People do not stop tracking food because they lack willpower. They stop because logging a meal takes ninety seconds, they eat four times a day, and six daily minutes of fiddly search-and-select is a tax that any sufficiently busy Tuesday will eventually refuse to pay. The habit doesn't break dramatically. It just doesn't happen one day, and then it never happens again. Habits die the way most things do: quietly, midweek.
Every second you shave off the logging flow is a real increase in the odds you're still doing this in March. So it is worth four minutes of your attention to learn which input method belongs to which food, which is what this is.
Barcode: exact, instant, and entirely dependent on someone having printed one
Point the camera, hear the beep, adjust the portion. Two to five seconds, start to finish.
Use it for: anything with a label. Yoghurt, protein bars, bread, the ready meal you're pretending was a choice, milk, packaged chicken, cereal, the sauce.
Why it wins: the numbers come from the manufacturer. No estimation, no theology about whether "chicken breast" means raw or cooked, skin on or off. For packaged food a barcode is ground truth, and nothing else in this article gets to say that.
Where it dies: the moment there's no barcode. Restaurants, anything homemade, loose produce. And the special small heartbreak of the barcode that scans perfectly and returns nothing, because that product never made it into the database.
The part everyone gets wrong anyway: the serving size. The scan gives you macros per 100 g or per serving; it does not know you ate two-thirds of the packet standing over the sink. This, not identification, is where barcode logging quietly fails. If you're going to weigh anything at all, weigh the packaged food you eat by the handful.
Photo: fast for real plates, approximate by nature, and honest about it
Photograph the meal. Get an estimate of what's on the plate and what's in it. Fix whatever looks wrong. Log it.
Use it for: homemade meals, restaurant food, anything mixed; anything where the alternative is logging six ingredients by hand like a medieval clerk.
Why it wins: the alternative for chicken, rice and broccoli is three searches, three portion guesses, three taps. A photo collapses that into one act. For composite meals it isn't somewhat faster. It's an order of magnitude faster, and orders of magnitude are the only speedups that change behaviour.
Where it fails: it's an estimate, and unlike some estimates it admits it. It cannot see the oil the chicken went into the pan with. It cannot tell full-fat yoghurt from fat-free by looking (nobody can, that's rather the point of yoghurt). It will guess the portion from apparent volume. Expect to correct it.
The right way to hold it: photo logging is not "record this meal exactly." It is "give me a first draft I can fix in five seconds." Which is enormously better than the realistic alternative, because the realistic alternative to an approximate log was never a perfect one. It was an empty one.
Search: still right for exactly one thing
Neither method deserves the food you eat constantly. If the same breakfast happens four days a week, log it properly once, save it as a meal, and add it forever after in one tap. No scanning, no photography, no ceremony.
For most people this covers more of their intake than they'd guess. Breakfasts and post-workout meals are ruts, in the good sense. Handle them with saved meals and you've deleted the majority of your logging time before touching a camera.
The actual rule
| Situation | Reach for |
|---|---|
| Has a barcode | Scan it |
| You eat it most weeks | Saved meal or recipe, logged once |
| Homemade or mixed plate | Photo, then correct |
| Restaurant, no packaging | Photo, then correct |
| Loose produce, weighed | Search and weigh |
On accuracy, and how much of it you actually need
Here is the sentence that would save people a great deal of anxiety if anyone ever believed it the first time: precision matters far less than consistency.
If your photo estimate runs 10% off, but 10% off in roughly the same direction every day, your trend is still true, and the trend is the only thing you make decisions from. Four days of laboratory-grade logging followed by nothing is worth less than twelve weeks of cheerful approximation. Considerably less.
Which is the same argument as reading the twelve-week trend rather than one day's number. What you are measuring is a direction. It was always a direction.
Doing it in OneRep
Barcode scanning, search backed by USDA FoodData Central, photo-assisted logging, recipes and saved meals all sit in the free tier; putting a paywall around the fast input methods would rather defeat the point of having built them.
You can also hand the photo to Coach and have the meal written straight into today's diary, along with anything else you'd like done to your log. Every write is shown for approval first, undoable in one tap. That runs on the monthly AI allowance every account gets, so you can find out whether the Coach subscription is for you before anyone asks you for money. Novel, I know.
Try it: Free, no trial window.
Compare: OneRep vs MyFitnessPal · OneRep vs MacroFactor