Scale weight is noisy
Day to day, the scale mostly measures water, food in transit, and yesterday's sodium. Swings of a kilo that have nothing to do with fat are completely normal. Reacting to them would mean reacting to noise — so Gravl Macros doesn't.
The trend line
Your trend weight is a smoothed average that gives recent weigh-ins more weight while letting older ones fade. It moves slowly and deliberately, which makes it the number worth watching: when the trend drops, you're actually losing; when a single morning spikes, the trend barely blinks.
The weight chart shows both — your raw entries as the solid line, the trend dashed through them. Pick a short enough period and each reading gets its own dot; over a longer stretch they are dropped, because a dot per day on three months of mornings reads as noise rather than data.
One month of mornings
The same person weighed every morning. The scale swings 1.8 kg end to end; the trend moves half a kilo.
- Scale weight
- Trend weight
How much a weigh-in still counts, later
A weigh-in is never thrown away — it fades. Its influence halves about every eight days, so the last three weeks carry most of the line and a single strange morning is outvoted by the ones around it.
How much one morning counts
The smoothing has one setting, and it is worth knowing because it explains everything else the trend does: each new weigh-in contributes 8% of itself to the line, and the previous trend carries the other 92%.
Run that forward and a weigh-in's influence halves roughly every eight days. Three weeks of mornings carry most of the line; anything older than a month is a rounding error. That is the whole mechanism.
Two consequences follow, and they are the two questions people actually ask:
- A bad morning cannot ruin your trend. In the figure above, a salty dinner puts 1.6 kg on the scale overnight. The trend moves 70 grams. You do not need to skip a weigh-in because you know it will read high.
- The trend lags real change, on purpose. If you genuinely drop a kilo, the line takes a couple of weeks to fully show it. That lag is the price of not chasing noise, and the expenditure algorithm is built around it.
There is no outlier filter
Worth saying plainly, because it is the opposite of what most people assume: the app does not detect strange weigh-ins and throw them out. Nothing is flagged, discarded or corrected. A holiday weigh-in three kilos above your normal goes into the line like any other, carrying the same 8%.
This is deliberate. A filter that decides which of your weights are "real" would eventually discard a real change — the first week of a diet break, say, or genuine water retention from a new training block. Diluting every reading equally is more honest than guessing which ones to believe.
The practical version: log every weigh-in, including the ones you don't like. The maths already assumes some of them are strange.
Your rate of change
The kg per week figure you see is not the difference between two weigh-ins. It is the slope of a straight line fitted through the last 20 days of your trend, by least squares.
That matters because it means your rate reflects three weeks of direction, not the accident of which two mornings you happened to compare. It also means it needs history: with fewer than three days of trend the app shows no rate at all, and it will not call a direction — gaining, losing or maintaining — until there are at least seven.
Between those, one threshold: a rate under 0.1 kg a week in either direction reads as maintaining rather than a very slow gain or loss. Below that, the movement is not distinguishable from noise.
Days you skip
Missing a day is not a problem. The app draws a straight line between your two real weigh-ins and fills the days in between, then smooths over the result.
This is worth knowing if you weigh in weekly rather than daily: the gaps get bridged, so your trend is a slope rather than a staircase. It still works. It is simply working from less evidence, and a single weekly weigh-in carries far more of the line than one of seven daily ones would.
Where your weigh-ins come from
All your sources land in one place: weights logged in Gravl Macros, imported from Apple Health or Health Connect (a smart scale, for example), brought in from another app, and weigh-ins from Gravl Workouts.
If the same day has several entries, the app keeps one value for that day, in this order of precedence:
| Priority | Source |
|---|---|
| 1 | A weight you logged in Gravl Macros |
| 2 | Apple Health or Health Connect |
| 3 | Imported from another app |
| 4 | A weigh-in from Gravl Workouts |
Your own entry always wins. A scale that syncs through Health later in the day cannot quietly overwrite a correction you typed yourself.
What it feeds
Trend weight is an input everywhere it matters: the expenditure algorithm reads its movement, the weekly check-in compares it against your goal pace, and your progress screens chart it. One good habit — stepping on the scale regularly, ideally at a consistent time of day — powers all of it.
One thing it does not feed: your body measurements. Those are smoothed separately and much more reactively, because a tape measure gets used every few weeks rather than every morning, and a line that took three weeks to notice a change would never catch up.