Turtle data
How we produce Turtle's data posts
Where the numbers come from, who is counted, how we keep members anonymous, and what the figures can and can't tell you.
Source
Every figure comes from the Turtle Method app: answers women give when they join, their morning diary, food diary, workouts and sleep synced from Apple Health or Health Connect, the in-app strength check, and app activity such as nudges and coach messages (timing only). We analyse a pseudonymised copy of the app data: names, emails and other direct identifiers are removed, and each member is represented by a one-way code. All analysis is read-only.
Who is counted
- Women only, as recorded in their profile. Staff and admin accounts are excluded.
- Each post sets its own minimum, such as 5 diary nights or 14 nights of sleep, so one-off entries don't skew the result.
- UK clock times use women whose account is billed in pounds (GBP), as a stand-in for living in the UK.
- Averages are per woman first: we work out each woman's own figure, then average across women, so the heaviest app users don't dominate.
Anonymity: the 50/10 rule
We publish aggregates only. Every group shown has at least 50 women, and no share is built on fewer than 10. Groups that don't meet the rule (often the youngest and oldest age bands) are left out rather than merged quietly. Group sizes are rounded ("about 420 women"), never shown as small exact counts.
Rounding
Percentages are rounded to whole numbers (halves round up), times to the minute, steps to the nearest 500 and kcal to the nearest 1. Group sizes are rounded to the nearest 10 under 1,000, the nearest 100 under 10,000, and the nearest 1,000 above that. Shares in a table may not add to exactly 100%.
How every figure is checked
Each post comes from a saved query, run once and kept with its results file. The pages are generated from those files, not typed in, and a separate check script re-derives every published figure from the results and fails if any page disagrees, or if any group falls under the 50/10 rule.
Update cadence
We refresh the data twice a year; the next update is due in April 2027. Each refresh re-checks the 50/10 rule and updates the dates on every page. If a refresh changes a headline, the page says so.
Caveats
- Not a random sample. Women who use a health and fitness app, wear a tracker or take a strength test may differ from women who don't.
- Patterns, not causes. We describe what the data shows; we don't claim one thing causes another.
- Logged, not measured. Food figures are what women logged, not what they ate. Sign-up answers are women's own estimates, and they can update them later.
- Devices vary. Watches and phones record sleep and workouts differently and can miss or misread them.
- Time zones. Wake-up dates use UTC, which is the same calendar morning for the UK and most members; clock-time findings are UK only.
- Not medical advice. These are population patterns, not guidance for any individual. This hub covers training, sleep, activity, steps and food logging only; it doesn't include menstrual cycle, symptom, medication or other health-condition data.
Each post
8 in 10 women end most nights with a phone or TV
Turtle's morning diary asks "TV or phone before sleep?" (yes or no) and how well you slept. We counted women who answered on 5 or more nights between 1 January and 30 September 2026, about 420 in all. Period: 2026-01-01 to 2026-09-30. Coverage: Mainly UK.
Women sleep about 25 minutes longer on Saturday nights
Sleep synced from Apple Health or Health Connect, 7 March to 30 September 2026: the longest sleep per morning, kept if 3 to 14 hours. Each woman needed 3 or more nights of a given weekday; clock times use UK time for women with a UK (GBP) account. Period: 2026-03-07 to 2026-09-30. Coverage: Worldwide (clock times: UK).
Nearly half of women average under 7 hours of sleep a night
Sleep synced from Apple Health or Health Connect, 7 March to 30 September 2026: the longest sleep per morning, kept if 3 to 14 hours. Women needed 14 or more nights; age groups with under 50 women are not shown. Period: 2026-03-07 to 2026-09-30. Coverage: Mainly UK.
Only 1 in 4 women say they walk 10,000 steps a day
The Turtle sign-up question "How many steps do you roughly do per day?", latest saved answer, women who joined between 2021 and October 2026 (about 74,000). Answers under 100 or over 50,000 were left out. Period: 2021-01-01 to 2026-10-07. Coverage: Mainly UK.
Walking is the workout women log most, and over half log strength training
Workouts synced from Apple Health to Turtle, 1 January to 30 September 2026, using Apple's workout type codes; about 1,100 women. Manual and Health Connect workouts are left out because their types are coded differently. Period: 2026-01-01 to 2026-09-30. Coverage: Mainly UK.
Monday is the busiest workout day, and nearly 1 in 3 workouts start before 9am
Workouts synced from Apple Health, 1 January to 30 September 2026, for women with a UK (GBP) account, start times in UK time. Walks are counted separately because many are recorded automatically. Period: 2026-01-01 to 2026-09-30. Coverage: United Kingdom.
Nearly half of women can hold a plank for a minute
Each woman's first Turtle strength check, May to July 2026, self-timed at home. Each figure uses one version of the test only (forearm plank, full floor push-ups, unsupported balance); groups under 50 are not shown, so there is no age split. Period: 2026-05-01 to 2026-07-31. Coverage: Mainly UK.
Praise gets acted on six times as often as a calorie warning
LeanShield suggestions sent 26 June to 30 September 2026. For each woman who got a nudge type 3 or more times, we took the share she acted on, then averaged across women. Medication prompts and housekeeping prompts are left out. Period: 2026-06-26 to 2026-09-30. Coverage: Mainly UK.
Women log food on over a third fewer Sundays than Mondays
Daily food diary totals, 5 January to 27 September 2026 (38 full weeks, so each weekday appears equally). A logged day is 500 to 5,000 kcal; women needed 20 or more logged days. Period: 2026-01-05 to 2026-09-27. Coverage: Mainly UK.
Evening is coach time: 29% of messages arrive between 6pm and 10pm
Timestamps of messages women sent to Turtle's AI coach, 1 June to 30 September 2026, for women with a UK (GBP) account, in UK time. Only the time was counted; message content was not read. Period: 2026-06-01 to 2026-09-30. Coverage: United Kingdom.
Licence and contact
Figures and charts are licensed under CC BY 4.0: free to use with credit and a link to the source page. Corrections and data requests: scott@turtlemethod.com.