Methodology
This page is the full data method behind the "Emotional eating, measured" series: what is recorded at each meal, which meals are excluded, how the headline figures are defined, the statistical tests used, the limits of what the data can show, and how the data is handled. Every per-article "How this was measured" note links here.
What is collected at each meal
Meals in the I feel food app are logged in two halves. Before eating, you move a slider to show how hungry you are, on a 0–9 scale, and pick what you are feeling from a list of 45 emotions. After eating, you do it again: how full you are now, and how you feel.
Hunger and fullness sit on the same 0–9 line: low means empty, high means full. The table below is the full scale.
| Reading | What it feels like |
|---|---|
| 0.0–0.4 | Famished: dizziness, fatigue |
| 0.5–1.4 | Strong hunger |
| 1.5–2.4 | Moderate hunger |
| 2.5–3.4 | Weak hunger: slight discomfort in the stomach |
| 3.5–4.4 | Neither hungry nor full |
| 4.5–5.4 | First feeling of fullness |
| 5.5–7.4 | Light to pleasant fullness |
| 7.5–8.4 | Strong fullness, discomfort |
| 8.5–9.0 | Overeating, feeling unwell |
The line between physical and emotional hunger sits at 3.5. Below it, the body is asking for food; at or above it, whatever is sending someone to the kitchen, it isn't the stomach. → Physical hunger vs emotional hunger
The dataset: date range and totals
- 1,534,951 meals logged by 50,692 people, 2021 to September 2026.
- 1,132,266 of those meals have a valid hunger reading before eating and a valid fullness reading after — the set used for every hunger/fullness comparison.
- 474,216 meals have at least one feeling named before eating.
- 403,738 meals have a feeling named on both sides, before and after — the set used for before/after feeling comparisons.
- 1,259 people are followed over time in the three-month comparison (a looser replication follows 2,155). → Does a food and mood journal work?
The data is observational, with no control group, and is not available for download; it is published only as the aggregates in this article series.
Exclusion rules: what counts as a valid meal
- A reading of exactly 0 means "not recorded", not "starving" or "stuffed" — the app's own logic — so it is excluded from any comparison that needs a valid reading on that side.
- A comparison between hunger and fullness needs both readings on the same meal: that is the 1,132,266-meal set above.
- A comparison by feeling needs a feeling named before eating (474,216 meals); a before/after feeling comparison needs one named on both sides (403,738 meals).
- The three-month comparison additionally requires at least eight valid meals in a person's first 14 days of logging and at least eight more in days 56–83 (1,259 people), or, in the looser replication, at least four meals in each of those windows with the later window widened to weeks 9–16 (2,155 people).
How the headline figures are defined
- Eating without physical hunger — a before-meal hunger reading of 3.5 or higher: "neither hungry nor full" or fuller.
- Ended uncomfortably full — an after-meal fullness reading of 7.5 or higher.
- Overeating zone — an after-meal fullness reading of 8.5 or higher.
- Still felt it afterwards ("feeling persistence") — the same feeling named again after eating, among the 403,738 meals with a feeling on both sides.
For example, of the 403,738 meals with a feeling on both sides, 33,124 began with an uncomfortable feeling and no physical hunger — eating without physical hunger, as defined above. → Does emotional eating help?
The statistical tests used
Only the three-month comparison reports a significance test. It is a paired t-test, each of 1,259 people compared against themselves: t = −8.87 for eating without physical hunger, t = −7.03 for overeating, t = −1.29 (not significant) for distress before eating. A looser replication on 2,155 people with a wider later window gives t = −10.73 and t = −7.88 for the first two. → Does a food and mood journal work?
Every other figure in the series — the percentages, ratios and rankings elsewhere on this page and in the articles — is a direct count or share over the meals defined above; the series does not separately publish a significance test for those.
Known limits
- No control group. The whole series is observational. Some of any measured change could happen through attention alone, and the data can't say how much.
- Self-selected, engaged users. The longitudinal comparison describes the 1,259 (or 2,155) people who kept logging for two months or more, drawn from the 50,692 who logged at least once. It describes engaged users, not everyone who installs the app.
- Measurement drift. People may learn to read the hunger/fullness slider differently over time, which would look like behaviour change even if nothing else did.
- Correlation, not causation. "People who tracked changed X" is what the data supports. "Tracking changed X" is a stronger claim than this data can make.
Privacy
Every figure on this page and across the series is a count or a percentage drawn from many people's logged meals. No individual's diary, meal photo or note is published.
I feel food's Privacy Policy and Consumer Health Data Privacy Policy describe how journal data is handled; analytics derived from it are anonymised and aggregated into counts and categories that cannot be linked back to any one person.
Cite this data
I feel food tracker · 1,534,951 meals, 2021 to September 2026
- Before eating: a hunger reading on a 0–9 slider and a feeling picked from a list of 45 emotions.
- After eating: the same two readings again, minutes later, on the same meal.
This page covers the whole series; see each article for the subset of meals it uses.
I feel food (2026). Methodology: how I feel food measures emotional eating. I feel food tracker data, 1,534,951 meals, 2021 to September 2026. https://www.ifeelfoodapp.com/about/methodology
BibTeX
@misc{ifeelfood2026methodology,
author = {{I feel food}},
title = {Methodology: how I feel food measures emotional eating},
year = {2026},
howpublished = {\url{https://www.ifeelfoodapp.com/about/methodology}},
note = {I feel food tracker data, 1,534,951 meals logged 2021 to September 2026}
} Changelog
- — Dataset and methodology established with the publication of the ten-part "Emotional eating, measured" series.
- — This page published, consolidating the per-article methodology notes and "Cite this data" boxes into one reference.