Three weeks is the shortest stall worth diagnosing#
If your weight has been flat for a week, you have learned almost nothing about your fat. If it has been flat for three weeks and your waistband agrees, you have learned something. The distance between those two sentences is arithmetic, not attitude. Day-to-day body mass carries a standard deviation of roughly half a percent of bodyweight, and a good rate of fat loss moves roughly half a percent of bodyweight per week. A whole week of real progress is about the size of a single day's ordinary variation — so a week-long flat spell is exactly what you would expect to see whether you were losing fat or not.
That one comparison is the decision rule. Everything below is how to apply it: how large the noise actually measures, why your weight runs on a weekly rhythm that makes a Monday-versus-Friday comparison worthless, how much averaging buys back, and what the popular "whoosh effect" is worth as evidence. Why a genuine stall happens and what to do about it are settled elsewhere in this cluster — this page is the prior question of whether you have one at all.
The noise has been measured, and it is bigger than the signal#
The most detailed record of ordinary weight variation comes from an unusual dataset: one healthy adult weighed under standardized conditions across nearly 30 years, 9,521 days of measurements. The standard deviation of the difference between consecutive days was 0.53% of body mass, rising to 0.69% across 7-day intervals, and the mean day-to-day difference was zero — the fluctuation neither grew nor shrank over three decades1. The authors' practical summary was that about 0.6% of body mass — roughly 450 mL of fluid in a 75 kg person — is the variability a clinician should expect to see and ignore.
Set that against the thing you are trying to detect. A sustainable rate of loss is 0.5–1% of bodyweight a week. For an 80 kg adult, the following is what each quantity is worth (the kilogram conversions are my arithmetic on the percentages each study reports):
| Quantity | Size for an 80 kg adult |
|---|---|
| Day-to-day standard deviation (0.53%) | ~0.42 kg |
| A domestic scale's own measurement error | ~0.1 kg |
| Weekly rhythm, weekend peak to weekday trough (0.35%) | ~0.28 kg |
| One week of loss at 0.5% per week | ~0.40 kg |
| The Christmas swing (1.35%) | ~1.1 kg |
Read the fourth row against the first. The instrument is not broken and the scale is not lying; it is reporting a real quantity whose week-to-week signal is the same size as its day-to-day scatter. Nothing you do to a single morning's reading — same time, empty bladder, same floor tile — shrinks that scatter much, because most of it is not measurement error at all. The device contributes about 0.1 kg. Your body contributes the rest, for the physiological reasons catalogued here.
The week has a shape, so compare like with like#
Before you average anything, there is a systematic error worth removing, because it is free. Body weight runs on a weekly cycle. Tracking 1,421 adults in a European weight-loss-maintenance study across hundreds of daily measurements each, researchers found weight peaking on Monday and bottoming out on Friday, with a weekly fluctuation of around 0.35% of body weight — larger in men (0.41%) than women (0.29%)2. Weight rose across the weekend and came back down through the working week, in every subgroup they looked at.
The consequence is immediate and nobody applies it. A Friday weigh-in compared against the previous Monday will show a loss of about a quarter of a kilogram that has nothing to do with your fat, and the reverse comparison will invent a gain of the same size. If you weigh weekly, weigh on the same weekday. If you weigh daily, use a full 7-day window so the rhythm cancels rather than leaking into your trend — the reason your body's accounting period is closer to a week than a day applies to how you read the output too.
The same dataset found a much larger seasonal artefact: weight rose 1.35% from early December to early January, and was still at least 0.35% above the pre-Christmas level the following March. A stall diagnosed in January is often a holiday still clearing.
What averaging buys, and where it stops#
Averaging is the only real fix, and it is worth knowing exactly how much it buys. If you treat consecutive days as independent draws, the standard deviation of a 7-day mean is the daily figure divided by the square root of seven: 0.53% becomes about 0.20%, or 0.16 kg for our 80 kg adult. That is the arithmetic behind every trend-weight feature, and it is mine, not any paper's — the studies above report the daily and weekly variability; the square-root step is textbook statistics applied to their numbers.
Now run the detection problem. Two weekly averages one week apart differ by about 0.40 kg if you are losing at 0.5% per week, and the noise on that difference is roughly 0.23 kg — under two standard errors, which is to say a coin-flip-ish signal you would be foolish to act on. Three weeks apart, the same rate has moved about 1.2 kg against the same 0.23 kg of noise, and the answer is no longer in doubt. That is where the familiar "two to three weeks" advice comes from, and it is why the advice is not arbitrary caution.
One honest caveat cuts against the arithmetic: consecutive days are not independent. Glycogen loads, salt loads and a heavy training week persist for several days, so the effective number of independent samples in a week is fewer than seven, and the real standard error of a weekly mean is somewhat larger than 0.16 kg. The three-week figure is therefore a floor, not a guarantee. Averaging more days always helps; it just helps less than the square root implies.
The "whoosh effect" is a story without a record behind it#
The folk explanation for the sudden multi-day drop after a flat stretch is the whoosh: fat cells supposedly hold water in place of lost fat, then release it all at once. It is repeated everywhere, and it does not survive a search of the literature. There is no peer-reviewed study establishing that mechanism, no measurement of the water volume it proposes, and nothing to cite for it that is not a blog post citing another blog post. This blog's rule is that a claim that lives only on content pages does not get laundered into a citation, so it does not get one here.
What is genuinely documented is the ordinary version: fluid shifts of exactly the size in the table above can mask a fortnight of fat loss and then clear over a couple of days, producing a drop that looks like a sudden event but is only the arrears of a trend that never stopped. You do not need fat cells to be storing water for that to happen — you only need the noise to be larger than the weekly signal, which the measured numbers say it is. The practical difference matters: the whoosh story implies waiting for a release, while the measurement story implies your average was already falling and you simply could not resolve it yet.
The four-question test#
When the trend really has been flat, the useful question is what else was flat with it, because the instruments fail in different directions.
| Question | Answer that means "noise" | Answer that means "stall" |
|---|---|---|
| How long has the 7-day average been flat? | under 3 weeks | 3–4 weeks or more |
| Are you comparing the same weekday, or full weeks? | no — fix this first | yes |
| Has your waist or the fit of your clothes moved? | yes, still changing | no, flat too |
| Anything obvious loading you with fluid — a holiday, a salt-heavy week, a new training block, a hormonal phase? | yes | no |
One piece of context makes the whole test easier to hold: a plateau is the expected end state, not a malfunction. Modelling of weight-loss trajectories puts the plateau after dietary restriction at roughly a year, arriving much later for GLP-1 drugs and bariatric surgery, and attributes the difference to how strongly appetite feeds back on intake rather than to changes in energy expenditure3. If you are three months in and the weekly average is still drifting down through the scatter, the model says you are on schedule and the scatter is the only thing you are actually looking at. The instruments that resolve the change your scale can't — and their own error bars — are covered in better ways to track progress.
FAQ#
How many days of weigh-ins do I need before a flat trend means anything?#
About three weeks of daily readings, averaged. Day-to-day variation has a standard deviation near 0.53% of body mass1, while a week of loss at 0.5% per week is the same order of magnitude — so one week of data cannot separate them. Three weekly averages can: by then a real 0.5%/week trend has moved several times the noise on the comparison. Fewer independent days than you think sit in each week, because glycogen and fluid shifts persist, so treat three weeks as the minimum rather than the target.
Why is my weight always highest on Monday?#
Because it is, for almost everyone. Across 1,421 adults tracked with daily home scales, weight peaked on Monday and reached its lowest point on Friday, with a weekly swing of about 0.35% of bodyweight — 0.41% in men, 0.29% in women2. Weekend eating and drinking patterns load the number, and the working week unloads it. It is not fat arriving and leaving; it is a rhythm, and the fix is to compare Monday with Monday or to average whole weeks.
Does the whoosh effect actually exist?#
There is no peer-reviewed evidence for the specific mechanism — fat cells holding water in place of lost fat and releasing it in one go. What is real is that fluid shifts of half a kilogram or more routinely mask a fortnight of fat loss and then clear over a day or two, which looks identical from the outside. The distinction is practical: nothing has to be "released" for the drop to arrive. Your average was already falling; the noise was simply larger than the week's progress.
Sources#
- Schneditz D, Hofmann P, Krenn S, Waller M, Mussnig S, Hecking M. Day-to-day variability in euvolemic body mass. Ren Fail. 2023;45(2):2273421.
- Turicchi J, O'Driscoll R, Horgan G, Duarte C, Palmeira AL, Larsen SC, Heitmann BL, Stubbs J. Weekly, seasonal and holiday body weight fluctuation patterns among individuals engaged in a European multi-centre behavioural weight loss maintenance intervention. PLoS One. 2020;15(4):e0232152.
- Hall KD. Physiology of the weight-loss plateau in response to diet restriction, GLP-1 receptor agonism, and bariatric surgery. Obesity (Silver Spring). 2024;32(6):1163-1168.
- Turicchi J, O'Driscoll R, Finlayson G, Duarte C, Palmeira AL, Larsen SC, Heitmann BL, Stubbs RJ. Data Imputation and Body Weight Variability Calculation Using Linear and Nonlinear Methods in Data Collected From Digital Smart Scales: Simulation and Validation Study. JMIR Mhealth Uhealth. 2020;8(9):e17977.
