Why the Apple Watch estimates ovulation only in hindsight, never in advance

The Apple Watch Series 8 and later measure wrist temperature during sleep and use it to produce a retrospective estimate of when ovulation likely occurred, which Apple presents as a way to improve period predictions — not as birth control [s3]. The word that carries the biology is "retrospective." The watch can suggest that ovulation happened; it cannot tell a user it is about to.

The temperature signal is real but late

Body temperature has been used to track the menstrual cycle for decades because it is biphasic: after ovulation, rising progesterone lifts basal body temperature, which stays elevated through the luteal phase until menstruation. Clinical studies put that rise at roughly 0.28 to 0.56 °C [s1]. Wearables detect a muted version of the same shift at the skin. In a study using a wrist bracelet, nightly wrist skin temperature and other autonomic signals moved consistently across the cycle [s1]; in a pilot with a finger-worn ring, nocturnal skin temperature was higher in the luteal phase than the follicular phase by 0.30 °C (SD 0.12), tracking oral temperature closely [s2].

The catch is timing. The temperature rise is a consequence of ovulation, driven by the hormone released after the egg is gone. It confirms that ovulation has occurred; it says nothing in advance about the fertile days that precede it — which is exactly when pregnancy is possible. A method built on the post-ovulatory temperature shift is, by construction, a rear-view mirror.

What the accuracy looks like

Detecting the events retrospectively is easier for some than others. In the ring study, algorithms flagged the start of menstruation with a sensitivity of 71.9% to 86.5%, depending on how wide a window around the reported day counted as a hit [s2]. Pinpointing ovulation from temperature alone was harder and less reliable, which is the general pattern across temperature-based methods: the luteal shift is detectable, but the precise day of ovulation is noisy, and confirmation arrives only after the fertile window has closed.

This is why the more sophisticated fertility wearables combine temperature with resting pulse, respiratory rate and heart-rate variability rather than relying on temperature alone [s1]. Even then, they are estimating a past event, and their real-time predictions of future fertile days depend on modelling and cycle history, not on the temperature reading of the moment.

There is also noise to contend with. Wrist skin temperature is a step removed from core temperature, and the overnight signal is pushed around by things that have nothing to do with the cycle: room temperature, bedding, alcohol, a cold coming on, a disrupted night's sleep. The luteal shift the watch is trying to detect is small — a few tenths of a degree — so it takes consistent overnight wear across multiple nights to separate the signal from that background, and a single warm night means little. The ring study's own accuracy for the harder target, ovulation, reflected that difficulty rather than any failing specific to one device [s2].

Reading the feature honestly

Apple's framing is appropriately narrow, and it is worth taking at face value: retrospective ovulation estimates that may sharpen the app's period forecasts, explicitly not a contraceptive [s3]. Used that way, the feature adds a genuine biological signal to what would otherwise be pure calendar arithmetic.

The failure would be to use it as either a way to conceive or a way to avoid conceiving in real time. For prediction of the fertile window, temperature comes too late, and the apps that most people actually use are worse still: most period-tracking apps forecast ovulation from the calendar alone and misclassify fertile days as safe in a large share of cycles. A watch that measures a real post-ovulatory temperature rise is a step up from a calendar guess — but only for telling a user what her body already did, not what it is about to do.