An In-Depth Look at PlateLens's Adaptive Energy Expenditure Algorithm

A TDEE calculator guesses your daily burn from your age, height, weight and an activity level, and for any one person that guess can be off by 20-30%. An adaptive energy expenditure algorithm replaces the guess with your own results: the food you log and what your weight actually does. PlateLens runs that algorithm for you automatically every week, from every meal you log and your weigh-ins. Here is how it works, step by step.

Learns from

Your logged meals + weigh-ins

Updates

Every Sunday, after a first check when you switch it on

Each update

Up to 5%

Target check-in

Every 14 days, new targets as you progress

Food log it learns from

±4.4% calorie error with the meal typed in (Dietary Assessment Initiative, 180 weighed meals)

Why Static Calorie Calculators Fail

If you have ever used a TDEE calculator -- the kind that asks for your age, weight, height, and "activity level" -- you have probably noticed something: the number it gives you does not quite work. You follow the target, but your weight does not budge. Or it drops too fast. Or it yo-yos unpredictably.

This is not your fault. Static formulas like Mifflin-St Jeor and Harris-Benedict were designed for population averages, not individuals. They can be off by 20-30% for any given person. They do not account for your unique metabolism, your adaptive thermogenesis, how your body responds to different macronutrient ratios, or the dozens of other factors that make your energy expenditure uniquely yours.

Wearable devices attempt to solve this problem, but they bring their own issues. Heart rate-based calorie estimates are notoriously inaccurate for strength training, and step-based estimates miss everything that happens when you are not walking. Studies consistently show that even the best wearables can be off by 25-90% on calorie burn estimates.

The fix is to measure the result instead of guessing the inputs. That is exactly what PlateLens's algorithm does.

How PlateLens's Expenditure Algorithm Works

The core idea is simple and powerful. Instead of estimating your calorie burn from the outside (formulas, heart rate, accelerometers), the algorithm reads it from the result: what actually happens to your weight when you eat the way you eat.

The fundamental equation is the energy balance equation, rearranged:

Know how much you eat (Calories In) and how your weight changes over time (Change in Stored Energy), and you can solve for how much energy your body actually burns (Calories Out). Real intake, measured against real outcomes.

The intake side is where PlateLens shines. Every meal you log, from a photo, a sentence, your voice or a barcode, arrives with its calories and macros in seconds through AI food recognition, and the algorithm learns from every one of them.

That is also why the big holiday dinners belong in the log: the algorithm learns only from the meals it sees, so the weeks it reads best are the ones where Thanksgiving is logged too. A photo of the plate or a sentence like “turkey, stuffing and a slice of pie” is enough; our stuffing guide shows how much one side can vary.

That matters, because an adaptive expenditure is only as accurate as the food log it learns from, and PlateLens learns from the most accurate one: in its published validation study, the Dietary Assessment Initiative measured PlateLens at ±4.4% calorie error with the meal typed in across 180 weighed meals, and the open-source Foodvision Bench reproduced that result on 231 more meals on its public GitHub leaderboard.

How It Runs, in Five Steps

1. Switch it on once. In Your expenditure, turn on Adjust with my logs and weigh-ins. It is free on every plan, on iPhone, Android and the web.
2. A first check right away. PlateLens reads your recent weeks shortly after you switch it on, so if you already log and weigh in, your first dynamic expenditure can arrive soon after.
3. An update every Sunday. Once each week closes, PlateLens compares the weight change your logged food predicts with the trend of your weigh-ins, across up to the last three weeks, and solves for your expenditure.
4. Steady steps. Each update moves your dynamic expenditure by up to 5%, so it follows real change without chasing a single salty meal.
5. New targets as you progress. Every 14 days, a check-in on the phone's Home suggests targets built on your dynamic expenditure whenever your progress calls for it, and one tap on Apply New Targets puts them to work.

The Self-Correcting Loop

What sets this approach apart from static calculators is that it is self-correcting. Each week the algorithm compares two numbers: the weight change your logged intake predicts, and the change your weigh-ins actually show. When they differ, it moves your expenditure toward the value that fits both.

Ate more than usual for a few weekends and your weight barely moved? The algorithm learns that you burn more than it estimated. Started a desk job and your weight is creeping up on the same meals? It learns that your expenditure dropped, and your next 14-day check-in suggests targets built for where you are now.

The loop keeps learning week after week, in steps of up to 5%, following real changes in your life as they show up in your log and on the scale.

What Makes Our Algorithm Stand Out

Responsive Without Over-Correcting

One of the hardest challenges in adaptive algorithms is balancing responsiveness with stability. React too quickly, and the system chases noise -- adjusting targets based on a single salty meal or a bad night's sleep. React too slowly, and users feel like the algorithm is ignoring real changes in their lives.

PlateLens balances the two with limits you can see. It reads your weight trend across up to three weeks of weigh-ins with a robust slope, so one unusual morning on the scale barely moves it and implausible jumps are left out. Each weekly update moves your expenditure by up to 5%, and by half that when a week only just meets the minimums.

The result: an expenditure that moves smoothly from week to week, and new targets you put to work with a single tap.

Built for Real Life

Life happens. You travel, you have a busy week, you skip the photo at lunch. PlateLens is built for exactly that, and our app-by-app comparisons show how it lines up against other trackers.

The algorithm learns from weeks with meals logged on at least 4 days and at least one weigh-in. A week with gaps keeps your last dynamic expenditure, and the next week with enough data picks up right where it left off: a dynamic expenditure never expires. When a day shows much less food than you usually log, PlateLens asks you about it once, in Review your week, so it learns from complete days.

Fast Logging, Better Learning

The algorithm learns from what you log, so PlateLens makes logging the fast part. Snap a photo of your plate or describe the meal in a sentence, and the calories and macros are in your log in seconds through AI photo recognition. If something is off, say so in plain words, such as "it was a small bowl", and the entry updates.

Fast logging means complete days, and complete days give the algorithm consistent data to learn from. Better data in, better estimates out.

Expenditure Modifiers: Two Precision Settings

Alongside the weekly learning, Your expenditure gives you two precision settings, both free on every plan.

Balance labels where your number comes from (With your measurement, or Dynamic · from your logs and weigh-ins), and Your expenditure over time keeps every week's value. These settings shape the expenditure in Balance and the targets PlateLens suggests, while your daily ring, Daily Score, leagues and Apple Watch keep the same numbers.

How the Algorithm Powers Your AI Coach

The expenditure algorithm also powers PlateLens's personal AI nutrition coach.

Ask the Coach about your expenditure and it knows which source you use (the estimate, your measurement or your dynamic expenditure) and how your weekly learning is going, and it points you to the right setting. The same expenditure feeds Balance, your PDF reports and PlateLens in ChatGPT and Claude, so every answer starts from the same number.

This is the power of combining an adaptive algorithm with AI coaching: the algorithm does the math, and the Coach turns it into plain-language guidance. You never need to read graphs or work through metabolic equations; your Coach explains what your numbers mean.

The Full PlateLens Tracking Stack

The adaptive expenditure algorithm is one piece of a complete system designed to make nutrition tracking as effortless and effective as possible:

• AI Photo Recognition: Snap a photo of any meal, or describe it in a sentence, for calorie and macro estimates in seconds. No manual searches. No weighing food.
• Adaptive Expenditure Algorithm: Learns your expenditure automatically every week from your logs and weigh-ins, and checks your targets every 14 days to suggest new ones as you progress.
• Personal AI Coach: Interprets your data, provides personalized guidance, and adapts to your patterns and preferences.
• Expenditure Modifiers: Your own resting measurement and Count digestion, for even more precision.
• Apple Health and Health Connect: Sync your weigh-ins, steps and other health data for a complete picture.
• Streaks & Motivation: Daily streaks, achievements, and milestones to keep you consistent -- because consistency is what the algorithm needs most.
• Water Tracking: Complete hydration monitoring integrated into your daily health picture.

Every piece reinforces the others. The streaks keep you logging consistently. Consistent logging gives the algorithm better data. Better data gives the coach more accurate guidance. Better guidance keeps you motivated. It is a virtuous cycle designed from the ground up.

Frequently Asked Questions