Physical activity level

Physical Activity Level measures how active your entire day is, not just your workouts. It is your total daily energy expenditure divided by your basal metabolic rate — the energy your body would use lying still. A PAL of 1.0 would mean complete bed rest. Around 2.5 is elite-athlete territory. Verve reads it from Apple Health's basal and active energy, averaged over seven days.

The four levels

The tiers come from the Institute of Medicine's Dietary Reference Intakes, which defined them to set energy requirements for populations: sedentary 1.00–1.39, low active 1.40–1.59, active 1.60–1.89, very active 1.90–2.50.

Read the trend, not the number

Both halves of this ratio are estimates, so the error compounds. Active calories come from your wrist, where systematic reviews put energy-expenditure error above 30%, and they miss effort that does not move your arm — cycling, rowing, lifting, carrying. Your resting figure is not measured at all; it is predicted from your height, weight, age and sex. Every wearable tested against doubly-labelled water under-estimates, for Apple Watch by roughly 394 kcal a day, and the bias grows at higher effort and higher body fat.

The practical consequence: a single PAL carries a band wide enough to straddle two tiers. Whether it drifts up over months does not. That is why Verve shows this as a trend read against the tier bands rather than as a figure to hit, and why the app's chart marks which band each week or month fell in rather than asking you to read a slope off a noisy line.

The one correction Verve makes

Apple reports basal and active energy but not the thermic effect of food — the energy spent digesting, conventionally about 10% of total daily expenditure. Their sum is therefore not total daily energy expenditure, so Verve divides by 0.9 to recover it. That correction is definitional rather than a judgement call, and it is the only adjustment applied.

Two further adjustments were tried and removed: an additive correction for occupational activity, and a rule that raised the tier when logged training contradicted a sedentary label. Both were judgement calls stacked on a base with more than 30% error, and the occupational one double-counted part of the very bias the thermic-effect correction already approximates. Layers of inference do not rescue a noisy instrument; they only obscure how noisy it is.