How do period trackers predict periods?

A period tracker uses the period starts you’ve entered to estimate when your next period might arrive. One basic method averages your previous cycle lengths and adds that interval to your latest start date.

A peer-reviewed comparison of prediction models looks at simple averages and more complex approaches that account for changing information and possible missed entries. These methods work from records; they cannot directly measure what your body will do next or know that the next interval will match the last.

A forecast worked through

Cycle length is the gap between period starts. It includes the days when you aren’t bleeding.

The mean is the arithmetic average of previous lengths; the median is the middle value when those lengths are put in order. Both were comparison methods in the prediction-model study by Li and colleagues. More complicated models can use your history alongside patterns in a wider dataset, or update a prediction as time passes.

These invented dates show how an average works. They are not a clinical example or a description of Veya’s algorithm.

Recorded period start	Next recorded start	Completed interval
1 April	29 April	28 days
29 April	29 May	30 days
29 May	30 June	32 days

The intervals total 90 days. Dividing by three gives a 30-day average. Adding 30 days to the latest recorded start, 30 June, produces an estimated next start of 30 July.

The arithmetic is correct even if the next period doesn’t start on 30 July: the uncertain part is whether the next cycle will equal the average. Four start dates have produced three completed intervals, because the latest cycle has only just begun.

Many app algorithms are proprietary. The study’s authors could not compare their model directly with those private algorithms, so this example cannot tell you how a particular app calculates its dates. Check the app’s own explanation of its method.

For help keeping the dates used in a calculation, see getting started with cycle tracking.

From a single date to a range

Repeating the shortest and longest previous intervals in the example would give a window from 28 July to 1 August. This illustrates a simple range; it is neither a guarantee that the next period will fall inside it nor a scientifically validated confidence interval.

A single date and a range are different ways to display a prediction; a range can make uncertainty easier to see. A spread based on past dates works differently from a model that estimates the probability of different outcomes. The Li study’s probabilistic model also considers uncertainty about whether a period went unrecorded.

Read the app’s explanation of how it chooses its window. If there isn’t one, you can still use the range for ordinary preparation, such as packing period products. A wider coloured strip doesn’t necessarily come with a known probability attached.

What temperature and test results add

Some approaches consider signs associated with ovulation alongside period dates. A positive urinary LH test suggests ovulation may be approaching; it does not confirm egg release.

Basal body temperature rises after ovulation and can help identify likely past ovulation retrospectively. That cycle’s temperature rise cannot give advance notice, and fever or insufficient sleep can affect readings.

Logging a test or temperature doesn’t tell you whether an app uses it in its forecast. Check the product’s explanation before treating a new entry as evidence that its prediction has become more certain.

How researchers tested accuracy

In a peer-reviewed evaluation of period-tracker predictions, Worsfold and colleagues entered constructed cycle histories into apps. The apps predicted the deliberately constant cycle well, but period-date predictions were less reliable when the profiles’ cycles varied. Different apps also produced conflicting results from the same histories.

The profiles came from existing data. The researchers did not follow new participants, and they tested app versions available at the time. Those results cannot give an accuracy percentage for every current tracker or for you personally.

Check what “accurate” refers to, too. Predicting a period start and predicting ovulation are different tasks; a result about one cannot be transferred to the other. Neither study says anything about how Veya’s planned estimates will perform.

Why your next period may fall outside the forecast

Natural variation makes prediction harder. More variable cycle histories were generally harder to predict in the Li study. Describing a pattern in previous cycles cannot make the next cycle follow it.

Missing entries matter as well. If a period happens but isn’t logged, the gap between recorded starts can look like one unusually long cycle. Li and colleagues found value in accounting for possible missed tracking. Real cycle changes also happen, so a long interval needn’t mean an entry is missing. Don’t add a period you cannot remember just to make the chart look regular.

An older pattern may become less useful after life changes. Stress, weight changes, breastfeeding and perimenopause are among the reasons periods may be late or absent. Some hormonal contraceptive methods also change bleeding. A forecast cannot identify which explanation applies to you.

Reading the result in your app

Check the underlying entries if a forecast surprises you, and mark uncertain dates as uncertain. Keep recorded starts separate from estimates. A blank entry cannot confirm that nothing happened; recorded patterns versus predictions explains how to read the two.

A predicted period date or fertile window is not contraception and cannot identify guaranteed “safe days”. Fertility-awareness methods require instruction from a trained professional. The NHS does not officially recommend fertility apps or monitoring devices.

Veya is pre-launch. Its planned estimates use your own logged history and are being designed to stay labelled as estimates, separate from logged events and recorded patterns. Planned basal-temperature logging may help mark likely past ovulation retrospectively; the app does not detect or confirm ovulation.

If your period hasn’t arrived

If pregnancy is possible, use a pregnancy test. An app’s forecast cannot answer that question. Test from the first day of a missed period, or at least 21 days after the last unprotected sex if you don’t know when your period is due. If the result is negative but you still think pregnancy is possible, wait a few days and test again. If a second test is negative and your period still hasn’t arrived, speak to a GP.

Arrange a GP appointment if you’ve missed three periods in a row or your periods have become irregular. You can seek help earlier if you’re worried.

Pelvic pain with possible pregnancy needs urgent advice: in the UK, use NHS 111; elsewhere, contact your local urgent medical service. Severe or worsening pelvic pain, or pain with faintness or heavy vaginal bleeding, needs emergency help, whatever date an app predicts. Call 999 or go to A&E in the UK, or use your local emergency number or emergency department.