The short answer#
A kilogram of body fat is about 7700 kcal. To genuinely gain one in a day you would have to eat 7700 kcal above maintenance in that day.
The scale can move that far by lunchtime. So whatever it is showing you between one morning and the next, it is almost never fat.
What is actually moving#
Salt, and the water it holds#
Sodium holds water. A takeaway, a restaurant meal or a bag of crisps can put a kilogram on the scale the next morning and take two or three days to leave. This is the single most common cause of an alarming Monday.
Carbohydrate and glycogen#
Glycogen — stored carbohydrate — is held with water attached, roughly three grams of water per gram of glycogen. A full glycogen load is a couple of kilograms of water above an empty one.
This is why the first week of a low-carb diet produces a dramatic drop that then stops, and why the number comes straight back when carbohydrate does. Neither direction is fat.
What has not left you yet#
Food and drink in transit weigh what they weigh. A large late dinner is still substantially present at 7am. Constipation shows up on the scale, and so does its resolution.
The menstrual cycle#
Water retention across a cycle routinely moves the scale by one to two kilograms, on a schedule that has nothing to do with what you ate. If you track a cycle, you will usually find the same shape repeating month to month — which means the useful comparison is to the same point last month, not to last week.
Hard or unfamiliar exercise#
Muscle repair holds water. A first session back after a break, or a new lifting programme, can add weight for several days. The scale going up after you started training is normal and temporary.
Everything else#
Hydration, alcohol, heat, altitude, a long flight, a different scale, carpet versus tile, clothes.
How long each one lasts#
| Cause | Typical size | How long |
|---|---|---|
| Salty meal | 0.5–1.5 kg | 1–3 days |
| Carbohydrate refill | 0.5–2 kg | 1–3 days |
| Gut contents | 0.5–1.5 kg | hours to days |
| Menstrual cycle | 1–2 kg | days, cyclical |
| New training load | 0.5–1.5 kg | 3–7 days |
| Actual fat change | 0.1–0.15 kg/day at most | permanent-ish |
The bottom row is the point. Real change is slow and small; everything above it is fast and large. Which is why a single reading is dominated by noise, and a run of readings is not.
Reading through it#
The fix is arithmetic rather than willpower. Draw a line through a run of weigh-ins and the noise cancels; take two readings and subtract, and you have kept the noise and thrown away the signal.
SmartCrunch reads a 7-day trailing moving average for everything that matters — the trend, the forecast date, and the maintenance figure the budget hangs off — and fits the slope by least-squares across every reading rather than by comparing the ends. It also will not attempt that calculation until your weigh-ins span at least a fortnight, precisely because a shorter series has no slope that can be distinguished from the table above.
When a fluctuation is not a fluctuation#
Two weeks is the honest waiting period. After that:
- The average is flat and you are logging accurately. That is not water. Your maintenance is probably not where you assumed, and the app’s adaptive method exists for exactly this case.
- The average stepped up and stayed for a fortnight. Look for a structural change — a new medication, a new job, a training block that ended.
- The average is moving as intended and the dots are jagged. Nothing is wrong. That is what this looks like.
One thing worth saying#
Sudden or persistent weight change that has no explanation in the list above is a medical question, not a tracking question. This app counts things; it does not diagnose anything and cannot tell you why a trend moved. The health disclaimer is the boundary.
Read next
- How often should you weigh yourself? Daily weighing gives the best signal and the worst individual readings. The fix is arithmetic, not willpower: read the average, never the morning.
- Maintenance calories, explained Maintenance is the intake at which your weight holds. Every calculator predicts it from a population; your own logged weeks can measure it instead.