Evidence & Methodology
Research-backed, AI-delivered
Every number, target, and insight inside Verdant is derived from published academic research studies and delivered to you through modern artificial intelligence. We do not guess. We read the literature, encode the consensus, and let AI translate it into calm, actionable guidance tailored to your patterns.
Energy & body-composition equations
- Basal Metabolic Rate (BMR)Mifflin–St Jeor equation (1990), validated against indirect calorimetry and endorsed by the Academy of Nutrition and Dietetics.
- Total Daily Energy Expenditure (TDEE)Harris / FAO / WHO / UNU physical-activity level (PAL) categories — Sedentary (1.2) through Very Active (1.9).
- Calorie targetsLose: −20% (~500 kcal deficit, ≈0.5 kg/week); Maintain: TDEE; Gain: +10–15% surplus (Aragon & Schoenfeld, 2013).
- Protein1.2–1.8 g/kg body weight depending on goal and life stage, calibrated from the Morton et al. (2018) systematic review.
- Fat & carbohydrate splitAcceptable Macronutrient Distribution Ranges (AMDR, Institute of Medicine) with preference-aware floors and ceilings.
- Fiber14 g per 1 000 kcal (Slavin, 2013; Academy of Nutrition and Dietetics position paper).
- Hydration35 ml/kg body weight (EFSA Scientific Opinion, 2010).
- BMI categoriesWHO standard thresholds (underweight <18.5, healthy 18.5–24.9, overweight 25–29.9, obese ≥30).
AI coaching & behavioural insights
Your AI coach does not invent advice. It observes your actual logging data — meal timing, macro distribution, hydration, and week-over-week trends — and frames feedback using the Fogg Behaviour Model and habit-loop research (Wood & Rünger, 2016). Small, specific, recent suggestions consistently outperform absolute scoring for long-term adherence.
The correlations between exercise energy expenditure and food intake (shown in the Energy Balance card) use Spearman rank correlation over a rolling 28-day window once enough paired wearable-and-log days exist. This is a standard non-parametric measure suited to day-to-day lifestyle data.
What the AI cannot do
Predictive equations carry an inherent ±10–15% margin of error versus measured resting metabolic rate. Activity multipliers are population averages. The app does not account for body-composition phase, thyroid status, medications (for example GLP-1 agonists, which can lower resting metabolic rate by ~5–10%), or adaptive thermogenesis. Always treat numbers as informed starting points, not clinical prescriptions.
Bottom line: We build on peer-reviewed nutrition and behavioural science, then use modern AI to make that science personal, timely, and kind. The research gives us the guardrails. The AI gives you the conversation.