What is a normal overnight reading
On a property with every outlet closed the normal overnight movement is none at all, because nothing should be drawing water. Over an 8 hour night, a movement of 4 litres is a rate of 0.5 litres per hour, and that is the point at which a movement stops being noise and starts being worth acting on.
Last reviewed 14 August 2026
The short answer
Nothing. A property with every outlet closed should use no water at all, so the honest expected reading is zero movement, and anything else is either a leak or something still running.
What matters is how much else. The severity bands on this site start at 0.5 litres per hour, and the table below converts that into what you will actually see on the register after a night.
| Movement over 8 hours | Rate | Band |
|---|---|---|
| 0 to 3.2 litres | under 0.4 L/hr | At or below what the meter can reliably register |
| 4 litres, or 0.004 m³ | 0.5 L/hr | The bottom of the first band worth acting on |
| 40 litres, or 0.040 m³ | 5 L/hr | Top of the small leak band |
| 480 litres, or 0.480 m³ | 60 L/hr | Top of the significant band |
Every figure in that table is the rate multiplied by eight hours, and nothing else. There is no assumption in it to disagree with.
How this works in practice
The reason overnight is the right window is that it is the longest stretch in which a household naturally uses nothing, and length is what makes the measurement good.
The rate is the volume divided by the hours. So a reading error of one litre, which is a single digit misread on the register, becomes a rate error of 0.5 litres per hour across a two hour test and 0.125 litres per hour across an eight hour one. Over two hours a misread digit is the whole threshold. Over a night it is a quarter of it.
Two practical consequences follow, and they matter more than the choice of interval.
- Write down the times, not just the readings. An interval you have estimated afterwards carries an error straight into the rate.
- Do not use water during the night, including a single flush. One flush of a modern dual-flush cistern is comparable to several hours of a small leak, and it will land in the result as though it were one.
A reading below the bottom of the table is not a clean result either, and that is worth understanding before you conclude anything from it. why a still meter does not rule a leak out covers what an unmoving register does and does not establish.
What changes the answer
Four things, and the first is by far the most common reason an overnight test reads high.
- Something automatic that was missed. A water softener regenerating overnight is the classic, because it is designed to run when nobody is using water. An ice maker, a dishwasher on a delayed start and a heating system on an automatic filling device do the same thing.
- A toilet that had been flushed shortly before the test started, and was still refilling when the first reading was taken.
- Which side of the boundary the meter is on. A boundary chamber meter includes the underground supply pipe, so an overnight movement may be a loss outside the building entirely. A meter fitted indoors after the stop tap does not measure that pipe, so the same loss would be invisible to it.
- Whether the loss is steady. A leak that only runs when mains pressure is high, which is often overnight, reads higher on a night test than a day one. That is not an error, it is the loss behaving differently, and the overnight figure is the more useful of the two.
Assumptions and sources
Aquadis+ DN15 and DN20 specification sheet
Itron · manufacturer source
- Q1 minimum flow rate, DN15, Q3 2.5, R160: 15.6 l/h
- Q2 transitional flow rate, DN15, Q3 2.5, R160: 25 l/h
- Q3 permanent flow rate, DN15: 2500 l/h
- Typical starting flow rate, DN15: 0.4 l/h
- Typical starting flow rate, DN20: 2 l/h
- Q1 minimum flow rate, DN20, Q3 4.0, R160: 25 l/h
- Flow rate at which accuracy is within +/- 5%, DN15: 3 l/h
- Flow rate at which accuracy is within +/- 2%, DN15: 5 l/h
- Minimum scale interval, DN15 and DN20: 0.02 l
Litres in a cubic metre
Exact by definition
SI definition. The litre is defined as one cubic decimetre, so one cubic metre is exactly 1,000 litres.
Hours in a day
Exact by definition
Definition of the day as 24 hours.
Days used for an annual figure
Editorial convention
Annual figures on this site are a daily rate multiplied by 365. Leap years are not modelled, so a leap year is understated by one day, which is 0.27 per cent.
Test length below which a reading error matters
Editorial convention
A meter test is a difference between two readings, so a misread of one small division is a fixed error spread over the length of the test. The smallest division on a DN15 register is 0.02 litres, so a one division misread over two hours is 0.01 litres an hour, which is about 2 per cent of the 0.5 litres an hour band boundary. Over a twenty minute test the same misread is 0.06 litres an hour, which is more than a tenth of it. The two hours is still a chosen round number rather than a derived one, and it changes nothing in the arithmetic: it only decides when the tool adds a note.
Where the first severity band ends
Editorial convention
A practical threshold set just above the 0.4 litres per hour starting flow quoted for a DN15 meter with the TVM register. It is not a meter accuracy tolerance. Against the STANDARD it sits about 31 times below Q1, the 15.6 litres per hour at which OIML R 49-1 first defines any permitted error. Against THIS MANUFACTURER it sits about 6 times below 3 litres per hour, the lowest flow at which Itron volunteers a typical accuracy that the standard does not require. Neither figure reaches down to 0.5, so nothing quantifies a reading here on either footing. Meters with other registers start higher, so this threshold sits above the most sensitive case rather than above every meter.
The conversions in the table are arithmetic and carry no assumption: rate multiplied by eight hours, with one cubic metre as 1,000 litres.
The 0.5 litres per hour threshold is a practical band edge rather than a measured boundary. It sits just above the best published starting flow of a DN15 domestic meter, which is what makes it a sensible floor: below it the reading is at the edge of what the instrument registers at all. That published figure is footnoted to one register type and others start higher, so the floor is if anything generous. It is not a meter accuracy tolerance, and pages that describe it as one have the right number for the wrong reason.
What to do next
What the movement justifies depends on which row of the table it landed in.
- Nothing, or less than the meter can register: repeat over a longer window before concluding anything, and check the bill against the same period last year.
- Between the bottom of the first band and the top of the small leak band: check the toilets with dye and the taps for a drip, because those are the two most common causes at that rate and both are findable without help.
- Above that: close the internal stop tap and repeat the test. If the meter still moves, the loss is on the underground supply pipe rather than inside the building, which changes who is responsible for it and what happens next.
- Any movement at all: photograph both readings with their times. That record is the evidence for a leak allowance later.
To turn the rate into money, what that rate costs over a year does the conversion.
Common questions
- How small a movement is worth worrying about?
- Below 0.5 litres per hour there is nothing to act on yet. That threshold is set just above the best published starting flow of a domestic meter, 0.4 litres per hour, and meters that start higher sit further inside the range the instrument cannot register at all, so a smaller reading is at the very edge of what it can see and cannot be told apart from residual draw.
- Is any overnight movement normal on a lived-in property?
- Not if the property is genuinely isolated. Movement usually means something automatic was missed rather than that a small draw is normal: a water softener regenerating, an ice maker, a heating system on an automatic filling device, or a toilet still refilling. Those are all findable, and finding them is part of the test rather than a failure of it.
- How long should the test run?
- Longer is better and overnight is usually the longest uninterrupted window available. The rate is the volume divided by the hours, so a reading error of one litre across 2 hours produces a rate error of 0.5 litres per hour, which is the entire threshold. Across 8 hours the same error is 0.125 litres per hour.
- The meter moved but only a little. Should I test again?
- Yes, and over a longer interval rather than the same one. A second short test repeats the same measurement error. A longer one shrinks it, and it also catches an intermittent loss that a two hour window can miss entirely.