Analytics breakdowns

breakdowns let you split your a/b experiment and personalization campaign results by a visitor dimension so you can see how each variation performs across segments for example, break down by device to compare conversion rates on mobile versus desktop, or by channel to see which traffic sources drive the most value how to use breakdowns open a campaign report and look for the break down by dropdown in the report settings select a dimension to split the results table into one section per segment value (e g , one section for mobile, one for desktop) select none to return to the default single table view when a breakdown is active, the experiment progress section is hidden because per segment sample sizes are smaller and the overall progress indicator would be misleading breakdowns can be combined with filters for example, you can filter to a specific date range and then break down by country to see how each country performed during that period available dimensions device — splits results by the device type recorded on the campaign event desktop, mobile, or tablet country — splits results by the visitor's country, resolved from their ip address when the event was recorded country codes are displayed as full names (e g , "united states" instead of "us") channel — classifies each visitor's traffic source into a marketing channel based on their utm parameters (source, medium, and campaign) see the channel classification section below for details referrer — splits results by the referring domain extracted from the campaign event's referrer url only the hostname is shown (e g , "google com" rather than the full url) reading the breakdown table when a dimension is selected, the report table splits into labeled sections — one for each segment value (e g , one section for "desktop", one for "mobile") each section shows its own variation rows with independent confidence, improvement, and confidence interval calculations a summary row at the top aggregates the results across all segments so you can still see the overall picture segment limits to keep the table readable, breakdowns are limited to the top 10 segments by total views if there are more than 10 segments, the remaining segments are aggregated into an others bucket this is most relevant for the country and channel dimensions, which can produce many distinct values channel classification the channel dimension classifies visitors into marketing channels following a simplified version of ga4's default channel grouping rules classification is based on the visitor's utm source, utm medium, and utm campaign — the values stored on the visitor record from their first arrival channels are evaluated in priority order the first matching rule wins channel rule direct source is empty, "(direct)", or "direct" and medium is empty, "(none)", or "(not set)" cross network campaign name contains "cross network" paid shopping medium indicates paid and (source is a shopping platform or campaign contains "shop") paid search medium indicates paid and source is a search engine paid social medium indicates paid and source is a social network paid video medium indicates paid and source is a video platform display medium is one of display, banner, expandable, interstitial, cpm paid other medium indicates paid (catch all for paid traffic not matching the above) organic shopping source is a shopping platform or campaign contains "shop" organic social source is a social network or medium is one of social, social media, social media organic video source is a video platform or medium is "video" organic search source is a search engine or medium is "organic" email source or medium contains "email" affiliates medium is "affiliate" referral medium is "referral" sms source or medium is "sms" audio medium is "audio" push medium contains "push" or "notification" unassigned everything else paid medium detection a medium is considered paid if it contains "cp" (covers cpc, cpv, cpm, cpa), equals "ppc" or "retargeting", or starts with "paid" recognized sources search engines google, bing, yahoo, duckduckgo, baidu, yandex, ecosia, ask, aol, naver, sogou social networks facebook, instagram, twitter/x, linkedin, pinterest, tiktok, snapchat, reddit, tumblr, quora, threads video platforms youtube, vimeo, dailymotion, twitch shopping platforms shopify, amazon, ebay, etsy, alibaba, walmart source matching is case insensitive and uses substring matching (e g , "google" matches "google com", "google/organic", etc ) faq do breakdowns affect statistical calculations? yes confidence, improvement, and confidence intervals are computed per segment each segment's control is compared only to the same segment's variations, not to the overall control can i use breakdowns with product breakdowns? no when a dimension is selected, per product breakdowns are not shown remove the breakdown to see product level data why is the experiment progress hidden during breakdowns? the progress indicator and winner detection are based on overall traffic and significance thresholds per segment sample sizes are typically smaller and may not meet the required thresholds individually to avoid confusion, progress is hidden when a breakdown is active why do i see "(not set)" as a segment? this appears when visitors do not have a value for the selected dimension for example, "(not set)" in a channel breakdown means the visitor had no utm parameters and did not match the direct channel rules, or the referrer was empty for a referrer breakdown do filters and breakdowns interact? yes filters are applied first, then the filtered data is split by the selected dimension for example, if you filter to "device = mobile" and break down by country, you will see country segments for mobile traffic only why does the channel breakdown show "unassigned"? visitors whose utm parameters do not match any of the recognized channel rules are grouped into the unassigned channel this typically happens when custom or non standard utm values are used is the "others" bucket a single segment? yes when more than 10 segments exist, segments ranked 11th and beyond by total views are merged into a single others row the others row's statistics are the sum of all the merged segments