Skip to main content
Accuracy tells you whether an investigation got the cause right. Time to usefulness looks at how quickly it helped your responders along the way. An investigation can help long before anyone names the cause, by telling responders something they didn’t know yet. For each incident, we look for the moments where the investigation told responders something useful. We call each one a useful signal, and the first one gives us the incident’s time to usefulness.

​
What counts as useful

An investigation gives a useful signal in one of three ways:

​
When a heads-up counts

Once your responders have established the incident’s cause, we grade each heads-up against it. A heads-up counts as useful when it was:
  • Accurate. It graded On target or better, on the same scale as investigation accuracy.
  • New. It told responders something they hadn’t already worked out. A heads-up that repeats what responders said in the channel doesn’t count, however accurate it is.
  • Early. It was sent before the root cause was understood, whoever found it.

​
When naming the cause counts

We only credit the investigation with naming the root cause when:
  • it was the first to state the cause
  • responders didn’t already know the cause when the incident was declared
  • it worked the cause out itself, rather than passing on something like an alert or an attached pull request

​
An example

1

The incident is declared

Checkout starts returning 500s and a responder declares an incident. This starts the clock.
2

A heads-up that doesn't count

Within a minute, the investigation posts a heads-up that checkout is failing. It’s accurate, but responders already said so when they declared the incident, so it isn’t new.
3

The first useful signal

Three minutes in, the investigation posts a heads-up that the errors all come from the payments service, which is timing out on its database connections. Nobody had said this yet, and it points at the real cause, so this is the first useful signal. The time to usefulness is 3 minutes.
4

A responder names the cause

A responder connects the timeouts to a deploy that cut the connection-pool size. The root cause is diagnosed here, but by a responder, so it isn’t a useful signal from the investigation.

​
Time to usefulness

An incident’s time to usefulness is the time from the incident being declared to its first useful signal, whichever kind of signal that is.

​
Seeing it in your dashboard

You can see time to usefulness under Speed on the Investigations page in your dashboard. It shows incidents grouped by their time to usefulness, as well as the median. In the investigations table, the First useful signal column shows each incident’s time to usefulness. Hover over it to see which signal it was. On the investigation’s timeline, the step that gave the first useful signal is marked First useful signal.

​
Which incidents are included

An incident counts toward time to usefulness when:
  • We could grade it. At least one of its heads-ups, or the investigation itself, has been graded against the incident’s cause. An incident with no known cause can’t be graded.
  • The investigation started within 5 minutes of the incident being declared. This keeps the measure about how quickly the investigation helped, not how long it took someone to start one.
Like the rest of the Investigations page, time to usefulness appears once at least 10 of your incidents are included.

​
FAQs

Time to diagnosis measures how long it took anyone, responder or investigation, to find the root cause, starting when the investigation started. Time to usefulness measures how long it took the investigation to guide responders in the right direction, starting when the incident was declared. A heads-up that points at the failing service is useful even if a responder names the cause later.

Measuring accuracy

How we grade investigations against what really caused your incidents.

Measuring time to diagnosis

How long your incidents take to reach a diagnosis once an investigation starts.

Measuring engagement

How responders read, steer, and act on what an investigation finds.

Incident channel experience

How investigations show up for responders while an incident is live, including heads-ups.