How Cycle Predictions Work, and Why an App Should Show Its Evidence
Data · 2025-02-20 · 7 min read
Most period apps give you a date. Few tell you how they got it. A study of more than 600,000 cycles shows why that matters, and what a prediction that shows its work looks like.
In a study of 612,613 ovulatory cycles from users of a fertility app, only 13% were 28 days long. The authors concluded that the widely held belief that ovulation happens consistently on day 14 does not hold for most people.
The 28-Day Myth
Many calendars and simple apps start from a textbook cycle: 28 days, with ovulation on day 14. It's a convenient average. It's also a poor description of most real cycles.
In 2019, researchers published one of the largest real-world analyses of cycle data in npj Digital Medicine. Working with anonymised records from Natural Cycles users, Jonathan Bull and colleagues analysed 612,613 ovulatory cycles from 124,648 users. The mean cycle length was 29.3 days, but the spread around it was wide. The study was funded by Natural Cycles, two of the authors are employees of the company, and two are its founders, so the data comes from one app's users, who are not a random sample of everyone who menstruates.
- 612,613: Cycles analysed
- 13%: Were exactly 28 days
- 65%: Fell between 25 and 30 days
- 10–30: Reported 95% interval for follicular phase length, in days
The paper attributes most of the difference in cycle length to the follicular phase, the time from the start of a period to ovulation. Its mean was 16.9 days, with a reported 95% interval of 10 to 30 days. The luteal phase after ovulation was steadier, averaging 12.4 days. In total, 65% of cycles were between 25 and 30 days long, and 13% were exactly 28. Cycle length also changed with age, and variation rose markedly after 40.
What a Prediction Actually Is
A cycle prediction is an estimate built from your own history. The simplest version averages your recent cycle lengths and counts forward. Better versions weigh some cycles more than others, use signals that change around ovulation, and say how sure they are.
- Period dates: the length of your recent cycles and how much they vary.
- Basal body temperature: a small, sustained rise after ovulation.
- Ovulation tests: LH readings that surge before ovulation.
- Other hormone readings: PdG and E3G from at-home test strips.
The authors of the 2019 study were blunt about calendar-only methods: apps that predict fertile days 'based solely on an outdated understanding of ovulation day variation could completely miss the fertile window'. They concluded that to identify the fertile period it is important to track physiological parameters such as basal body temperature, and not just cycle length.
Why Showing the Evidence Matters
A date on a screen looks precise whether it rests on twelve regular cycles or two irregular ones. When an app conceals its reasoning, you can't tell the difference, and you can't judge how much weight to give it.
An estimate that shows its evidence lets you check it against what you know about your own body. If a prediction is unsure because your last three cycles varied by a week, you should be able to see that. If a temperature rise moved the estimate, you should be able to see that too.
A prediction is a probability, not a promise. The honest version tells you how confident it is and what it is based on.
How HushCycle Shows Its Work
Every prediction in HushCycle comes with a confidence meter and a 'Why this estimate?' sheet that lists the evidence behind it. Newer cycles count more: HushCycle weights your last six cycles from 1.0 to 1.5, and says so on screen with the caption 'Newer cycles count more.' If your cycles vary, you can tell the app in Settings → Cycles → Cycle pattern: Regular, Somewhat, Irregular or Not sure.
The fertile-window estimate runs from five days before the estimated ovulation day to one day after. You can log LH, PdG and E3G test readings as numbers, and basal body temperature, where HushCycle looks for a rise of 0.20 °C held for three mornings. With a supported Apple Watch, you can choose to read wrist temperature from Apple Health; it is off by default, and HushCycle never writes to Health.
HushCycle is careful about what an estimate is for. It is not birth control and never estimates safe days. Its information is general, not medical advice, and it never diagnoses. All of the math runs on your iPhone, and the logs it learns from never leave it.
Sources
- Real-world menstrual cycle characteristics of more than 600,000 menstrual cycles, Bull JR, Rowland SP, Scherwitzl EB, Scherwitzl R, Gemzell Danielsson K, Harper J, npj Digital Medicine 2, article 83, 27 August 2019. https://www.nature.com/articles/s41746-019-0152-7