The habit · Decision 7 of 12

Why track anything at all

Logging is usually studied inside a wider programme. The results below do not isolate the effect of Snapkin, photo logging, or logging alone.

Written and maintained by Mattias Geniar. Last updated .

What the app does
Asks for a photograph rather than a database search, because the biggest problem in this category is how few people are still logging after a month.
How sure we are Fair , Real, but small, indirect, or thinly measured.
Regular logging is associated with greater weight loss, and some digital programmes help. Installing an app alone does not guarantee a benefit. Motivation and accompanying support matter.

The evidence for it

Three results about the same behaviour: a pooled estimate from randomised trials, a comparison of behaviour-change techniques, and a study of how often people log.

−2.87 kg compared with a control group, pooled across 12 randomized trials of diet-and-activity self-monitoring interventions Berry et al. (2021) · 95% CI −3.78 to −1.96 · intake down 182 kcal/d
13% of between-study heterogeneity explained by self-monitoring in a meta-regression; not 13% extra weight loss Michie et al. (2009) · 122 studies, 44,747 people
2.7 × a day how often people who lost 10% or more of their weight logged, compared with 1.7 times a day for those who lost less Harvey et al. (2019) · 142 participants; month-six frequency association, not causation

What exists instead compares ways of logging, or compares whole programmes that happen to include logging. The best of it: Berry et al. (2021) pooled 12 randomised trials of digital self-monitoring and found −2.87 kg (95% CI −3.78, −1.96), with calorie intake down 182 kcal a day. Michie et al. (2009) looked across 122 studies and 44,747 people. Of the 26 behaviour-change techniques they examined, self-monitoring explained more of the difference between studies than any other. Harvey et al. (2019) found that people who lost 10% or more logged 2.7 times a day, against 1.7 times for those who lost less than 10%. Among those still logging at month six, time spent did not differ significantly by weight-loss group. This does not mean logging time is irrelevant or prove that increasing frequency causes weight loss.

Different interventions produced different weight-loss estimates Weight difference against a control group, with 95% confidence intervals, on one axis.
no difference
Diet-and-activity interventions, 12 trials −2.87 kg

Berry et al. 2021. Calorie intake down 182 kcal a day alongside it

Being handed MyFitnessPal, 212 patients −0.30 kg

Laing et al. 2014, primary care, six months, p = 0.63. The interval crosses zero

−4.2 kg +1.4 kg
Berry pooled diet-and-activity interventions, often with tailored advice. Laing tested introducing MyFitnessPal in primary care and did not detect a significant weight benefit. These findings do not isolate food logging or predict the effect of Snapkin.

And the usual warning from the same research: Burke et al. (2011) found the link every time, and then said plainly that “the level of evidence was weak because of methodologic limitations.”

Prediabetes is where the evidence is strongest, and it is not about the app

The Diabetes Prevention Program (3,234 people, followed for 2.8 years on average) is the landmark study. An intensive lifestyle programme, aiming for 7% weight loss and 150 minutes of activity a week, reduced the incidence rate of type 2 diabetes by 58% relative to placebo (95% CI 48–66) (4.8 versus 11.0 cases per 100 person-years). Metformin, a drug, cut them by 31% (17–43). Average weight loss was 5.6 kg. The Finnish Diabetes Prevention Study (522 people) found the same 58% reduction on its own.

Writing down what you eat was part of both. In the American programme, Wing et al. (2004) reported that “dietary self-monitoring was positively related to meeting both weight loss and activity goals.” In the Finnish study, dietary advice was based on three-day food records.

On how much weight loss is enough: the usual “5 to 7%” figure is a programme target taken from what those two trials aimed for and reached. It did not come from a study designed to find the right amount. The real relationship is gradual, with no cut-off point. Hamman et al. (2006) found an association with 16% lower diabetes risk per kilogram lost (HR 0.42, 95% CI 0.35–0.51 per 5 kg), and weight loss was the strongest predictor. This is an observational association within the trial, not a guaranteed risk reduction from each kilogram. Magkos et al. (2016) showed that 5% weight loss on its own improved insulin sensitivity in fat, liver and muscle.

Diabetes incidence fifteen years after the original DPP assignment Share of each group that had developed type 2 diabetes at the 15-year follow-up.
Intensive lifestyle programme 55%

The group that lost 5.6 kg and reached its activity target

Metformin 56%

Original metformin assignment

Placebo 62%

Original placebo assignment; later offered lifestyle training

0 70% of the group
At 2.8 years the lifestyle group’s risk reduction was 58%. At the 15-year follow-up, more than half of every group has developed diabetes, and the gap is seven percentage points. The long-term follow-up includes changes in treatment: all groups were offered lifestyle training after the randomized phase. These are cumulative outcomes by original assignment, not fifteen years of an unchanged placebo comparison. They demonstrate substantial remaining risk, not proof that everyone will eventually develop diabetes.

The evidence against, and the two numbers we like least

Almost nobody keeps logging

Helander et al. (2014) followed 189,770 people who downloaded and opened a free photo food-logging app at least once.

Active use in one free photo-logging app The Eatery: 189,770 people who opened the app at least once in 2011–2012.

2.58% used it actively did not

Active meant at least ten photographs over at least one week: 4,895 people (2.58%). This is not a standard 30-day retention rate. The app already used photographs, so the result cannot establish that photos solve abandonment or that tapping effort caused it.

Apps in general do very little

Chew et al. (2022) put the pooled effect of apps at 2.18 kg at three months, falling to 1.63 kg at twelve, and called the effects “minimal in their current states.” A separate review studied apps combined with human coaching (Chew et al., 2023). It found −2.15 kg against varied controls and explicitly called for research on added benefit over an app alone. Comparing pooled estimates across reviews cannot establish that coaching doubles the effect. Snapkin has no human coach.

Self-reported calories barely match real calories

Freedman et al. (2014) pooled five studies with 2,265 people in total. Each compared what people reported eating with doubly labelled water, the most accurate way to measure how much energy a person really uses. Under-reporting gets worse with body weight, by a further 5 to 7 percentage points at a BMI of 30 compared with 25. The correlation between reported and true calorie intake was r = 0.21 to 0.31.

How much of what people eat never makes it into the log Under-reported energy against doubly labelled water, the reference method.
A single 24-hour recall 15%

Freedman et al. 2014, pooled over five validation studies, 2,265 people

A food frequency questionnaire 28%

The same pooling. The more the method asks you to remember, the worse it gets

0 35% under-reported
Photo-based methods are not immune: Ho et al. (2020) found image-based logging came up 448 kcal a day short against doubly labelled water. That pooled shortfall does not identify how much came from omitted food versus inaccurate estimates; both can contribute.