This page explains the science behind our recommendations. It is not medical advice. Consult a healthcare professional before starting any diet or exercise program.

Methodology

Every recommendation ScanFitPro makes traces back to published, peer-reviewed research. This page shows our work — the studies, formulas, and guidelines that drive every number you see.

How the estimate works

ScanFitPro uses a vision-language model (Claude Sonnet) to analyze two photographs — a front-facing and a side-profile — alongside your height, weight, age, and biological sex. The model estimates body fat percentage and returns a confidence range.

Why two photos?

A front photo captures overall body shape and muscle definition. A side photo captures abdominal depth — the sagittal dimension that a front photo alone cannot resolve. Research shows that combining two poses improves prediction accuracy (R² for percent fat improved from 0.39 to 0.64 in Wong et al. 2021).

Confidence range

We report a ±5–7 percentage point confidence band based on Bland-Altman Limits of Agreement from published validation studies. Purpose-trained CNNs achieve 2–4.5% error vs. DEXA; our VLM approach is less precise, so we use a wider, more honest range.

What it can tell you

  • Overall body fat percentage (within ±5–7% in real-world conditions)
  • Broad fat distribution pattern (android vs. gynoid)
  • Regional fat estimates (abdominal, gluteofemoral) — moderate accuracy
  • Changes over time — relative tracking is more reliable than absolute estimation

What it can't tell you

  • Visceral fat volume (requires MRI/CT/DXA)
  • Hydration status or lean mass composition
  • Bone mineral density
  • Metabolic health markers — body shape correlates with risk but does not diagnose it

Key citations

Reading your number in context

Not everyone carries body fat the same way, and a single percentage can mean different things for different people. We don't ask about family background or ethnicity — the estimate uses your photos, height, weight, age and (optionally) biological sex, and nothing else. The research below is here so you can interpret your own number, not because we adjust it behind the scenes.

Why we don't ask about ethnicity

The underlying science is real — Asian populations carry more visceral fat at a given BMI (Wang 1994, WHO 2004), and Black adults carry more lean mass at equal BMI (Heymsfield 2016). We still don't ask. It was the single most-abandoned question in setup, it arrived before we'd shown you anything useful, and personalizing a target from it is a short step from stereotyping. Read the ranges below as context and talk to a clinician about your own risk.

WHO Asian-BMI cutoffs

For Asian populations, observed health risk starts at BMI 22–25 kg/m² (vs. 25+ for Europeans), and high risk at BMI 26–31 kg/m² (vs. 30+). Public health action points are set at 23.0, 27.5, 32.5, and 37.5 kg/m².

Age-related changes

Body composition changes with age: sarcopenia (muscle loss) accelerates after 60, and menopause causes lean mass to decline ~0.5% per year while fat mass increases ~1.7% per year. Metabolic rate is stable from 20–60, then declines ~0.7% per year (Pontzer 2021). Worth knowing when you compare yourself to a population average.

Published body fat reference ranges

Synthesis derived from WHO Asian-BMI cutoffs, ethnicity-body fat differences, age-related composition changes, and ACE/ACSM norms. Not a direct quote from any single source, and not a target the app sets for you — you choose your own goal during setup.

AgeSexStandardAsian-phenotypeHigher-leanHigher-%fatSarcopenic
20–39Men10–17%9–15%12–19%11–18%
20–39Women18–24%17–22%20–26%19–25%
40–59Men14–20%12–18%16–22%15–21%
40–59Women21–27%19–25%23–29%22–28%22–28%
60–79Men16–22%14–20%18–24%17–23%17–24%
60–79Women23–30%21–28%25–32%24–31%24–32%
≥80Men18–25%16–23%20–26%19–25%19–26%
≥80Women25–33%23–31%27–34%26–33%26–34%

Key citations

Daily calorie burn

Resting energy expenditure

We calculate resting energy expenditure using the Mifflin-St Jeor equation (1990), validated as the most reliable predictive equation by Frankenfield et al. (2005). This is multiplied by an activity factor (1.2 sedentary → 1.9 very active) to estimate total daily energy expenditure (TDEE) — the daily calorie burn shown on your home screen.

When biological sex isn't provided

The Mifflin-St Jeor equation differs by sex by roughly 166 kcal. If you haven't told us, we average the two equations rather than defaulting to one, and we say so on the estimate — picking either would bias your number in a direction we have no basis to choose.

Caloric deficit

Per AHA/ACC/TOS 2013 guidelines, a 500–750 kcal/day deficit produces ~0.5–1% body weight loss per week. We enforce a safe calorie floor (1500 kcal/day for men, 1200 for women, 1350 for undisclosed) and report the effective deficit honestly when the floor binds.

Protein

Protein is set at 1.6–2.2 g/kg of lean body mass per day during a caloric deficit to preserve muscle (Morton 2022). For adults 65+, we lower this to 1.2–1.6 g/kg of total body weight to reflect the older adult evidence base.

Key citations

Recommended daily steps

Your home screen shows a daily step target built from the Tudor-Locke & Bassett (2004) pedometer classification — five activity tiers (sedentary, low active, somewhat active, active, highly active) that map directly onto the activity level you set in your profile.

How the target is set

The target is the floor of the tier one step above your own — sedentary steps up to 5,000/day, active steps up to 12,500/day, and so on. Already in the top tier? The target holds there rather than climbing indefinitely.

The body fat nudge

If your latest estimate is still meaningfully above your goal, we add 1,000 steps — the same NEAT (non-exercise activity thermogenesis) lever this page's progress suggestions already recommend for a plateau, folded directly into the number instead of left as a suggestion nobody sees.

Why weight isn't in the target

Body weight doesn't change how many steps count as "enough" — that's a step-count question, not a body-mass one. It changes how many calories those steps burn, so we use it only for the secondary calorie estimate next to your target, using a commonly cited ~30-40 kcal per 1,000 steps for a typical adult.

Key citations

Complete citation list

All 16 citations, organized by section. Click any citation to see the full reference and DOI/PubMed link.

Body fat estimation

Key citations

Reading your number in context

Key citations

Daily calorie burn

Key citations

Recommended daily steps

Key citations