METHODOLOGY
Calm language requires disciplined limits.
How Trendline separates observations, sustained direction, analytic availability, source context, and claims it refuses to make.
Observation is not interpretation
A recorded measurement is a point in time. Trendline preserves that fact while deriving a slower, deterministic view intended to reduce the influence of ordinary fluctuation.
Deterministic and inspectable
The same readings, focus, and settings produce the same result. Trendline does not use a generative model to invent health conclusions or quietly change its interpretation.
Availability and confidence are different
Weight analytics in the iPhone app become available after seven distinct eligible Weight Reading Days; Composition follows the same seven-day rule for eligible Composition readings. Before admission, the app shows exact progress instead of manufacturing a conclusion.
Once an analytic has been earned, a different device, changed source metadata, or a new observation method does not start a new waiting period. Sparse, stale, or conflicting evidence can qualify a specific claim or widen a planning window, but provenance alone does not erase the history or silence a supportable result.
Focus changes meaning
Lose Weight, Build Lean, and Maintain are distinct interpretation contexts. A downward direction, upward direction, or stable range is not universally good or bad outside the focus selected by the person using the app.
Sources provide context, not proof
Apple Health can combine manual entries, devices, and apps, and it prioritizes data sources. Trendline retains the source context Apple makes available because a change in equipment or algorithm can resemble a change in the body. A source label can help someone investigate a discontinuity; it cannot prove what caused it, and it never creates a customer-visible reset by itself.
The energy examples are website tools
The formula and adaptive TDEE instruments on this website are free browser tools. They are not current features of the Trendline iPhone app. The equations below explain those website estimates and their limits.
| Equation | What it assumes | Where it fails |
|---|---|---|
| Mifflin‑St Jeor | Body mass, height, age and sex predict resting rate in an ordinary adult. | It cannot see body composition, so at one weight it can under-read a muscular person and over-read a softer one. |
| Harris‑Benedict | The same inputs, fitted to a 1919 cohort. | That sample was smaller, younger and leaner than the population now using it. |
| Katch‑McArdle | Lean mass alone predicts resting rate. | It needs a body-fat percentage, whose measurement error moves directly into the answer. |
Composition stays qualified
Consumer body-composition values are estimates. Research shows acute hydration can materially alter bioelectrical-impedance estimates. Trendline treats them as consistency signals and avoids tissue-level certainty.
What the method does not do
- Diagnose, treat, or assess medical risk.
- Promise a goal date or prescribe a fixed calorie target.
- Declare an exact amount of fat or muscle gained or lost.
- Turn a maintenance boundary into a medical threshold.
- Hide uncertainty behind a single confidence score.
Effect size 0.23, 95% confidence interval −0.00 to 0.47, p = 0.055 (Benton, Hutchins and Dawes, 2020). Pooled across all 26 studies the estimate is 0.33; that figure has no published interval here, so it is named rather than drawn.
Primary references
- Apple Health developer documentation
- Apple: manage Health data and sources
- Jeong et al. (2023): acute hydration and bioelectrical impedance
These sources support the boundaries above; they do not validate Trendline as a medical device.
See the rules applied
The free browser instruments run their own deterministic math in the open, with stated limits and nothing uploaded.