Missing monitoring data

Gaps are shown as gaps. What each kind of gap means and whether it is a problem.

Monitoring data has gaps. Every system that measures the real world does. What matters is that a gap is drawn as a gap, and never filled with an estimate, an interpolation or a zero.

The kinds of gap

Before monitoring began

A provider monitored for twelve days has seventy-eight unmeasured days inside a ninety-day window. Those days are drawn as unmeasured, and the ninety-day figure is withheld with a note saying how much of the window exists. The metric is shown from the first day, so a reader can see it exists and is filling up, but the number waits until it is true.

A reporting gap

Days inside the record with no measurement: usually a period when too few workers were reporting. These look identical to the pre-monitoring stretch on the timeline, deliberately, because from a reader's point of view they are the same thing: a day nobody measured.

A region with no coverage

Nobody is running a worker there. The region is absent from the breakdown rather than shown as failing. If you want a region covered, the answer is a worker in it: see Installing a worker.

Not enough agreement

Measurements exist, but too few independent workers to be confident. Reported as "not enough data", with the history still visible.

What is never done

• No interpolation. A missing Tuesday is not filled in from Monday and Wednesday.
• No zeros. An unmeasured day is not 0% uptime. It is not a number at all.
• No borrowed figures. A provider's missing region is not estimated from a similar provider's.
• No rounding a gap away. A ninety-day average is not computed over the twelve days that exist and labelled ninety.

Reading a sparse timeline

Coverage is stated under the strip: "12 of 90 days measured" means what it says. A provider with a sparse timeline is not a provider with poor uptime, and the page is careful never to let the two look alike.