually want to use this feature

If you’ve read this far, the recommendation list is short.

Wear the device every night for at least three weeks before you read anything into the chart. The baseline is the product; without consistent wear there is no baseline.

Treat trends as the signal and individual nights as noise. Two or three nights of elevated deviation, especially alongside an elevated resting heart rate and a drop in HRV, is the cluster worth paying attention to. One night above baseline after a glass of wine is not.

For cycle phase, the wearable is a reasonable supplement to other signals if you wear it nightly. For fertility decisions in either direction, whether you’re trying to conceive or trying not to, layer it with cervical mucus observation and LH strips, or use a dedicated basal body thermometer cleared for fertility use. Do not use a wrist deviation chart as a contraceptive.

For fever, use a thermometer. A $15 ear or oral thermometer answers the question your watch was never designed to answer. The watch is good at telling you something might be off 36 hours from now. The thermometer is good at telling you whether you have a fever right this minute. These are different jobs.

If you’re picking a device specifically for temperature, the practical differences are about wear site and sampling cadence more than about sensor quality. Oura’s nightly finger temperature is the cleanest dataset because the finger sits in a stable microclimate inside a fist under a duvet. Apple’s wrist temperature has the benefit of a second sensor for ambient compensation. Whoop’s data is solid but locked behind a subscription, and the Oura versus Whoop tradeoffs go well beyond temperature. The wrist watches with full smartwatch features split between Apple and Fitbit on different ecosystems, and the Withings ScanWatch 2 covers the same ground for people who do not want a smartwatch aesthetic. Our best sleep trackers roundup covers the broader category.

The version of this feature that exists in 2026 is genuinely useful and genuinely limited. It is a long-running trend monitor strapped to your wrist or finger, and within those limits it works. The mistake is reading it like a thermometer. It is not one, and the manufacturers, to their credit, never claimed it was.