How these numbers are worked out
How long a disruption lasted
The time from a notice going up to its stated “works are now complete” update. Incidents whose only end signal was a scheduled finish are counted separately - a published plan is not evidence the works ended then. End times are read from the notice text (see the accuracy evaluation); no time on this page is a measurement taken in the field.
About one incident in twenty never reports an end at all - the count is shown beside the county's disruption count each month. Those stay out of the figure above: checked both ways, they run shorter than average, so leaving them out moves the typical duration up, not down. They still count in full in the availability percentages and person-hours, where leaving something out would mean calling it zero.
When a notice closed
A notice counts as closed on the day its own update reports the works complete. One that never says so counts as closed the first time a twice-daily check finds it gone, so those dates run late and the closed-in-month lists are minimums.
What counts as a disruption
Only hard disruptions count against availability: bursts, reservoir / treatment-plant / pump-station interruptions, pump failures, power outages and unplanned repairs. Restrictions, low pressure, discolouration and routine works are shown but never count. A closed notice with no usable end is charged the typical length of a completed notice of the same kind of works, rather than nothing at all.
Repeating works
Works announced as a repeating window - “daily from 10pm until 7am, from 9 to 27 July” - are charged only the hours inside the windows, and are treated as a restriction whatever the title says.
How many people a notice affects
Each notice is assumed to affect the Census 2022 Small Areas whose centres lie within 500 m of its pin. Availability weighs each notice by how many people it covers and how long it stands: the percentage is the share of all that person-time no notice covered. One notice is charged at most 14 days.
What the letters mean
Supply availability over the month, and nothing else:
- A 99.9% or better
- B 99.75% or better
- C 99.45% or better
- D 99.0% or better
- E 98.7% or better
- F below 98.7%
Why the ! sits beside the grade, not inside it
Whether the water is safe to drink is a different question from how much of it there was, and a letter cannot answer both - so an active boil-water / do-not-drink / do-not-consume notice is a marker beside the grade, never folded into it.
The ten largest
The ten largest disruptions of each finished month rank the same hard disruptions as individual incidents, with person-hours clipped to the month so they sum against the county figures.
Full methodology: the full methodology, on GitHub.