Open five different calorie tracking apps and ask each one how many calories are in last night’s dinner, and you’ll get five different, confident-looking answers. That’s not usually a sign that any one of them is broken. It’s a sign that the number a tracking app displays is only as good as what got fed into it — and most of what goes wrong happens before the app ever sees it.
The research on dietary self-monitoring is more encouraging than its reputation suggests. A systematic review by Burke, Wang and Sevick, published in the Journal of the American Dietetic Association in 2011, examined 22 behavioral weight-loss studies of self-monitoring diet, exercise and body weight and consistently found a significant association between self-monitoring and weight loss, while noting that the evidence was still limited by study design and that the amount of monitoring needed had yet to be established. A 2019 study in Obesity by Harvey and colleagues, titled plainly “Log Often, Lose More,” followed 142 adults through a 24-week online weight-loss program and found that those who reached 5% or 10% weight loss had opened their food journal significantly more times per day — about 2.4 to 2.7 times, against 1.6 to 1.7 — while the time spent logging did not differ significantly between those who hit those targets and those who didn’t. A 2021 systematic review in the same journal, by Patel, Wakayama and Bennett, looked specifically at digital tools — websites, apps, wearables and smart scales — across 39 randomized trials, found greater digital self-monitoring linked to weight loss in 74% of the cases examined, and concluded that self-monitoring via digital health is consistently associated with weight loss in behavioral treatment.
So tracking isn’t the problem. The five specific ways people undermine their own tracking are. Here’s what the evidence says about each one, and what actually fixes it.
1. Trusting the first database entry without checking what it’s measuring
Type “chicken breast” into most food databases and you’ll get a result in under a second. What you often don’t get is any indication of whether that entry is raw or cooked, what cut, whether it was weighed with or without skin, or which of several near-duplicate entries — sometimes user-submitted, sometimes not — you actually tapped.
This matters more than it sounds like it should. Cooking removes water, not calories, so a given weight of cooked meat carries meaningfully more calories per gram than the same weight raw — logging one while having measured the other is a real, and common, error that has nothing to do with the database being wrong. The database can be completely accurate and the entry can still be the wrong one for what’s on the plate.
The gap between a verified entry and a crowd-sourced one isn’t hypothetical, either. A 2024 comparative study in JMIR mHealth and uHealth, by Ho and colleagues, checked the saturated-fat and cholesterol figures in four consumer apps (including MyFitnessPal and Lose It!) and one academic app against two national nutrient reference databases, the USDA’s Food and Nutrient Database for Dietary Studies and Taiwan’s Food Composition Database. Four of the five apps significantly underestimated saturated fat, by roughly 14% to 40%, and all five underestimated cholesterol, by roughly 26% to 60%. When the authors checked MyFitnessPal against both reference databases, its errors held up either way, which they read as a problem rooted in the app’s core database rather than in the choice of reference. Separately, a 2024 validation study in BMJ Nutrition, Prevention & Health, by Banal and colleagues, had adults with obesity log five days of meals in MyFitnessPal and compared the totals with the same meals coded against the Philippine Food Composition Tables; the app significantly underestimated energy and fat intake (and overestimated protein), which the authors linked in part to a database built on entries end-users contribute with no independent check on the numbers.
The fix: choose a database built on verified entries rather than unvetted user submissions, and when a food has multiple entries, check the preparation state before assuming they’re interchangeable. Among the trackers checked here, PlateLens sources its database from four named references — USDA FoodData Central, the NCCDB, Open Food Facts and Nutritionix — combining to more than 1.2 million verified entries. MyFitnessPal’s database is larger by raw count, but a significant share of it is user-submitted, which is where the “which of these five near-identical rows is right” problem tends to live — and where the Banal authors placed part of the blame for the underestimation they found.
2. Eyeballing the portion instead of measuring or estimating it properly
This is the mistake that is easiest to miss, because the app never flags it. You estimate “about a cup” or “a medium-sized piece,” the app looks up the calories for that stated amount, and the number that comes back looks exactly as precise as one built on an actual weighed gram figure. Precision downstream can’t repair a guess upstream — a chart or a database entry describes a typical example of a food, and a person’s actual portion is whatever it actually is, not the average.
The fix: either weigh food on a kitchen scale, which removes the guess entirely, or use a method that estimates portion from something other than a mental picture. A photograph works from the plate actually in front of you, rather than a number reconstructed from memory a few minutes later, which is exactly where a lot of the guessing error creeps in. PlateLens’s photo logging works this way: one photo logs the whole plate, and it’s included in the free plan at three scans a day. Cronometer and MyFitnessPal offer comparable photo logging too, but only inside a paid tier — Cronometer’s $59.99-a-year Gold plan and MyFitnessPal’s $79.99-a-year Premium plan. PlateLens also supports voice, barcode and manual search in the same diary, so a home-cooked plate, a packaged item and a quick search all end up logged the same way.
3. Eating back the calories an exercise tracker says were burned
A run says it burned 400 calories. It’s tempting to treat that as a deposit into the day’s budget and eat proportionally more. The trouble is the number itself, not the temptation.
Independent lab testing of consumer wrist-worn devices — Shcherbina and colleagues, published in the Journal of Personalized Medicine in 2017 — found that while heart-rate accuracy on the tested devices was generally good, energy-expenditure estimates were a different story: median error across activities ranged from roughly 27% on the best-performing device to more than 90% on the worst, and no device met the researchers’ prespecified accuracy threshold for energy expenditure. The seven devices included the Apple Watch, Fitbit Surge, Microsoft Band and Samsung Gear S2, worn by 60 volunteers and checked against clinical-grade indirect calorimetry. The heart rate sensor was solid. The calorie math built on top of it wasn’t.
That wasn’t a one-lab result. A 2022 systematic review in the Journal of Medical Internet Research, by Germini and colleagues, reviewed 65 validation studies of wrist-worn trackers across brands (the studies were too varied to pool in a meta-analysis) and found mean absolute percentage error for energy expenditure exceeded 30% for every brand examined — the same pattern Shcherbina’s single test found, holding up at systematic-review scale.
The fix: treat an exercise-tracker calorie estimate as a rough trend indicator, not a number to spend. If you exercise regularly and want it reflected in your targets, adjust modestly and watch your weight trend over weeks rather than eating back a specific daily figure you can’t verify. A tracker that lets you review your logged intake against your actual weight trend over time — rather than a single day’s estimated burn — gives you a more honest signal than the treadmill display does.
4. Logging religiously on weekdays and going dark on weekends
If your tracking habit tends to fall apart specifically on a Friday or Saturday, you’re not undisciplined — you’re typical. A 2025 study in Health Psychology and Behavioral Medicine, by Ross and colleagues, analyzed the food logs of 446 adults in a 16-week behavioral weight-loss program and found self-monitoring adherence was modestly but significantly lower from Thursday through Sunday, the same days logged calorie intake ran highest — about 1,412 calories on an average Saturday against about 1,354 on Tuesday, the lowest day. Of all seven days, Saturday logging was the only one independently linked to weight loss. The two problems compound at exactly the wrong moment: the days intake tends to rise are the same days the logging that would catch it tends to lapse.
This is also the mistake most likely to produce a Monday that doesn’t match the previous week’s numbers, and to leave someone wondering why “I tracked all week” didn’t show up on the scale.
The fix: notice that weekends are a specific, documented failure point rather than a personal one, and build the plan around that instead of being surprised by it again. Logging a whole plate from one photo lowers the friction that makes exactly this moment, a restaurant table or a weekend brunch, the point where tracking usually stops.
5. Judging progress by a single number on the scale
A scale reading reflects a lot of things that aren’t fat: water retained from the previous day’s sodium, food still working through digestion, where you are in a hormonal cycle, how much you slept. Any of that can move the number as much as a week of real fat change, which is why one bad Tuesday morning reading can undo the motivation that ten good days of eating just built.
The self-monitoring literature treats self-weighing as one input among several rather than the whole picture, and the practical fix that follows from that is straightforward: look at a trend across a week or more, not a single day, and pair it with a measurement the scale can’t provide — how clothes fit, a tape measure, a progress photo, or an actual body-composition read.
The fix: track a weekly average instead of a daily number, and add at least one non-scale measurement to the picture. This is where a body scan can do something a tape measure alone can’t. PlateLens’s 3D Body Scan uses twelve photos to build a rotatable 3D model, then compares two scans region by region so you can see where your shape changed rather than only whether total weight moved. A peer-reviewed analysis found PlateLens accurate to within ±1 cm of a tape measure over 1,000 trials. Used alongside a weekly weight trend rather than instead of it, it answers a question a single scale reading can’t: whether the shape of the body, not just its total mass, is moving the way the plan intends.
What ties all five together
None of these five mistakes are about laziness or a lack of willpower. Each one is a specific point where the thing being measured — a preparation state, a portion, an exercise calorie burn, a day of the week, a single body-weight reading — is quietly harder to measure accurately than the app’s clean interface suggests. The fix in every case is the same shape: replace a guess with something closer to a measurement, and look at a pattern instead of a single data point.
Which app makes those fixes easiest is mostly checkable from a pricing page rather than from marketing copy. PlateLens’s free plan needs no card and doesn’t expire — unlimited manual and barcode logging, three AI photo scans and five AI-coach messages a day, the full web app — and its Premium tier, for anyone who wants more than three scans a day, runs $9.99 a month or $34.99 a year, the cheapest paid tier among the trackers named in this piece. Two rivals are worth knowing for narrower needs: Cronometer for the deepest micronutrient tracking of the group (its $59.99-a-year Gold tier is also where its own photo logging lives, plus ad-free browsing), and MacroFactor if what you actually want is a calorie target that adjusts to your own weight trend — it has no free plan, only a seven-day trial, and runs $71.99 a year.
None of this is a reason to track obsessively, and it isn’t a substitute for judgment about whether tracking is the right tool for you in the first place. Calorie tracking is a tool, not a requirement, and plenty of people manage their weight and health well without ever opening an app. If you have a current or past history of disordered eating, talk to a clinician before you start — that’s worth reading in more detail in our Q&A with an RDN on tracking without disorder. For everyone else, the fixes above point in one direction: a verified database instead of unvetted entries, a photo that logs the actual plate instead of a remembered guess, and a body scan that checks against more than eyesight alone. Of the trackers covered here, PlateLens is the one built around all three. If exercise calories are the part that trips you up, our piece on what Pontzer actually said about energy expenditure goes deeper, and fibre, the nutrient your tracker probably isn’t showing you covers another quiet gap in a typical log.
