Skip to main content

Methodology

How Metric turns Apple Health signals into estimates and trends.

Metric combines supported Apple Health measurements over stable time windows, maps each available signal to age- and sex-aware reference behavior where relevant, and displays the result with data-coverage context. Estimates are for wellness tracking and should be interpreted alongside their inputs—not as standalone conclusions.

Last reviewed July 16, 2026

Biological-age estimate

Metric’s biological-age feature compares a set of cardio-fitness, cardiovascular, body-composition, movement, and sleep signals with reference ranges. Most inputs use rolling 90-day data so that one unusual day has less influence.

Each available input produces an age-like contribution. Metric starts missing contributors at chronological age instead of redistributing their weight among the signals that happen to be present. This keeps sparse data from giving one measurement disproportionate influence.

Current biological-age model inputs and maximum contribution weights
SignalWeightRole
VO₂ Max18%Cardiorespiratory fitness
Resting heart rate14%Baseline cardiovascular demand
Heart-rate variability (SDNN)10%Autonomic and recovery context
Body-fat percentage10%Preferred body-composition signal
Sleep duration10%Typical nightly sleep duration
Daily steps9%Everyday movement volume
Lean-mass ratio9%Body-composition fallback when body-fat percentage is unavailable
Body mass index5%Broad body-size context
Systolic blood pressure4%Pressure when the heart contracts
Activity load4%Active energy relative to basal energy
Sleep regularity4%Night-to-night timing consistency
Diastolic blood pressure3%Pressure between heartbeats

Coverage, freshness, and fallbacks

  • VO₂ Max, BMI, and body-composition measurements can contribute with fewer samples because they are often recorded less frequently.
  • Resting heart rate, HRV, steps, activity load, and sleep generally require repeated samples before they contribute.
  • Body-fat percentage is preferred for body composition. Lean-mass ratio can act as a fallback when body-fat percentage is unavailable and both lean mass and body mass are present.
  • Activity load uses active energy in relation to basal energy when both are available.
  • Older or missing data lowers the coverage shown in the app. Rolling-window values outside the model’s plausible ranges do not contribute.

Aging rate

Metric averages the most recent 90 generated daily biological-age entries and begins showing aging rate once at least seven entries are available. A daily entry can still be generated when source data is sparse because missing contributors default to chronological age; the separate coverage indicator shows how much eligible data supports the result. Aging rate divides the rolling estimate by chronological age and limits the displayed ratio to 0.10×–3.00×.

A result below 1.00× means the rolling estimate is younger than chronological age; a result above 1.00× means it is older. Because the value is derived from the same underlying inputs, a person should examine the contributors and data coverage rather than interpreting the ratio in isolation.

Workout strain and readiness context

Metric assigns workouts a training-load value using duration and heart-rate intensity, then tracks short-term and longer-term load with exponential decay. Acute load uses a roughly five-day decay on workout days and a faster roughly 2.5-day decay on rest days; chronic load uses about 32 days. The model is bootstrapped from up to 90 days of workouts.

The displayed 0–100 score is oriented as freshness: a higher score represents lower accumulated strain, while a lower score represents higher strain. The score considers short-term load relative to a longer-term baseline, absolute acute load, and today’s workout load. It is training context, not a directive to ignore fatigue, pain, illness, or professional guidance.

Model changes

Metric may adjust reference ranges, data-quality rules, weights, and presentation as the product improves. This page is updated when a material change affects how a public-facing result should be interpreted. The review date at the top identifies the current public methodology version.