Four estimate methods. Twelve profiles. One live site check.

Download the evidence

The complete inputs, outputs, formulas, and rounding rule are published so the result can be checked independently.

License: CC BY 4.0. The profiles are synthetic and do not represent real people or clinical recommendations.

External implementation check - captured August 20, 2026

Same inputs and same factor produced the same maintenance estimate

Profile P08 is a 35-year-old male, 6'0", 210 lb, with no body-fat input. At the identical 1.375 activity factor, all three captured calculators returned 2,648 kcal/day. That agreement is expected because each uses Mifflin-St Jeor and the same multiplier for this setup.

What the check shows: a correctly implemented equation is not the moat. The useful differences are how a tool labels activity, exposes uncertainty, sets deficit guardrails, explains its assumptions, and helps a person test the estimate against observed data.

Why two named tools are not in the output row

This is a transparent implementation check, not an accuracy trial. None of these calculator outputs was compared with measured energy expenditure.

Benchmark table

Activity factor is applied only to the two BMR equations. The weight-only methods are exactly 10 or 15 calories per pound. The blend is the arithmetic mean of all four displayed estimates.

Where the spread gets widest

These bars show each profile's full range relative to the largest range in the dataset. Larger body weights widen the distance between weight x 10 and weight x 15 by construction. Activity level does not affect those shortcuts at all.

How the test was run

1. We created six female and six male synthetic profiles spanning ages 24 to 60, 130 to 250 lb, 5'3" to 6'2", and four activity bands.
2. Imperial values were converted with 0.45359237 kg per pound and 2.54 cm per inch.
3. Mifflin-St Jeor and revised Harris-Benedict BMR were calculated at full precision, then multiplied by the listed activity factor.
4. The two sheet-derived shortcuts were calculated from body weight alone.
5. The blend was calculated from unrounded component values. Displayed values were rounded only at the end.

The profiles were selected to expose sensitivity across age, size, sex, and activity. They are not population-weighted, so the average spread describes this test set rather than all adults.

What this benchmark can and cannot tell you

It can audit model behavior

The table shows exactly how changing model inputs changes the result. It also makes an important design flaw visible: averaging a detailed equation with a weight-only shortcut does not automatically make the answer more accurate.

It cannot measure anyone's metabolism

No synthetic benchmark determines true energy expenditure. The equations estimate population averages, and the activity factor remains a large assumption. A person's consistent intake and weight trend can eventually provide more individualized evidence.

It does not rank competing websites

Many calculators use similar equations but make different choices about activity, body fat, rounding, calorie floors, and result presentation. This test isolates model behavior rather than declaring a site-level winner.

Sources and review status

• Mifflin et al. (1990), a new predictive equation for resting energy expenditure
• Roza and Shizgal (1984), revised Harris-Benedict equations
• Frankenfield et al. (2005), comparison of predictive resting metabolic rate equations
• NIDDK Body Weight Planner, a dynamic model for longer-term planning

Analysis and software by Builderfox LLC. This benchmark has not yet been reviewed by a registered dietitian or clinician. Report a reproducibility issue through the contact page or review the correction log.

Creation note: Software was used to calculate and cross-check the dataset. AI-assisted research and drafting were manually reviewed against the linked primary sources and the published data files.