Nobody has demonstrated that the app is the active ingredient#
You can count calories with a notebook, and the reason is not nostalgia — it is that the software has never been shown to be the part that works. A Cochrane review of 18 studies in 2,703 people found that a smartphone app made little to no difference to body weight at 12 months (MD −2.5 kg, 95% CI −6.8 to 1.7; low certainty) and none at 24 months (MD 0.7 kg, 95% CI −1.2 to 2.6; moderate certainty), concluding that the evidence "does not demonstrate a clear benefit" of apps for adults or adolescents with overweight or obesity1.
The cleanest single trial makes the point sharper still. Two hundred and twelve primary-care patients were randomised to usual care with or without help installing a free calorie app. The app arm genuinely tracked more — 2.0 additional days per week of calorie-goal tracking (95% CI 1.1 to 2.9; p<0.001) — and lost no more weight: a between-group difference of −0.30 kg (95% CI −1.50 to 0.95; p=0.63)2. Extra tracking days, purchased by the tool, bought nothing. That is the finding a pen-and-paper method rests on, and the rest of this article is about what such a method has to contain instead — since a page cannot do the one thing the app was for, which is look things up. This is for people who still want a total, and are willing to produce it themselves.
Head to head over six weeks, the notebook held#
There is a direct comparison, and it is small enough that its main use is to bound the difference rather than to measure it. Fifty adults, mean age 26, were randomised for six weeks either to a dietary self-monitoring app or to a paper diary. Days recorded came out at 18.5 (SD 14.1) for the app and 15.5 (SD 10.1) for paper — a gap of three days that did not approach significance (p=.67). Weight fell 0.4 kg (SD 1.6) in the app arm and 1.4 kg (SD 2.7) on paper, with no significant difference between the arms (p=.33)3.
Do not over-read the direction. Fifty people over six weeks cannot resolve a one-kilogram difference, the sample was young and mostly male, and the paper arm's within-group significance is a fragile thing to hang anything on. Two of the authors were employed by the company that built the app being tested — a disclosure worth naming, and one that runs against the result rather than for it, which is a mildly reassuring direction for a conflict to point. What the trial supports is modest and sufficient: over a short horizon, the medium did not decide the outcome. The known long-run cost of paper is that it decays faster than a phone does, which is a separate finding and belongs with the setup decision rather than with the method.
A record that cannot look anything up has to round#
Here is the objection worth taking seriously: without a database, every number you write is a guess, and a diary full of guesses seems worthless. The evidence on deliberately coarse records says otherwise.
Thirty-eight adults from racial and ethnic minority groups were randomised for twelve weeks to detailed self-monitoring — logging every food and drink in a full tracking app — or to a simplified version consisting of a daily checklist of high-calorie "red zone" foods. Weight loss was effectively identical: −3.4 kg (95% CI −4.6 to −2.2) detailed against −3.3 kg (−4.4 to −2.2) simplified. Adherence was not: participants completed the diet-tracking task on 49% of days in the detailed arm and 97% in the simplified one, and satisfaction with the tracking method ran 56% against 95%4.
Thirty-eight people is a pilot and twelve weeks is short, so the equal weight loss is "not distinguishable here", not "proven equal" — and a checklist is not a calorie count, so this does not license throwing the number away. What it does license is the resolution question. A record can drop a great deal of precision and still change behaviour, which is exactly the trade a paper log makes.
A paper log cannot look anything up. Every number in it is one you produced rather than one you accepted — that is its cost, and also the only reason it stays short enough to keep.
And the precision you are giving up is smaller than it looks, because the database you would have consulted is not a fixed point either: the same food is legitimately calculated several different ways, and two apps can disagree without either being wrong. Rounding a chicken thigh to the nearest fifty calories puts you inside the spread the lookup would have handed you anyway. That is my reading of the disagreement, not a published equivalence — but it sets the standard correctly. You are not competing with truth. You are competing with a number that was itself an estimate.
The method, in one page#
A workable paper system has four parts and takes ten minutes to set up.
Write the anchor list first, on the inside cover. Fifteen to twenty foods you actually eat, each with one rounded number for the portion you actually take — not per 100 g, per your serving. Reference values for the usual suspects are collected in calories in common foods; copy the ones that describe your kitchen and ignore the rest. This list is the whole database, and it is small because your diet repeats.
Record in countable units, never in adjectives. "Two slices", "one scoop", "half the tray", "three fingers of cheese". The unit does not need to be standard; it needs to be one you will reproduce next Tuesday. A page that says "some pasta" is a page you cannot total.
Give each line a range, not a point. Two columns — a low and a high — take no longer to write than one and stop you performing a precision you do not have. Sum both at the end of the day and you get a band. If the band is 400 calories wide, that is a fact about the day, not a defect in the notebook.
Price the invisible things separately. Oil, butter, dressing and drinks are the lines a mental tally drops, and fat costs more than twice as much per gram as anything else on the plate. One habit — a standing assumption for the cooking fat, written in every time a pan was involved — closes most of the gap that no anchor list can. Running the calories-per-gram check on a suspicious day is the one piece of arithmetic worth doing by hand.
| Entry | What you write | What you do not write |
|---|---|---|
| Breakfast | oats, 1 scoop, 150–180 | 163 |
| Coffee | flat white, 1 large, 130–170 | "coffee" |
| Lunch | leftovers, half the tray, 500–700 | a reconstructed ingredient list |
| Cooking fat | pan used at dinner, 100–150 | nothing at all |
| Day | 1,850–2,300 | 2,041 |
The right-hand column is where paper logs usually die: not from imprecision but from the attempt to be precise, one ingredient at a time, until the page is a chore. Setting the resolution low on purpose is what keeps the record daily, and daily is the property that has evidence behind it — the argument the counting method makes at length.
Two things paper genuinely cannot do, and it is worth deciding in advance whether you mind. It has no running total during the day, so it reports a verdict at bedtime rather than a signal at lunchtime. And it cannot remind you, which means the cue has to come from somewhere else — a fixed slot, a page left open on the counter, a pen that lives beside the kettle. Those are solvable. The database was never the hard part.
FAQ#
How do I get a calorie number without looking anything up?#
From a short anchor list you write once. Fifteen to twenty foods, each with one rounded figure for the portion you actually serve yourself, covers most of a repeating diet; anything outside the list gets a deliberately wide two-number range. Rounding to the nearest 50 keeps you inside the spread a database would have given you anyway, since the same food is legitimately calculated more than one way.
Does counting on paper actually work without a database?#
The comparisons are small but unembarrassing. Randomised over six weeks, a paper diary and a tracking app produced no significant difference in days recorded (15.5 vs 18.5, p=.67) or in weight change (p=.33)3. More broadly, a Cochrane review of 18 app trials found no clear benefit from the software itself1. The medium is not what produces the result.
How much detail can I drop before the log stops working?#
More than most people expect. Randomised against full app-based tracking, a daily checklist of high-calorie foods produced essentially the same twelve-week weight loss (−3.3 vs −3.4 kg) while being completed on 97% of days against 49%4. That is a pilot in 38 people, so read it as a direction — but the direction is that resolution is the cheapest thing in a food record to spend.
Sources#
- Metzendorf MI, Wieland LS, Richter B. Mobile health (m-health) smartphone interventions for adolescents and adults with overweight or obesity. Cochrane Database Syst Rev. 2024;2:CD013591.
- Laing BY, Mangione CM, Tseng CH, et al. Effectiveness of a smartphone application for weight loss compared with usual care in overweight primary care patients: a randomized, controlled trial. Ann Intern Med. 2014;161(10 Suppl):S5-12.
- Ahn JS, Lee H, Kim J, Park H, Kim DW, Lee JE. Use of a smartphone app for weight loss versus a paper-based dietary diary in overweight adults: randomized controlled trial. JMIR Mhealth Uhealth. 2020;8(7):e14013.
- Patel ML, Cleare AE, Smith CM, Rosas LG, King AC. Detailed versus simplified dietary self-monitoring in a digital weight loss intervention among racial and ethnic minority adults: fully remote, randomized pilot study. JMIR Form Res. 2022;6(12):e42191.
