Experiment Library timeline

The Library's timeline shows every experiment on a single time axis. You can see what ran and when, what's running now, what's scheduled, and what's expected to come next in your pipelines.

Opening the timeline

In the Library, switch the Table / Timeline toggle to Timeline.

The timeline always shows every experiment, archived ones included, so you see your whole testing history in one place. The status views (Running, Draft and so on) don't apply here, but search and Filter do. Switching back to Table returns you to the view you had selected.

Choosing a time range

The timeline opens on All time, from your first experiment to the latest planned end. Choose 3 months, 6 months or 12 months for a closer view.

Scroll sideways to move through time. The timeline opens scrolled to today, so current and upcoming tests are in view and your history is to the left.

Reading the bars

Colour shows the status: running, deployed, ended or draft.

The line style shows how sure a date is:

• Solid: what actually happened
• Dashed outline: a scheduled start or stop
• Faded: an estimate

A running test with no estimated end yet fades out to the right. The Today line runs down the whole chart.

Under each experiment's name you'll see its dates, marked "(est.)" for an estimate or "(sched.)" for a scheduled date. Hover over a bar for more detail: its status, dates, length and why it ends where it does.

A warning icon on a bar means the experiment is running at the same time as another test it may interact with.

Where the estimates come from

• Running tests: the end is estimated from how quickly the test is collecting traffic, using the same projection as its Statistics tab. If a stop is scheduled, that date is used instead.
• Drafts: a draft appears once it's scheduled or queued in a pipeline. Its length comes from its run-length plan, so only drafts that have a saved plan are drawn with a length. See "Plan how long an experiment needs to run".
• Pipelines: experiments are grouped by pipeline and show their stage. A waiting stage is placed after the stage before it, following that stage's rule: it can start when the first test concludes, when all have concluded, or only when someone moves it on by hand.

Experiments that can't be placed

Some experiments can't be given honest dates yet, for example:

• a draft with no saved plan
• a pipeline that hasn't started or is paused
• a stage waiting behind a test with no estimated end

These are listed under could not be placed, below the timeline, each with the reason. Nothing is drawn at a date nobody knows.

Adjusting the name column

Drag the edge of the Experiment column header to make it wider for long names. You can also focus the edge and use the arrow keys. Double-click the edge to reset it. Your browser remembers the width.