Report breakdowns
A breakdown splits an A/B test's results table into segments, such as mobile versus desktop or one row group per country. Each segment shows its own views, reach, goals, comparison metric, confidence and improvement, so you can see whether a variation wins everywhere or only for some visitors. Campaign reports break down by Device, Country, Channel or Referrer. Widget reports have their own breakdown chart with more dimensions.
How it works
- In the Results card of an A/B test report, open the report settings and set Break down by.
- The table regroups into one block per segment, each with the control and every variation.
- The statistics (confidence, interval, improvement) are recalculated within each segment.
- Segments are sorted by views. The top 10 are shown, and the rest are merged into Others.
The test progress bar and verdict are hidden while a breakdown is active. They only apply to the whole test.
Dimensions
Dimension | Segments by | Where the value comes from |
|---|---|---|
Device | Desktop, Mobile, Tablet | The visitor's device when they saw the campaign |
Country | Country name | The visitor's location when they saw the campaign, from their IP address |
Channel | Marketing channel, e.g. Paid Search, Email | The UTM parameters saved on the visitor's profile, classified into channels (see below) |
Referrer | Referring domain, e.g. google.com | The referrer of the page on which the visitor saw the campaign |
A segment shows (not set) when the value is unknown, for example when the visitor's location couldn't be determined, or when the page had no referrer.
How channels are classified
Channels follow Google Analytics 4's default channel groups and are computed from the UTM source, medium and campaign saved on the visitor's profile. Each UTM parameter keeps the latest value seen in a URL the visitor opened (see UTM parameters). The first matching rule wins:
Channel | Rule (simplified) |
|---|---|
Direct | No UTM source and no UTM medium |
Cross-network | Campaign contains cross-network |
Paid Shopping | Paid medium, plus either a shopping source (shopify, amazon, ebay, etsy, alibaba, walmart) or a campaign containing shop |
Paid Search | Paid medium and a search source: google, bing, yahoo, duckduckgo, baidu, yandex, ecosia, ask, aol, naver, sogou |
Paid Social | Paid medium and a social source: facebook, instagram, twitter, x.com, t.co, linkedin, pinterest, tiktok, snapchat, reddit, tumblr, quora, threads, fb.com, fb.me |
Paid Video | Paid medium and a video source: youtube, vimeo, dailymotion, twitch |
Display | Medium is display, banner, expandable, interstitial or cpm |
Paid Other | Any other paid medium |
Source or medium contains email | |
Organic Shopping | Shopping source, or a campaign containing shop |
Organic Social | Social source, or medium social, social-media, social_media or social media |
Organic Video | Video source, or medium video |
Organic Search | Search engine source, or medium organic |
Affiliates | Medium affiliate |
Referral | Medium referral |
SMS | Source or medium sms |
Audio | Medium audio |
Push | Medium contains push or notification |
Unassigned | Anything else |
A medium counts as paid when it contains cp (such as cpc or cpm), equals ppc or retargeting, or starts with paid. Sources match when they contain the listed text, so m.facebook.com counts as facebook. Values are compared in lowercase.
Channel uses UTM parameters only, not the referrer. A visitor from an organic Google result without UTM tags is classified as Direct. Channel reflects the visitor's saved UTM values at the time you view the report: if a visitor who first arrived from an email later comes back from a tagged paid ad, they move to the paid channel, including in earlier date ranges.
Widget report breakdowns
Widget reports have a separate Breakdown card that counts one widget event type by one dimension:
Option | Values / default |
|---|---|
Event | View, Click, Submit, Conversion (default), Completion |
Break down by | Page, Device (default), Country, Trigger, Visitor's referrer, Visitor's UTM source, medium, campaign, content or term, First visit page, and Product (for Click and Conversion, on widgets with products) |
Limit results | 5, 10 (default), 20 |
Widget breakdowns count events. They don't run statistical tests.
Behavior details and limitations
- Smaller segments, wider intervals. Each segment has a fraction of the traffic, so confidence intervals are wider and results take longer to become significant.
- Expect false positives across segments. With 10 segments at a 90% significance level, one segment showing a "significant" result by chance is normal. Use breakdowns to check consistency or to form a new hypothesis, not to pick winners segment by segment. See Statistics methodology.
- Referrer is page-level. It's the referrer of the page where the campaign was seen. When a campaign is seen on a second or later page, the referrer is your own domain, so your store's domain is often the largest segment. For where the visitor originally came from, use the Referrer report filter, which uses the visitor's first-touch referrer.
- Device and country can vary per visitor. They're taken from each exposure, so a visitor who saw the campaign on both their phone and laptop can appear in two segments.
- A/B tests only. Personalization reports don't offer Break down by. Use report filters such as Device or Country instead.
- Breakdowns combine with filters. Filters are applied first, then the remaining data is split into segments.
Examples
- Mobile versus desktop: break down a product-page layout test by Device. If the variation wins on mobile but loses on desktop, follow up with a personalization that shows it on mobile only.
- Market check for a price test: break down by Country to confirm the higher price doesn't hurt conversion in your second-largest market.
- Traffic quality: break down by Channel to see whether a landing-page variation helps paid social visitors, who often bounce more than email subscribers.