Sometimes a change is great on desktop and terrible on mobile, or vice versa. Device and country targeting let you test on a specific slice of your audience instead of averaging across everyone.
Device type
Three buckets: desktop, tablet, mobile. We classify based on viewport width — <768px = mobile, <1024px = tablet, the rest = desktop.
Useful patterns:
- Mobile-only experiments when you're testing thumb-friendly redesigns.
- Desktop-only for hover-driven UI patterns that don't apply on touch.
- Tablet excluded to keep the test sharp — tablets are often a small enough segment that they add noise without adding signal.
Country
Geo-IP based, computed at the edge before the tracker fires. Pick one or more country codes. Useful when:
- You want to test pricing, shipping, or copy that's specific to one market (e.g. SAR pricing for Saudi Arabia).
- You're rolling out a feature country-by-country.
- Your conversion rates differ enough by region that mixing them dilutes the test.
Browser
Less common but occasionally useful — when a CSS quirk only affects Safari, or you're testing a feature that requires a specific browser API.
Combining targeting types
Multiple targeting types stack with AND logic. URL contains /checkout AND device = mobile AND country = SA → only mobile checkout users from Saudi Arabia. The narrower you target, the more traffic you need before the result is significant.
Watch your sample size
Every targeting rule cuts your sample. Mobile + Saudi Arabia + a single product page might give you only 20 conversions a week. That's not enough to detect anything but a huge effect. Before adding a rule, check your traffic — if it's already small, broaden instead of narrow.
What's next
Once your targeting is set, pick a goal and start the experiment. If it's running but no data is coming in, double-check your targeting rules — overly tight rules are the most common cause of "ghost" experiments.