The Evolution Index expresses a person's functional capacity as a single number from 0 to 100 — normed to their age and sex, alongside a band and a Capacity Age. It exists to make decline visible early enough to reverse, and to direct effort at the one input that will move the most. This is how it works, and why we trust it.
The Provisional Index is free. You enter the measurements yourself and you create your account on that page. It needs numbers a watch or a gym can give you — a VO₂ max estimate, grip, body composition — and it tells you how much each estimate widens the error. The certified assessment is taken in person at 32 Grosvenor Square on calibrated equipment. If you have none of that to hand, the Field Test needs only a staircase and a chair.
The Index exists to direct work, not replace it. Every report ends with one intervention, never a list.
Cardiorespiratory fitness, strength and movement are independent risk pathways. Excellence in one does not offset failure in another — and the scoring enforces it.
Where a measure is included for engagement rather than hard outcome evidence, we say so — here, and in the report. Where a measurement device cannot carry the claim, the value is recorded as context and does not score.
Every metric carries a stated minimum detectable change. Below it, we report no change. Almost nobody in this market publishes their error. We do.
Three domains, weighted by the strength of the evidence linking each to all-cause mortality and loss of independence. The weighting is deliberate: cardiorespiratory capacity carries the most, because the evidence behind it is the strongest.
How well the body takes in, transports and uses oxygen — the single most powerful predictor of longevity in the battery.
The muscle and connective tissue that carry you through life and protect you when things go wrong. Four measures in v1.1: grip and lean tissue carry equal and largest weight, then lower-body power, then the hang.
The ability to get down to the world and back up from it — mobility, balance and control.
Standardisation is everything: the same conditions, the same scripted words, the same equipment at every retest — because a change in the test can masquerade as a change in the person. Each test below lists what it captures, how it is administered, and its published error — the minimum detectable change (MDC) below which we report no change at all.
Ramp or Bruce protocol to volitional exhaustion on treadmill or cycle, with the modality fixed for the client's history. A true maximum is confirmed against objective criteria; where it can't be, it is recorded as a peak and flagged.
A complete stop. At the instant the maximal test ends the client comes to a full halt and stands motionless — no walking cooldown, no continued pedalling, no sitting, no hands on knees, no leaning on the rail. The timer starts at the moment of cessation. The score is peak heart rate minus heart rate at exactly 60 seconds. Posture and cessation dominate this measure, so both are scripted and held identical at every retest; any deviation is recorded and the value flagged rather than quietly compared.
Seated, elbow at 90°, a calibrated hydraulic dynamometer. Three maximal efforts per hand with rest between; the best dominant-hand value is used. The same device is used at every retest — brands are not interchangeable.
Whole-body fat-free mass in kilograms divided by height in metres squared, measured fasted and normally hydrated on the same device each time. FFMI replaces appendicular lean mass in v1.1: it is available from every reference-grade body-composition method rather than DEXA alone, it is more robust to limb-segmentation differences between scanners, and it is what the devices in the field actually report.
The acquisition method decides how the value is used, because a number is only as good as the instrument that produced it:
One of three, fixed for the client's history. Countermovement jump — hands on hips, best of three, height in centimetres — for the private cohort. The timed five-repetition sit-to-stand is the validated alternative for corporate and older clients. New in v1.1, the single-leg sit-to-stand is the most discriminating of the three: the client rises from a graded seat, unaided, on the weaker leg, and the score is the lowest seat height they can clear. Seats are stepped in fixed increments and the same stack is used at every retest.
The single-leg version replaces the farmer's carry in the battery. It isolates one limb, so it exposes the asymmetry a bilateral test conceals; it needs a box and nothing else; and unlike a loaded carry it does not confound leg capacity with grip, which the battery already measures twice. Power is a more sensitive early-warning signal than maximal strength.
A passive hang from a fixed-diameter bar, feet clear, to release. Scored as work — hang time multiplied by bodyweight — so that heavier clients are not penalised for carrying more tissue. Single attempt, capped for safety.
Barefoot, on a clear floor: sit down and stand back up using as little support as possible. Points are deducted for each hand, knee or loss of balance used. Deduction rules are posted and one assessor owns a client's scoring across their history.
Eyes open, hands on hips, timed to the first loss of position, best of two. The published mortality evidence uses a 10-second pass/fail; we use a timed scale, with 10 seconds sitting near the base of the range.
Barefoot, hips below the knee crease, heels flat, torso self-supported, held for time. Continuous and directly trainable — it represents the very capacity the domain claims to measure.
Every raw measurement — a VO₂ figure, a grip in kilograms, a balance time — converts to a 0–100 metric score through the same anchor system, so that a score means the same thing on every test. A 55 is a 55 whether it came from a treadmill or a squat hold.
Scores between anchors are interpolated smoothly; for tests where lower is better — the timed sit-to-stand, the seat height in the single-leg rise — the scale simply runs the other way. Thresholds are set at a reference age and adjusted continuously for the individual, so nobody gains or loses points overnight on a birthday.
The top band is named Extending in v1.1. The previous name, Compounding, borrowed a financial metaphor that implied returns accrue on their own. They do not: capacity at this level is held and pushed further only by continued work, and the name now says so.
If any single domain scores below 40, the overall Index is capped at 59 — the top of Holding — no matter how strong the arithmetic elsewhere. A superb VO₂ max cannot buy back sarcopenic muscle. This is the methodological expression of the central claim: these are independent pathways, and they do not trade against one another. We never apply the cap silently — the report states the uncapped score, the cap, and the domain responsible.
Capacity Age is the chronological age at which your current results would be typical — the age at which they'd sit right at the median. It's the translation device: one number for precision (the Index), one for meaning (the age). It is a functional statement and nothing more — not a biological age, metabolic age or heart age. That discipline is deliberate, and it is what keeps the measure honest.
Thresholds are published at a reference age of 55 and scaled continuously from there, so the norms move with the person rather than stepping on a birthday. Lean tissue is scaled on a gentler curve than the performance tests: whole-body fat-free mass carries bone and organ tissue, which does not fall away at the rate limb muscle does, and a multiplicative curve borrowed from the strength tests would make the FFMI standard implausibly easy in the eighth decade.
Version 1.1 adds an 18–29 age group, and with it a correction to the model. Below 30 the norms plateau: an 18-year-old and a 30-year-old are scored against the same expectation. Extending the linear curve below 30 implies that capacity keeps rising through the early twenties at the same rate it falls after 55, which the population data does not support — peak capacity is reached in the late twenties and holds. Without the plateau a fit 22-year-old would be scored against a standard nobody has measured, and would be marked down for it.
Capacity Age is floored at 30 for the same reason. Below the plateau the norms are flat, so a Capacity Age of 24 would be an artefact of the arithmetic rather than a finding. Results at or beyond the plateau are reported as <30, and the Index continues to separate performance above that point.
Resting heart rate and heart-rate variability are recorded where a client has them, from a wearable or a morning reading. They are enhancement inputs: they never move the Index. Not one point, in either direction.
What they do is narrow the confidence interval reported around the Index. The composite carries a published interval of ±3 points. A single test session is a snapshot, and a snapshot of a person who slept badly, travelled, or arrived under-fuelled is a noisier estimate of their capacity than the same session backed by months of autonomic data. Continuous resting data tells us how representative the day was — so it buys precision, not score.
The narrowing is weighted by how much history a reading represents:
With both inputs supplied at 90 days the interval narrows to ±2.0 points, and no further — the remaining uncertainty belongs to the tests themselves, and no amount of wearable data removes it.
A narrower interval does not lower the threshold at which we report a change. The composite minimum detectable change stays at 4 Index points for every client, whatever their wearable history, until the test–retest study on the validation roadmap replaces the estimate with a measured figure. Tightening the reporting rule on the strength of data that never entered the score would be exactly the kind of quiet inflation this protocol exists to avoid.
The scale is anchored on established clinical and population science, then extended by Evolution's own assessment experience where the literature runs out:
The exact cut-points that convert a raw result into a score, the age-scaling that personalises them, and the precise weightings within each domain are Evolution's proprietary calibration. The method above is open — anyone can inspect and reproduce the logic, and that openness is what makes the Index citable. What is ours is the calibration those norms are tuned to, sharpened against the Evolution client dataset. You are not being asked to trust a black box; you are being shown the whole machine, minus the settings we earned.
Every metric has a stated minimum detectable change, and the composite Index carries an error of ±3 points — narrowing to no less than ±2.0 where enhancement inputs are supplied, as described above. A change of four points or more between assessments is real and reportable; anything smaller is stated as unchanged, with the reason given. Retest at twelve weeks, not sooner — the delta is the product, and a delta smaller than the error is not a delta.
The Index is publishable now as a provisional instrument. It becomes progressively more defensible — and more genuinely proprietary — through a deliberate validation roadmap:
A scoring system nobody can inspect gets ignored. One that can be inspected — and that states plainly where it is still provisional — gets trusted, and eventually cited. That is the whole idea.