How the coach works
The weekly review, in plain English
Once a week, on the day you pick, FuelKey works out what your body really burns from your own logs and weight, then suggests at most a 150-calorie change to your plan as a real food swap. You approve every change. The key difference from other apps: it only learns from days you finished, so a forgotten dinner can’t trick it into cutting your food.
What it does each week
- It only counts days you finished.A day counts once every planned item is ticked eaten or skipped. If you forget to log dinner, that day looks like you ate less than you did. A coach that counts every day would decide you burn less and cut your food. Ours leaves unfinished days out.
- It reads your weight trend, not today’s number.Daily weight swings with water, salt and food. FuelKey smooths it into a trend, and no single day can move that trend by more than 1.2%, so one salty dinner doesn’t look like fat gain.
- It ignores your first week.The first week of a new plan is mostly water and stored carbs moving around, not fat. Those swings are set aside so week 2 doesn’t bring a scary, wrong correction.
- It starts from the formula and shifts to your data.Day one uses the standard formula for your height, weight, age and activity. As finished days add up, your own numbers take over: fully after 21 finished days. It needs at least 4 finished days and 3 weigh-ins in the last three weeks before it uses your data at all.
- It measures your real burn.Over the last three weeks: what you ate on finished days, adjusted by how your weight trend moved (about 3,500 calories per pound). Eat 3,000 a day while holding steady, and you burn about 3,000.
- It changes your plan a little, with real food, and asks first.Any change is capped at 150 calories a week. Protein stays fixed; one carb or fat portion is resized (for example, ¾ cup more rice at lunch). Nothing changes until you approve it.
How we tested it
Before building it, we ran the design against 2,000 simulated people over 8 weeks, each with a known true burn and a formula estimate off by up to 15%. Then we made the logging sloppy on purpose: a share of days missing a meal, the way real life goes.
Typical (median) error in the burn estimate after 4 weeks, in calories a day (lower is better):
| Days missing a meal | 0% | 10% | 20% | 35% |
|---|---|---|---|---|
| A coach that counts every day | 173 | 195 | 249 | 380 |
| FuelKey’s coach | 178 | 172 | 164 | 163 |
With 1 in 5 days missing a meal, the count-every-day coach ended up telling people to eat about 212 calories a day less than they should. FuelKey’s was off by about 22. The sloppier the logging, the bigger the gap.
We also tested the 150-calorie weekly cap: the target reaches the right level in about a week for most people, and within 3 weeks for 9 in 10. A smaller cap adds about a week; a bigger one saves about one but makes the plan jumpier.
Honest caveats: the comparison coach is a simple model of the “count every day” approach, not any specific app’s private formula, and the simulated noise levels are reasonable assumptions rather than measurements from real users. The direction of the result is solid; the exact numbers aren’t.
What it won’t do
- Change your plan without asking.
- Move more than 150 calories in a week.
- Lower your protein.
- Learn from days you didn’t finish.
FuelKey records what you tell it and does the arithmetic. It isn’t medical or dietary advice; if you have a medical condition, check changes with your doctor.