snapkin
The research behind Snapkin
Snapkin analyses what you have eaten and puts a number on it. You should be able to check where that number comes from, so this is where we write down our reasons.
Written and maintained by Mattias Geniar. Last updated .
12 decisions, 4 questions, 87 sources. Every one of them linked.
None of this is medical advice. Snapkin is a general wellness tool: it does not diagnose, treat, prevent or monitor any disease, and it is not a medical device. The terms say what that means in practice.
Question 1 of 4
The camera
What a picture of a plate can tell you, and what it cannot.
Why we think a photograph is good enough
External photo benchmarks, the difference between average and individual error, and why portions and preparation matter.
23–34% mean absolute energy error for four commercial models in one benchmark; not Snapkin validation
Why we track added sugar, and what we mean by it
Why free sugars differ from the US added-sugars label, and why specialist validation cannot establish photo accuracy.
0.98 ICC for specialist estimates of packaged-food added sugar; not Snapkin’s accuracy
Question 2 of 4
The targets
What the app does with the number once it has one.
How we calculate your calorie target
The arithmetic for loss, gain and maintenance, the evidence behind it, and why an energy surplus cannot promise kilograms of muscle.
3 objectives a deficit, a surplus, or estimated maintenance; different calorie and protein targets
Why protein comes first
What eating more protein does to appetite, what it protects while you are losing, and where the effect stops.
−441 kcal/d mean intake change in 19 people given food at 30% protein; no parallel control group
Why your steps and workouts do not change your calorie target
Most other trackers add your workout to what you may eat. The review that changed our mind, and the fair complaint against that choice.
8 studies 270 participants in the review’s energy section; errors varied widely
Why the meal it preselects depends on where you are
Meal times come from national time use diaries rather than one country’s habits, so lunch lands when your country eats. And what that still cannot know about you.
23 countries of real time use diaries decide when the app calls it lunch
Question 3 of 4
The habit
Whether logging food is worth doing at all, and who it is bad for.
Why track anything at all
What logging is worth in kilograms, why the app asks for a photo instead of a database search, and the honest limits of the evidence.
−2.87 kg pooled difference in diet-and-activity self-monitoring interventions; not logging alone
Tracking and eating disorders
The page that argues against our own product: what the link between tracking and eating disorders really is, the limits of the experimental evidence, and what the app refuses to do.
r = 0.13 pooled association for weight-related self-monitoring; 95% CI −0.02 to 0.28, not proof of safety
Rewards, and what we will not give one for
Nothing here can be earned faster by eating less. The studies behind that rule, and what the app counts instead.
> 3 days consistent logging was associated with less regain in one study; not a proven minimum
How much setup is worth asking for
Which setup questions change the plan, what can wait, and the evidence for testing a shorter introduction.
70 people in a tutorial usability study; subscription conversion remains a separate question
Question 4 of 4
The scale
How fast weight should move, and how to read a scale that jumps around.
How fast the app will let you lose weight
Why the pace stops at 0.8% of body weight a week: a small trial with greater lean-mass gains in the slower group, and why the default sits well under the cap.
+2.1% lean-mass gain in athletes assigned 0.7% weekly loss; the app’s 0.8% cap was not tested
Why the weight chart draws a line you did not measure
The heavy line is not your weighings. What one morning on a scale really tells you, and why smoothing cannot remove every source of variation.
0.4 kg illustrative variation at 82 kg; smoothing reduces scatter but cannot measure fat loss
How these pages are maintained
These pages are in English only, on purpose. Machine-translating a medical claim into ten languages is ten chances to get one wrong.
Who writes them, and what the company behind them is registered as, is on the about page. Every decision on these pages has a feature page describing what it does in the app: Photo logging, Protein first, Home screen widget, It learns you, Honest numbers, Weight tracking.
Every figure here was read from the paper or its abstract, not remembered. Where we could only reach an abstract, the page says what the abstract says and nothing more. Where a claim rests on a preprint or on one small study, it is labelled. If you find something wrong, tell us at [email protected] and we will correct it here.