Unique visitors is the number of distinct people (in practice, distinct browsers or devices) that visited a site during a date range, with each one counted once no matter how many times they came back.
Sessions count visits, so one unique visitor who returns three times adds one visitor and three sessions.
Google Analytics 4 calls this metric "Total users" and defines it as "the number of unique users who triggered any event in the specified date range." GA4 tells users apart with the _ga cookie, which is "used to distinguish users" and lasts 2 years by default.
So a person who clears cookies or switches from phone to laptop shows up as two visitors.
Unique visitors vs sessions
A session is one visit. GA4 says a session ends after 30 minutes of inactivity by default, so coming back after lunch starts a new session but not a new visitor.
| Unique visitors | Sessions | |
|---|---|---|
| What it counts | Distinct browsers or devices | Separate visits |
| Returning three times in a week | 1 | 3 |
| Always the larger number? | No, smaller or equal | Yes, larger or equal |
| Typical A/B test use | Unit of randomization and denominator for conversion rate | Denominator for "per session" metrics |
The choice changes the number you report. Say a store has 10,000 unique visitors, 14,000 sessions and 300 orders in a week.
Conversion rate is 300 / 10,000 = 3.0% per visitor, but 300 / 14,000 = 2.14% per session.
Both are correct; they answer different questions. Always state which one a conversion rate uses before comparing it with a benchmark or another test.
How unique visitors show up in A/B testing
Most testing tools assign the visitor, not the session, to a variation and keep them there on every return visit.
That is why sample size, baseline conversion rate and lift are usually calculated per unique visitor.
If you analyze per session instead, a few heavy returners count many times, and the observations stop being independent.
Visitor counts per arm are also your first quality check.
With a planned 50/50 split, arms with very different visitor counts point to a sample ratio mismatch, which usually means a bug in assignment or tracking.
Anatta's tabbed buy box test shows the other side: it ran an intentional 85/15 split, and its 34,730 and 6,144 visitors matched that plan.
Examples from the database
Real tests often mix per-visitor and per-session metrics in one report, so read the unit next to each number:
- Blend Commerce's visible mobile search bar test for Aeromotive reported +88% conversion rate, +109% revenue per user and +30% add to cart per session. Those three figures use different denominators and can't be added together or compared directly.
- GrowthRock's size guide test counted 20,000+ sessions per variation and reported +21% conversion rate and +46% revenue per session, with under 200 conversions per variation. GrowthRock itself says to treat it with caution.
- Blend Commerce's subcategory pills test for Wilson's Pet Food was paused after 5,342 visitors at an 83.1% probability to beat control. Its +22.86% revenue per visitor is a directional result, not a settled one.
Blend CommerceBlend Commerce
GrowthRockWhere unique visitor counts go wrong
- Cross-device use: one person on phone and laptop counts as two visitors unless you identify logged-in users.
- Cleared or blocked cookies: returning visitors look new, which inflates unique visitors and lowers per-visitor conversion rate.
- Different tools, different counts: your testing tool and your analytics tool rarely agree exactly, because they set their own cookies and filter bots differently. Compare trends within one tool, not raw numbers across tools.
Related terms
See sample size for how many unique visitors a test needs, browse the full glossary, or search all test results to see how other teams report their visitor counts.