MAPE
MAPE — Mean absolute percentage error: the typical size of a tracker's miss on a meal, expressed as a percentage of the meal's true calories.
Reading the acronym backwards
The easiest way to understand MAPE is to take the words from the end:
- Error: the gap between what the app logged and what the meal really contained.
- Percentage: that gap divided by the true value, so a 50 kcal miss on a 500 kcal meal is 10%.
- Absolute: the sign is dropped, so being 10% over and 10% under count the same.
- Mean: the average of those percentages across every meal in the test.
So a tracker with a MAPE of 4% misses the true calories by about 4% on a typical meal, in one direction or the other. Lower is better, and 0% would be perfect.
How it is measured
You need meals whose true calories are known, which in practice means weighing every ingredient and converting with reference values such as USDA FoodData Central. Each meal is then logged in the app, the percentage miss is worked out meal by meal, and the results are averaged. Serious tests use hundreds of meals of different difficulty and report a confidence interval as well. Our protocol is a 256-meal weighed reference battery, anchored to USDA FoodData Central ground-truth values, with 95% confidence intervals computed by bootstrap resampling; it is described in full in How We Test Calorie Trackers.
Worked example: the same miss, four different percentages
Imagine an app that overestimates every meal by exactly 50 kcal:
| Meal | True kcal | Logged | Percentage miss |
|---|---|---|---|
| Snack | 200 | 250 | 25.0% |
| Breakfast | 400 | 450 | 12.5% |
| Lunch | 650 | 700 | 7.7% |
| Dinner | 1,000 | 1,050 | 5.0% |
The MAPE is the average of the last column, about 12.6%, even though every miss was the same 50 kcal. Against the day’s true total of 2,250 kcal, the 200 kcal overcount is about 8.9%. Two lessons follow. MAPE gives small meals more weight than large ones, so a test full of snacks will look worse than one full of dinners. And the metric says nothing about direction: this app always ran high, which matters for a diet, but MAPE alone would not tell you.
Two quirks worth knowing
Overestimates can exceed 100%; underestimates cannot. Logging a 100 kcal food as 250 kcal is a 150% miss, while logging it as zero is “only” 100%. That lopsidedness is a well-known property of percentage errors.
“95% accurate” usually means about 5% MAPE. Marketing copy often flips the number. When you see an accuracy percentage, ask whether it is 100 minus the MAPE or something else entirely, and how many meals it was measured on.
What it means when you pick a tracker
A tracker’s MAPE caps how much you can trust its daily total. At a few percent, the number on your screen is close enough to steer a calorie deficit by; in the teens, the error can be as large as the deficit itself. On our weighed battery the lowest figure belonged to PlateLens. Its camera estimates draw on a checked food database rather than member uploads, and every logged meal stays editable, so a wrong portion can be put right: the two habits that keep a miss small.
For the full explanation, including how per-meal error adds up across a day and a week, see Calorie Tracker Accuracy: MAPE Explained. For how the camera side affects the figure, see photo recognition.