How Makuro estimates calorie needs

Total daily energy expenditure (TDEE) includes resting needs, activity and digestion. Makuro starts with Mifflin–St Jeor, an equation developed in healthy adults, multiplied by a selected activity factor.¹ A systematic review found it predicted resting expenditure within ±10% of the measured value in 82% of non-obese and 70% of obese adults, with individual errors from −18% to +15%. The same review noted that the equations had not been validated in US-resident Asian and other ethnic minority groups.²

Adding activity adds error. Even when each person's activity level was taken from doubly labeled water measurements, the U.S. National Academies' 2023 equations predicted total expenditure with a typical error of about 250–340 kcal per day.³ An equation built from 6,497 doubly labeled water measurements had a mean absolute error of 11.2%.⁴ A self-selected activity level is coarser still: in a Japanese study, almost all activity questionnaires underestimated measured total expenditure, some by more than 600 kcal per day on average.⁵ Makuro therefore treats the starting estimate as uncertain by about ±15%.

If enabled, Apple Health active energy changes the baseline to (predicted resting expenditure + average active energy) × 1.1, bounded to 1.2–2.4 times predicted resting expenditure. The digestion allowance and bounds are product heuristics. Wearable estimates have substantial error; a 2017 validation found poor energy-expenditure accuracy across the devices tested.⁶ Makuro applies the same ±15% uncertainty to this baseline. That is our judgement, not a measured property of current watches.

Energy balance links what you eat, what you burn and changes in stored energy. If you log intake and weigh yourself, expenditure ≈ average intake − energy stored or released. Researchers use this "intake-balance" approach with doubly labeled water and body scans. A mathematical version needing only body weight matched it within 40 kcal/day on average over six-month periods, with individual differences of about 215 kcal/day.¹² Makuro uses a simpler short-window form of this idea. That study does not validate it.

The fit. Within the last 28 days, Makuro uses every stretch where each day between weigh-ins has complete intake records. Repeated weigh-ins on one day are combined by median. Within each stretch it fits weight minus cumulative intake ÷ 7,700 kcal/kg against time. The slope of that line, found as the median of slopes between weigh-ins at least seven days apart, estimates expenditure. This is the Theil–Sen estimator,^15,16 computed only within uninterrupted stretches, following the seasonal Kendall construction.¹⁷ It resists isolated bad weigh-ins. A missed day splits a stretch; it does not throw away the data before it.

How much to trust it. Makuro estimates the fit's own uncertainty with the standard error of a trend line. The correction for day-to-day correlated scale noise comes from climate statistics.¹⁸ Scale noise is measured from your data with a robust spread (median absolute deviation) that one outlier cannot inflate.¹⁹ The fit and the ±15% starting estimate are then combined by inverse-variance weighting, a standard way to combine two independent estimates. More, cleaner and more regular data earns more weight. In our simulations, with 28 days of daily weigh-ins and realistic 0.4 kg scale noise, the formula kept about 4–5% of the weight. It kept 9–12% when noise persisted across days, and 22–40% with only 14 days of data. With strongly persistent noise, the fit's uncertainty was about 30% smaller than it should be, a known limitation of this correction.¹⁸ These weights are not a clinically validated confidence interval. For comparison, a modeling study found that more than 28 days of daily weighing is needed to estimate a change in intake with a 95% confidence interval under 300 kcal/day.¹³

Requirements. The most recent fully logged stretch must end with a weigh-in from the last three days, and at least seven weigh-in days and 14 fully logged days are needed. Otherwise Makuro shows the formula estimate rather than inventing one. Because slopes need weigh-ins at least seven days apart inside one stretch, a missed day every six days or less prevents an adaptive estimate. These thresholds are engineering choices. Implausible fits beyond 0.6–1.6× the starting estimate get limited weight as a guard against logging errors.

Limits. The conversion of 7,700 kcal per kilogram fits people with more than about 30 kg of body fat and overstates stored energy for leaner people.⁸ Over two weeks, most weight change is lean tissue and water, not fat.⁹ This matters little near a stable weight and more when weight is changing quickly. Glycogen is stored with about three to four times its weight in water,¹⁰ and weight follows a weekly rhythm, typically higher after weekends.¹¹ A few days of eating more or less can therefore hide or exaggerate real change. Expenditure can fall about 10% in the first month of calorie restriction,¹⁴ and body weight responds slowly to lasting changes in intake.⁷ A window spanning a diet change mixes two states. Consistent under- or over-logging carries straight into the estimate. That also keeps the estimate in the same units as your logging, which is what the calorie target uses. Makuro's estimator has not been validated against doubly labeled water.

The weight chart shows a smoothed trend (an exponentially weighted moving average²³) for readability; the estimator fits the daily weigh-ins directly.

• Explicit fasting: an intentional zero-calorie fasting day is kept as zero intake. Fasting means a full day without energy intake, not simply skipping breakfast.
• Partial or missing: these days split the fitted stretches. We do not count them as zero or fill them with the average of other days. Without enough recent complete data, the estimator returns the formula estimate.
• Today: today's unfinished intake is excluded. A current morning weigh-in can close the preceding interval.
• Unmarked low intake: the 500 kcal screen is a logging heuristic, not a biological cutoff or a recommended minimum. It cannot detect a partially logged day above that threshold. Mark incomplete days honestly.

Log all energy intake, including drinks, oils and snacks. Weigh under similar conditions, preferably in the morning. Consistent logging makes the estimate more useful but does not make it a laboratory measurement. A fasting mark records what happened; it is not a recommendation to fast or proof that a day matched your nutrition target.

Editing day marks or correcting past food records refreshes expenditure, not the week's calorie budget. Automatic targets update weekly when usable data permits. An existing adaptive budget is retained during a formula fallback; explicit goal edits or manual target recalculation can update it sooner.

Targets combine the expenditure estimate with your goal and chosen pace. Defaults use approximately −0.6% body weight per week for fat loss, −0.2% for recomposition, +0.25% for gain, and zero for maintenance. Conversion at 7,700 kcal/kg is an initial planning approximation, not a predicted trajectory. Athlete studies support considering slower loss to preserve lean mass; their results do not define a universally safe pace.²⁰

The deficit is capped at 25% of estimated expenditure and the surplus at 15%, within absolute limits. A stored pace cannot reverse the selected goal. The 1,200 kcal floor is a software guardrail, not a guarantee of adequate nutrition for an individual.

Protein preferences scale with body weight and goal. Resistance-training research found an average breakpoint near 1.6 g/kg/day; this is not an individual requirement or evidence that everyone benefits from the highest setting. ²¹ Recommendations of 2.3–3.1 g/kg during contest preparation refer to fat-free mass, not total body weight. ²² Makuro's highest options are preferences, not validated prescriptions for every user.

Fat starts near 25% of energy with a weight-based preference. Neither that percentage nor 0.5 g/kg guarantees hormonal health or essential-fat adequacy. Fat and protein are fitted inside the calorie budget, reserving 20% of energy for carbohydrate; carbohydrate then fills the remainder. That reserve is a product allocation, not a physiological minimum. Energy accounting uses protein/fat/carbohydrate at 4/9/4 kcal/g.

Review progress weekly. Adult equations and athlete studies do not validate this program for children, pregnancy, breastfeeding, eating disorders or clinical conditions. Individual targets in these situations need qualified professional input.