Audience targeting
Run each experiment only for the visitors you choose, and control how much of that traffic enters the test.
Why target an audience
By default an experiment is eligible for every visitor to your storefront. Audience targeting narrows that pool so the experiment only runs for visitors who match the rules you set. This is useful when a hypothesis only applies to part of your traffic — a mobile-only layout change, a landing page for one paid campaign, a message for first-time visitors, or an offer for shoppers in a specific country.
Targeting also keeps your results clean. When an experiment runs only for the audience it was designed for, the conversion and revenue numbers reflect that audience instead of being diluted by visitors the change was never meant to reach.
Targeting decides who is eligible to enter an experiment. Among eligible visitors, each person is still assigned to one variant at random and stays on that variant for the rest of their visit.
Rule types
You can add one or more targeting rules to an experiment. Each rule checks a property of the incoming visit and either includes or excludes the visitor.
| Rule | Matches on | Options |
|---|---|---|
| Device | The device category of the visit. | Mobile, tablet, or desktop. |
| New vs returning | Whether the visitor has been seen before. | New visitors or returning visitors. |
| Country | The visitor's country. | Is in, or is not in, a list of countries. |
| UTM parameter | A UTM value on the landing URL (for example utm_campaign). |
Equals, or contains, a value you specify. |
| Referrer | The referring URL that sent the visitor to your store. | Contains, or equals, a value you specify. |
| Cookie | A cookie present in the visitor's browser. | Exists, or equals a value you specify. |
Device
Restrict the experiment to mobile, tablet, or desktop visits. Use this when a variant is designed for one form factor — for example a compact mobile hero, or a desktop-only comparison table.
New vs returning
Target visitors who are new to your store or ones who have visited before. First-time visitors and returning shoppers often respond differently to the same change, so this is a common way to keep a test focused.
Country
Include or exclude visitors by country. Choose "is in" to run the test only in the countries you list, or "is not in" to run everywhere except those countries — handy for regional pricing, shipping messages, or language.
UTM parameter
Match a UTM value on the URL the visitor arrives with, such as utm_source, utm_medium, or utm_campaign. Use "equals" for an exact value or "contains" for a partial match. This is the usual way to run an experiment only for traffic from a specific ad or email campaign.
Referrer
Match the referring URL that sent the visitor to your store. Use "contains" to match any referrer that includes a string (for example a search engine or a partner domain) or "equals" for an exact referrer.
Cookie
Match on a cookie in the visitor's browser. Choose "exists" to target anyone who has the cookie set, or "equals" to require a specific value. This lets you reuse a segment your store or another tool already marks with a cookie.
UTM, referrer, and cookie rules read values that are present when the visitor lands and is assigned. Set them to match how your campaigns actually tag traffic.
How multiple rules combine
When an experiment has more than one rule, all of them must match for a visitor to be eligible — the rules combine with AND. For example, a rule set of device is mobile and country is in India only enters visitors who are both on mobile and in India.
To reach a broader audience with a single value type, use the options built into a rule rather than adding conflicting rules. Country accepts a list of countries in one rule, and "is not in" lets you express an exclusion without a second rule.
Because rules are combined with AND, adding many rules narrows the audience quickly. If a running test is getting little traffic, check whether its rules are more restrictive than you intended.
Traffic allocation
Separately from targeting, you set what share of eligible visitors actually enter the experiment. A visitor who matches all targeting rules is first checked against the traffic allocation percentage; only the allocated share is assigned to a variant, and the rest see your normal storefront and are not counted in the test.
- 100% — every eligible visitor enters the experiment.
- Lower than 100% — a random portion of eligible visitors enters, and the remainder are held out. Use this to limit exposure while you gain confidence in a change.
Within the visitors who do enter, traffic is split between the variants. Allocation controls how many people join the test; the variant split controls how they are divided once inside.
Effect on results and segments
Targeting shapes which visitors your results are based on. Because only matching, allocated visitors are assigned, the conversion rate, revenue, and probability-to-be-best on the results page reflect exactly that audience.
You do not have to target narrowly to analyze narrowly. Even when an experiment runs for a broad audience, results include segment breakdowns — by device, new vs returning, country, UTM, referrer, and cookie — so you can see how each segment responded after the fact. A common pattern is to run broad, read the segment breakdowns to find where a variant wins, then launch a follow-up experiment targeted at that segment.
The segments available in results mirror the targeting rule types, so an audience you can target is also an audience you can analyze.
Related pages
- Quickstart — set up and launch your first experiment.
- Results and statistics — read outcomes and segment breakdowns.
- Campaigns — group related experiments to report together.
- Revenue attribution — how sales are credited to variants.