Does tracking work, and how consistent is consistent?

What was measured

Oreshko and Heikkinen analyzed six months of records from 5,132 people using a commercial food-tracking app and published the result in the Journal of Medical Internet Research in 2026. Everyone in the group had a weight-loss goal and a weight recorded every month from June to November 2025. The question was simple: does how often somebody logs predict how much weight they lose?

It does, and the relationship climbs steadily across the whole range.

Under 2 days a week. 6.84% of body weight lost on average, with 52.0% of that group losing at least 5%.
2 to 3 days, 7.35%, and 56.5% reached 5%.
3 to 4 days, 7.77%, and 59.8%.
4 to 5 days, 7.95%, and 63.9%.
5 to 6 days, 9.82%, and 73.9%.
6 to 7 days, 11.33%, and 80.0% reached 5%.

Adjusted for age, sex and starting body mass index, each additional day of tracking per week was associated with 1.35 times the odds of reaching a 5% loss. Across the whole group mean loss was 9.3 kg, and 70.1% reached 5%.

How much weight that finding carries

A number this quotable deserves its design stated plainly, because the design is what decides how far it travels.

It is an association, not a demonstration that tracking caused the loss. This was a retrospective look at records that already existed. Nobody was assigned to track more or less and there was no control group, so people who log daily may simply be people who are more committed to everything else as well. The authors say as much: motivational differences cannot be excluded.

The group is a narrow slice. The 5,132 are 1.2% of the 412,193 people who signed up in the period, after filtering for a weight-loss goal, at least 21 days of logging, and a weight recorded every single month for six months. They were older and heavier than the people excluded. This describes the most engaged fraction of an already selective group rather than the average person who downloads a tracker.

The study was run by the app’s own developer. Both authors work at MyNetDiary, the company whose app supplied the data, and the first author is its chief executive. That is declared in the paper. It is peer-reviewed and indexed, which is not nothing, but it is not an independent trial and anyone citing it should say so.

Some of the loss may not be the tracking. The authors note that medication was not measured, and name GLP-1 receptor agonists specifically: across the period studied, some of that weight loss may reflect drugs rather than logging.

With all of that attached, the finding is still worth having. Among people who are trying, more consistent logging goes with better results, and the distance between occasional and near-daily is not marginal.

Why frequency would matter more than precision

The mechanism is not mysterious, and two well-established findings point straight at it.

The first is that estimating intake is hard and errs in one direction. Freedman and colleagues pooled five large validation studies against recovery biomarkers in the American Journal of Epidemiology (2014) and found people under-report energy intake by about 28% using a food frequency questionnaire and about 15% on a single 24-hour recall, with body mass index and age both predicting how much. In the extreme case, Lichtman and colleagues studied ten people who believed themselves resistant to dieting and reported in the New England Journal of Medicine (1992) that they under-reported intake by 47% while their measured metabolic rates were normal. That was a small and deliberately selected group, but it demonstrates the thing that matters here: the gap is real and it is invisible from the inside.

The second is that this drift is much of what a plateau is made of. Hall and colleagues concluded from the shape of real weight-loss curves in The Lancet (2011) that the stall people hit at six to eight months implies the original reduction in intake was “rapidly and progressively relaxed” rather than metabolism collapsing.

Put those together and a consistent log is doing something specific. It is not producing an exact number, it is producing a comparable one: your own error stays roughly constant if your method does, so the direction and rate of change in your own data remain readable even when the absolute total is off. Skipped days are what break that, because the days people skip are not a random sample of their days.

What follows for anyone logging

Frequency first, precision second. A rough entry on a bad day is worth more to your own trend than a perfect entry on four days out of seven.

Log the awkward days especially. They carry the information, and they are the ones that get skipped.

Keep the friction low enough that consistency is realistic. This is the one part a piece of software can genuinely help with, and it is the whole reason logging speed is worth caring about.

None of that tells you what to eat or what to weigh, which are questions for you and, where health is involved, a clinician. It is about whether the record you keep is good enough to read. See why weight loss stalls for what to do when the trend goes flat, and what a calorie deficit is for what the numbers in it mean.

How Loopa handles this

Loopa is built around exactly that finding, which is why every way of logging is free and none of them is rationed: describe a meal in plain words, photograph the plate, search the database, scan a barcode, repeat a saved meal, or say it out loud. Food, weight and water logging is free forever (no ads, no daily cap, no card) because a tracker that rations the thing it exists to do is a tracker people stop opening. If you have been meaning to start, get Loopa free and log your first meal tonight.