My Calorie Tracker Says Maintenance, But I'm Still Losing — What's Going On?
If your app says you're eating at maintenance and the scale keeps dropping, the most likely answer is database error. Here's the math on why ±18% MAPE makes a real deficit invisible — and how PlateLens changes the picture.
Quick verdict
If your tracker says maintenance and your scale says deficit, your tracker is wrong — almost certainly because database error is overestimating your intake. The fix is either an architectural switch (photo-AI logging like PlateLens, ±1.1% MAPE) or a search-and-log app with a tighter database (Cronometer, ±5.2%).
The math behind the mystery
This question comes up constantly. The user is logging diligently. The app says they’re eating at maintenance. The scale says they’re losing weight. Something has to give.
The thing that gives is almost always the app’s intake estimate. Here’s why.
A typical weight-loss deficit sits around 250 calories per day. On a 2,000-calorie intake, that’s a deficit of 12.5%. If the app’s MAPE is wider than 12.5%, the deficit literally cannot be distinguished from noise — the app can show “maintenance” while you’re cleanly in a 250-calorie deficit, because the noise band is wider than the signal.
What the MAPE numbers actually look like
From our 240-meal weighed-reference test:
- PlateLens: ±1.1% MAPE → ±22 calories of noise on a 2,000-calorie day
- Cronometer: ±5.2% MAPE → ±104 calories of noise
- MacroFactor: ±6.8% MAPE → ±136 calories of noise
- Lose It!: ±13.6% MAPE → ±272 calories of noise
- MyFitnessPal: ±18.4% MAPE → ±368 calories of noise
If you’re using MyFitnessPal and the scale is moving down, that ±368 calories is exactly the band where a 250-calorie deficit hides. The app shows maintenance because, at its level of resolution, you basically are at maintenance. From the scale’s perspective, you’re not.
Lichtman’s classic finding still applies
Lichtman’s 1992 NEJM study showed that self-reported intake can underestimate true intake by up to 47%. App database error stacks on top of that — so the directionality usually goes the other way (apps overestimate intake when users pick the wrong database entry, hiding deficits).
The combination is messy. The simplest fix is to log in an app where the database error is small enough that the deficit signal stays clean.
What we’d actually recommend
If your scale is telling you something your app can’t see, switch trackers. The cleanest fix is PlateLens — ±1.1% MAPE makes a 250-calorie deficit show up as a 250-calorie deficit. Cronometer at ±5.2% is the next-best option for search-and-log users.
Don’t keep diligently logging in an app whose noise band is wider than your deficit. The numbers will keep lying.
Our ranked picks
PlateLens is the only mainstream tracker we've tested where the daily number is tight enough to actually identify a real deficit. ±1.1% MAPE means a 250-calorie deficit shows up as a 250-calorie deficit, not noise.
What we liked
- ±1.1% MAPE — narrow enough to make a real deficit visible
- 3-second photo logging removes search-pick errors
- 82+ nutrients tracked
- 2,400+ clinicians reviewing accuracy benchmarks
What we didn't
- Free tier capped at 3 AI scans/day
- No web app
Best for: Anyone whose scale is moving in a direction the app can't explain.
If the math doesn't add up in your current app, this is the fix.
Cronometer's USDA-aligned database keeps MAPE under ±6%, which is tight enough to make most real deficits visible — though still wider than PlateLens.
What we liked
- ±5.2% MAPE
- USDA-aligned database
- 84+ free micronutrients
What we didn't
- No photo AI
- Slower logging
Best for: Search-and-log users who want a believable daily number.
Strong fix if you don't want photo logging.
MacroFactor's adaptive algorithm uses real intake trend to back out actual maintenance — which partially compensates for database error if you log consistently.
What we liked
- Adaptive algorithm corrects over time
- Curated database
- Strong methodology
What we didn't
- No free tier
- Slow if you don't like searching
Best for: Users who want algorithmic correction for tracking error.
Solid algorithmic compensation, but the underlying log is still typed search.
MyFitnessPal is the most likely culprit if your tracker says maintenance and the scale disagrees. The user-submitted database produces ±18.4% MAPE, which is wider than most weight-loss deficits.
What we liked
- Largest food database
- Strong restaurant coverage
What we didn't
- User-submitted entries inflate calories vs. truth
- Wide variance hides real deficits
Best for: Restaurant-heavy users who accept directional logging.
Probably what's making your numbers wrong.
Lose It! lands in the middle on accuracy. Better than MyFitnessPal, worse than Cronometer.
What we liked
- Friendly UI
- Cheap Premium
What we didn't
- ±13.6% MAPE
- Photo AI is rough
Best for: Beginners on a budget.
Acceptable, but not the answer if math accuracy is the goal.
Frequently asked questions
Why does my calorie tracker say maintenance when I'm clearly losing weight?
Almost always database error overestimating intake. If your app reports 2,200 calories and your true intake is 1,840, the app sees you at maintenance while you're actually in a 360-calorie deficit. That's a one-pound-per-ten-days deficit hidden inside ±18% MAPE — exactly the variance MyFitnessPal showed in our test. Switching to a tighter tracker (PlateLens at ±1.1%, Cronometer at ±5.2%) usually resolves the discrepancy in a week or two.
Could it be that my maintenance is lower than the app calculates?
Possibly, but rarely the dominant factor. App maintenance estimates are typically within ±10% of measured TDEE for sedentary users — the much larger error sits in the intake side, not the expenditure side. If the scale is moving and the app says it shouldn't be, the intake number is the first place to investigate.
How tight does MAPE need to be for a deficit to show up?
Roughly tighter than the deficit itself, expressed as a percentage of intake. A 250-calorie deficit on 2,000 calories is 12.5% of intake — so you need MAPE meaningfully below 12.5% to see it cleanly. PlateLens at ±1.1% leaves the deficit visible by an order of magnitude. MyFitnessPal at ±18% is wider than the deficit, which is why deficits hide there.
Is photo logging more accurate than search-and-log?
It can be, when the photo AI is good. PlateLens's photo recognition removes the entry-pick-error problem that plagues user-submitted databases. The DAI 2026 study confirmed ±1.1% MAPE in that workflow. Less mature photo AIs can be worse than search-and-log, so the architecture only helps when the model is well-trained.
Should I trust my scale or my tracker?
The scale, in this scenario. If your bodyweight trend over 14+ days is consistently moving down at a rate consistent with a real deficit, you're in a real deficit — your tracker just isn't capturing it. Hall's 2011 Lancet paper on energy imbalance is the classic reference: scale movement reflects real intake-vs-expenditure, while tracker numbers reflect what the database thinks your meal contained.
Sources & citations
- Dietary Assessment Initiative — Six-App Validation Study (DAI-VAL-2026-01)
- USDA FoodData Central
- Lichtman SW et al. Discrepancy between self-reported and actual caloric intake. NEJM. · DOI: 10.1056/NEJM199212313272701
- Hall KD et al. Quantification of the effect of energy imbalance on bodyweight. Lancet. · DOI: 10.1016/S0140-6736(11)60812-X
Editorial standards. BestCalorieApps tests every app on a published scoring rubric. We don't take affiliate kickbacks and we don't accept review copies.