What you actually spend: TDEE explained

Every plan has to start somewhere, so most apps hand you a single number and call it your metabolism. In reality that number is a reasonable guess, and the more reliable way to know your true energy expenditure is to watch what actually happens to your weight as you eat.

The four pieces of your daily number

Total daily energy expenditure, or TDEE, is usually broken into four parts.

• Resting energy (BMR/RMR) — what your body uses just to keep running, roughly 60-75% of the total for most people.
• The thermic effect of food (TEF) — the energy cost of digesting and processing what you eat, about 10% on a mixed diet.¹
• Non-exercise activity (NEAT) — fidgeting, walking around, standing, taking the stairs, all the incidental movement that isn't a workout.
• Exercise activity — deliberate training, typically the smallest slice for people who aren't athletes.

That decomposition is tidy on paper, but each piece varies a lot between people and even day to day for the same person. TEF alone isn't a flat 10% regardless of what you eat — meals higher in protein tend to cost more energy to process than meals higher in fat, so two days with identical calories can have genuinely different thermic costs depending on what those calories were made of.¹

NEAT is the wildcard

Of the four pieces, NEAT is the one that swings the most and gets the least attention. In a classic overfeeding study, researchers fed volunteers an extra 1,000 calories a day for eight weeks and watched what happened to their energy expenditure. About two-thirds of the rise in total expenditure came from people unconsciously moving more — more fidgeting, more incidental activity — and how much someone's NEAT increased closely predicted how well they resisted gaining fat.² Later work from the same research group found NEAT can differ by roughly 2,000 calories a day between two people of similar size, depending on their job and how they spend downtime.³ That's a bigger swing than most formulas' entire activity-factor range, and it's largely invisible — nobody feels their own NEAT rising or falling.

This is one of the main reasons two people with the same height, weight, age, and stated activity level can eat similar amounts and see different results. A formula can't see their fidgeting.

Formulas are a starting point, not a verdict

Equations like Mifflin-St Jeor are among the better population-level tools researchers have, and reviews comparing predictive equations against measured resting energy find it performs about as well as any of them for community-living adults.⁴ "About as well as any of them" still means real individual error — the same review shows accuracy sliding for specific groups, like people with obesity or certain age ranges, even for the top-performing equation.⁴ A formula gives you a sensible number to start eating around. It was never meant to be a verdict on your personal metabolism.

There's also a common shortcut worth retiring: the idea that a pound of fat is worth an exact, fixed 3,500 calories, so you can reverse-engineer your precise deficit down to the calorie. Kevin Hall's dynamic energy-balance modeling, the basis for the NIH's Body Weight Planner, shows that expenditure itself shifts as you lose weight and as your body adapts — so the old static rule tends to overestimate long-term results, and the real relationship between a deficit and weight lost bends and slows over time rather than following a straight line.⁵ A short-window approximation like 7,700 kcal per kilogram — the figure Kinra's engine uses to translate a few weeks of energy balance into a rate of change — is a reasonable engineering approximation for a short window, not a precise physical constant you can extrapolate over months.

Metabolic adaptation adds another layer. In a well-known follow-up of contestants from a weight-loss competition, resting metabolic rate was still measurably suppressed relative to expectations six years later, even as most participants had regained much of the weight — a reminder that expenditure can shift in ways formulas don't anticipate, and that the size of the shift varies a lot from person to person.⁶ That single study doesn't mean everyone's metabolism adapts the same way after dieting; it means adaptation is real, individual, and worth accounting for rather than assuming away.

Why the better approach is to learn it, not guess it

Put those pieces together — variable NEAT, diet-dependent TEF, formula error, and adaptation over time — and it's clear why no equation, however well-built, can hand you a number that stays right for months. A more dependable approach, and one that's gained real traction in serious nutrition-tracking tools, is to treat the formula as a first guess and then continuously refine it using what actually happens: your logged intake compared against your weight trend over several weeks.

This is the same dynamic-energy-balance thinking behind Hall's NIDDK work, and it's roughly the approach some other nutrition-tracking apps have built their methodology around — start with an estimate, then let real outcomes correct it.

Kinra works this way too. It starts you on a Mifflin-St Jeor estimate times an activity factor, then smooths your daily weigh-ins into a steadier trend line so one noisy morning never moves your plan. Over rolling two-to-four week windows, it compares what you actually ate against how your trend weight actually moved, and blends that learned number in with the original formula estimate — leaning on the learned figure more once there's enough data to trust it, and holding steady when the data is thin or noisy. When an adjustment is warranted, it's a small, damped nudge, capped and rounded, never a sudden swing.

None of this claims to measure your metabolism in a lab or predict your weight months from now — it's a moving best estimate, updated by your own numbers instead of someone else's average. If you have a medical condition, are pregnant, or have a history of disordered eating, it's worth working with a qualified clinician alongside any nutrition app, including this one.