An outlier is a single reading that sits clearly above or below your usual level without any change in your body fat. Typical triggers are a salty restaurant meal, a high-carb day, a hard leg session, travel or, for many women, the menstrual cycle.
How much an outlier moves the trend
A well-smoothed trend absorbs single outliers. How strongly it reacts depends on the method.
Example: Your level is 80.0 kg and one morning you read 81.5 kg. That same day the trend rises to 80.15 kg with exponential smoothing, the default in Lean is law, to 80.11 kg with a 14-day average and to 80.21 kg with a 7-day average. None of the methods takes the outlier seriously; the differences are in the tenths. This is a worked example.
Moving averages have one quirk: the outlier keeps acting while it sits in the window and then drops out all at once. That makes the line jump after seven or fourteen days. With exponential smoothing it fades evenly.
Log it or delete it?
Log outliers. The day after a restaurant dinner is a real value, not a measurement error. Recording only the good readings distorts your history more than it improves it. The option to ignore a reading is meant for genuine measurement errors, for example when you weighed yourself with clothes on. When in doubt, keep the value; the trend evens it out anyway. A short note about the day helps you explain the spike later.