Skip to content
Beta Beta is open.
Join the beta
Life & Health Metrics
Science · Methodology

How biological age and aging speed are computed.

The methodology rests on patterns within locomotor activity — not raw step counts, which are too noisy on their own. The patterns examined are intensity transitions, duration of sustained activity, recovery cadence between bouts, and perpetuity of activity over weeks. These have been shown in peer-reviewed work to correlate with biological age, mortality risk, and chronic-disease incidence.

Pipeline

From sensor to BioAge — five stages.

The journey from raw phone-sensor data to a personalised aging indicator is a deliberate sequence. Each stage is inspectable; nothing is opaque.

  1. Step 01

    Sensor

    Locomotor activity (steps, motion intensity) is read from the phone or from connected wearables.

  2. Step 02

    Preprocess

    Streams are normalised, calibrated across devices, and require a minimum window before any reading is produced.

  3. Step 03

    Model

    A peer-reviewed BAA model translates the activity patterns into a biological-age estimate.

  4. Step 04

    Calibrate

    A confidence rating is computed alongside. Less data, more variance, or signal gaps lower the confidence; users see both the number and how much weight to put on it.

  5. Step 05

    Surface

    BioAge, 28-day aging-speed trend, and confidence are surfaced together in the app. Never a daily figure.

Scope

What it doesn't claim.

Three boundaries that are part of the methodology, not addenda to it.

  • Not a clinical diagnosis

    BioAge is a wellness signal validated in peer-reviewed research, not a medical diagnostic. The model does not detect, treat, cure, or prevent disease, and is not a medical device.

  • Not a recommendation engine

    The methodology informs decisions; it does not make them. The product never tells the user (or the user's advisor) what to do. It shows how decisions reshape given the trajectory.

  • Requires ≥10 days of data

    No reading is produced before the minimum window is met. A short tail of activity is too noisy to interpret as biological age, and the product will not surface a premature number with false-precision.

Biological age

A 50-year-old can have a biological age of 45 (slower aging, in better physical condition than the average 50-year-old) or 56 (faster aging, accumulated wear and tear). Biological age is a slow-moving indicator: it changes over weeks and months in response to sustained patterns of activity, recovery, stress, and lifestyle.

Aging speed

The signed gap, in years, between biological and chronological age. Positive values mean the body is running ahead of the calendar; negative values mean it is running behind. Aging speed is not read off a single day — it is derived from a long-term trend in the BAA signal, so day-to-day noise does not move it. Sustained patterns of activity, recovery, and stress do. The result: a person who begins exercising regularly, sleeping better, or managing stress will see aging speed bend first; biological age, the integrated quantity, follows over months.

  • 0biological age tracking chronological — body keeping pace with the calendar
  • +4biological age four years ahead of chronological — body has accumulated more wear than the calendar suggests
  • −3biological age three years behind chronological — typically the result of sustained positive lifestyle patterns

Why locomotor activity

Three properties make locomotor activity the right input class:

Already available

Modern smartphones record steps and movement continuously, with no additional hardware, sensors, or appointments required.

Longitudinal by design

A blood test or DNA assessment is a snapshot. Locomotor data accumulates daily, automatically, capturing patterns of activity, intensity, recovery, and consistency. The indicator improves with the length of the record.

Scientifically validated

The algorithmic foundation has been validated against external biomarkers — DNA methylation clocks, blood-based markers, and other clinical inputs — with consistent agreement across the methods.

Aging trajectories — what the output looks like

Each user has a biological-age trajectory plotted against chronological age. The vertical distance from the diagonal is aging speed in years at that age; the slope of the line shows whether that distance is widening, holding, or narrowing over time. Three categories emerge in the target audience (age 40+):

Four-panel chart of aging trajectories. Top-left: healthy ager (green) — trajectory advances slower than chronological, aging speed becomes negative over time. Top-right: tracking calendar (grey) — trajectory follows the diagonal, aging speed near zero. Bottom-left: age-related strain (red) — trajectory advances faster than chronological, aging speed grows positive. Bottom-right: population overlay of all three categories, 17 users each.
Illustrative output. Each panel plots biological age (vertical) against chronological age (horizontal); the dashed diagonal is the reference where aging speed = 0. Healthy ager — trajectory advances slower than chronological time, so aging speed becomes increasingly negative. Tracking calendar — trajectory follows the diagonal, aging speed near zero throughout. Age-related strain — trajectory advances faster than chronological time, aging speed grows positive. The bottom-right panel overlays seventeen users from each category to show the spread within the target audience. Per-panel slope values in the legend are a technical rate-of-change reading; the headline metric in-product is aging speed in years.

Financial projection layer

The financial-projection model computes retirement projections from a small set of inputs the user provides — current savings, planned contributions, expected investment return, target retirement age, payout scheme — combined with publicly available actuarial statistics and information about retirement products.

The model is intentionally transparent. A user, advisor, or provider analyst can inspect each input, follow the calculation, and see how the output changes when an assumption changes. There is no opaque black box; there is no hidden parameter.

Current limits

The model is intentionally simple at present:

  • ·Inflation is not yet modelled separately from the return assumption
  • ·Contribution-growth profiles are not yet user-configurable
  • ·Family and household plans are not yet supported

These are deliberate simplifications that keep the model accessible, easy to explain, and easy for non-specialist users to interrogate. The model is on a clear path of progressive sophistication, including through future integrations with retirement-product providers.

Read the publications.

Six peer-reviewed papers underwrite the methodology. Each entry on the publications page has a one-line takeaway.