Methodology

Most apps set your “maintenance calories” from a formula and never revisit it. The problem: those formulas are population averages, and your body adapts — metabolic rate falls during a diet by more than the loss of tissue alone would predict.⁴ A number fixed on day one drifts wrong within weeks.

Makuro instead learns your expenditure from the one signal that can’t lie: the energy-balance identity. Over any period, energy stored equals energy in minus energy out — ΔEnergy = Intake − Expenditure. Rearranged, if we know your intake and how your body-energy stores changed (via weight), we can solve for expenditure.¹ This is the same “intake-balance” principle used in metabolic-ward research.

How Makuro applies it, step by step:

• A rolling 28-day window. Recent behavior only — old data doesn’t describe your current metabolism.
• A smoothed weight trend, not the scale. Daily weight is mostly water, sodium, and gut contents. We fit the trend (see §2) and take its slope by least-squares regression.
• Recency-weighted intake. Your average intake is weighted toward recent days (10-day half-life), and implausibly low days (likely un-logged) are excluded rather than counted as fasting.
• Converting weight change to energy. The weekly trend slope is converted to an energy flux using the classic tissue energy density of ~7,700 kcal per kg. Expenditure ≈ average intake − (trend slope × 7,700).
• Blending with a baseline by confidence. Before you have enough data, we lean on the Mifflin–St Jeor equation — the predictive formula validated as most accurate against indirect calorimetry.⁵⁶ As clean weeks of data accumulate, the estimate shifts toward your measured energy balance and away from the formula.

Why not just trust “3,500 kcal = 1 lb”? Because as a predictive rule it’s wrong: the body’s weight response to an intake change is dynamic and slow, and static rules badly overestimate loss.²³ Makuro uses the constant only in reverse — to read energy out of an already-observed trend — which sidesteps that error, and the confidence-blend and windowing absorb what remains.

What about Apple Watch? Wearables' calorie totals carry large errors — a Stanford validation found energy-expenditure errors of roughly 27–93% across devices, even when heart rate was accurate.¹² So Makuro never feeds wearable calories into the energy-balance measurement. But measured movement does beat a self-reported activity level — so if you opt in, your activity data informs only the formula phase, and its influence fades to zero as your logged data takes over. Measured beats self-reported at the start; the scale beats the wrist in the long run.

A single weigh-in can swing a kilogram overnight on water alone, so the raw scale number is a poor progress signal. Makuro fits an exponentially weighted moving average — each day’s weight updates a running trend, with recent days weighted more and older ones decaying smoothly. This is the long-established approach to separating a real body-weight trend from daily noise.⁷

Your weekly rate of change is read off that smoothed trend, not off two raw weigh-ins. The practical payoff: a salty dinner doesn’t read as a gain, and a dehydrated morning doesn’t read as progress. It’s the same reason your expenditure estimate (§1) is stable rather than lurching day to day.

Targets are built on your current, adaptive expenditure (§1), then adjusted for your goal:

• Calories. A moderate deficit to lose fat; a modest surplus to gain; maintenance to hold steady. Recomposition sits just below maintenance — a slight deficit, with high protein and training doing the compositional work. Deliberately moderate — see the rate note below.
• Protein: your preference within the evidence-based range. Meta-analysis puts the average muscle-building benefit near 1.6 g/kg/day with an upper confidence bound around 2.2,⁸ and lean dieters benefit from higher intakes — roughly 2.3–3.1 g/kg of fat-free mass — while cutting.⁹ So Makuro lets you pick where you sit (Low / Moderate / High / Extra High), and the range itself is goal-aware: about 1.6–2.8 g/kg while losing or recomping, 1.4–2.4 g/kg while maintaining or gaining. Higher protein in a deficit does double duty, preserving lean mass while you lose fat.
• Fat: ~25% of calories, keeping essential-fat needs met, with carbohydrates filling the remainder to fuel training. Macronutrient energy values are the standard 4 / 9 / 4 kcal per gram for protein / fat / carbohydrate.

On how fast to lose. Faster isn’t better: at a 0.7%-per-week loss rate, athletes gained lean mass and lost more fat than a group losing 1.4% per week.¹⁰ The evidence-based sweet spot for preserving muscle is roughly 0.5–1% of body weight per week.⁹¹¹ So Makuro sizes your deficit to your bodyweight to land in that band — rather than applying one flat number to a 50 kg and a 95 kg person alike. Prefer a specific pace? Set it in the app and Makuro converts it into your daily target the same way — still clamped to a sensible band.

Honesty is part of the method. Predictive equations are population estimates; individual metabolic rate varies. Energy-balance estimation needs about 2–4 weeks of consistent weight and intake logging before it’s personalized — until then you’re seeing an informed formula, not your measured expenditure. The ~7,700 kcal/kg constant is an approximation, and the first week or two of a new diet includes glycogen-and-water shifts that no model reads perfectly; Makuro’s windowing and smoothing limit, but don’t erase, that effect. Makuro is a tracking tool, not medical advice.