Audiences

An audience is a saved set of conditions that decides who sees a test. Build it once and any test can use it.

Every test asks who it runs on at step four. You can write the conditions there and move on, or pick an audience you saved earlier. Rollouts ask the same question in the same way.

A test with no audience runs on everybody. That is the right answer more often than people expect, because narrowing a test also slows it down. A test needs the same number of visitors either way, so half the traffic means about twice the wait.

#What you can target on

Eleven things, grouped the way the picker groups them.

Language matches on the first part. A browser reports en-GB and the list offers en, so picking English catches British and American visitors alike.

To target on something only your own site knows, such as a plan or a logged-in role, write it into a cookie and target the cookie on the test itself.

#Ready-made audiences

Twenty-three audiences are written for you, in six groups: Device, Browser, Operating system, Visitor, Traffic source and Time of visit. Picking one on the Audiences screen creates an audience of your own that you can then edit. Picking one inside a test copies its conditions straight onto that test.

They cover the ones people write by hand every time, including the awkward ones:

Start from one of these when it is close, then change a condition. It is quicker than writing the referrer list for search traffic yourself.

#How to set one up

1. Open Audiences in the sidebar and click New audience.
2. Type a Name, such as "Mobile shoppers in India".
3. Pick a ready-made, or click Add condition and choose a field, a comparison and a value.
4. Add more conditions, and choose whether to match All or Any of them.
5. Read the summary back, then save.

The screen writes each condition out as a sentence, so "Device is Mobile" reads as what it is before you save it.

#What each label means

A condition takes up to forty values, so one row can list forty countries rather than forty rows listing one each.

#Comparisons

The last two need no value, because the only question is whether anything arrived. Use is not set on Referrer to mean direct traffic.

Give a comparison more than one value and it renames itself, so "is" becomes "is one of" and "contains" becomes "contains any of". The sentence keeps reading as a sentence.

#Groups, and when you need one

One group covers most audiences. Reach for a second when you want two different kinds of visitor in the same test.

"Mobile visitors from Google" is one group, matching All: Device is Mobile, and Referrer contains google.

"Mobile visitors, or anyone from a paid campaign" is two groups, joined by Any group. Neither condition belongs with the other.

Two levels is the limit. Conditions combine inside a group, groups combine between themselves, and there is no third level.

#Good to know

• A test takes a copy of the conditions when you apply an audience. Editing the audience afterwards reaches the next test that applies it, never one already running. Your live tests cannot move underneath you.
• An audience a test came from cannot be deleted. The refusal names how many tests took their conditions from it. Point those elsewhere first.
• The Used by column counts those tests, and Who is in it on the same list reads the conditions back as a sentence.
• An audience with no conditions is refused. Nothing to narrow is not an audience, it is what a test does without one.
• Narrow with care. Conditions multiply, and "Mobile, in India, on Chrome, at the weekend" can be a slice too small to ever reach an answer.

#Related articles

• Your first A/B test: where an audience gets applied, at step four.
• Metrics: the number the test is judged on.
• Events: what your site reports about each visitor.