Due to the individualized average, information about the actual distribution of water use is obscured. That could be between household members, or households vs. businesses. The unit of observation is the water billing unit, but the unit of analytical claim is one average German inhibitant. There could be other analytical claims calculated from the water billing units, e.g. regional or neighborhood-based water use. Or there could be other observation units than the billing units / households, e.g. smart meters at every single tap, to allow for other analytical claims.
It is not documented, how the measurement is standardized. However, it is interesting to note that the dataset is specifically about drinking water use and not e.g. wastewater. If water is taken from the tap for cooking or drinking, it does not directly produce wastewater in the same household. Whereas the freshwater use of bathing, showering or the toilet flush should balance out with the wastewater produced by the same process.
By reading the dataset it occured to me that in most households, different water connections are built in for different purposes already: the kitchen tap is generally used for food / drinking or washing hands, the dishwasher usually has a separate connection, all the toilet connections have a different purpose, etc. This makes distinguishing different water usecases easier, but uncertainties remain: Without a dishwasher, it is difficult to distinguish between cooking and washing hands or dishes from the kitchen tap.
The general concept of water use is divided for the dataset into categorical variables of different water usecases. It seems as if the dataset grounds roughly on the distinction of water pipes / connections: in the kitchen, for the toilet, in the garden, for the dishwasher / washing machine, etc. One category however is a re-combination of broken-down use cases: room cleaning, car cleaning and garden wartering are counted as one category.