Use case

Watch a launch day

You ship at nine. The question for the next few hours is not "how did the quarter go" but "is this working", and it is a different question with a different instrument.

The situation

On a normal day nobody needs to watch analytics. On a launch day everybody does, and the reports built for months are the wrong shape for it. By the time a daily figure is meaningful, the moment to react has gone.

The other half of the problem is that a launch attracts attention you did not ask for, and some of the traffic arriving is not people.

How it is done

1. Open the live view. Sites and apps each have one. It answers "is anything arriving at all", which is the first thing that goes wrong and the fastest to rule out.

2. Sample recordings up, briefly. A launch is exactly when the extra cost is worth it, and exactly when a new page meets people who have never seen it. Turn it back down afterwards.

3. Watch stability, if it is an app. Crash-free rate for the new version alone, not in total.

4. Expect noise, and act on the alert. A site ID is public, so anybody can send hits to it. Skomi flags an address sending implausible volume and hands you the one fact you need to exclude it.

5. Write the day down. An annotation now saves an argument in three months.

Where this stops

It is measurement, not monitoring. Nothing here pages you at two in the morning, and none of it is uptime monitoring. If the site is down, Skomi sees no traffic and so do you, which is not an alert.

Real time is not a trend. Data appears in seconds; meaning takes longer. A quiet first hour is normal and says almost nothing. See how long before analytics tells you anything.

Recording started today misses yesterday. Sampling up on the morning of a launch gives you the launch, not the week before it.