Method
A short note on where the numbers come from and how far to trust them.
Where the numbers come from
The flat has three water meters: on the staircase riser, in the bathroom and in the wardrobe. They are read once a month, around the 10th.
A month's usage is this month's reading minus the previous one. A month here means the month of the reading, so “November” covers roughly 10 October to 10 November.
Litres per day and comparisons
Litres per day is the month's usage divided by the number of days between readings.
Each month is compared with the one before, with the all-time average and with the same month a year earlier. Green means less water, orange means more.
Litres per day by who is home
Using all months together, we fit two numbers: litres per day with two people at home, and with one. With nobody home the water is shut off, so usage is zero.
The fit is least squares with no intercept. Only the all-time result is shown.
What a day is made of
For each meter, usage is split into three parts: a baseline while the water is on, an amount per person per day, and an amount per laundry load.
Laundry is a lower bound: regular loads and post-workout loads are counted, the rest are unknown. Treat “per load” as a rough guide, not a precise figure.
How far to trust it
R² is the share of month-to-month variation the model explains: 1 is all of it, 0 is none. Ours is around 0.5–0.6, so it gets the overall picture but is rough for any single month.
The meters are not read on exactly the same day and hour, so a hundred litres or two can slip from one month into the next.
What is not here
The source spreadsheets are private. Nothing here says who was home or when — only usage, readings and all-time estimates.
The pages make no external requests: no analytics, third-party fonts or scripts.