The traffic sources report
The report lives in the Content tab of YouTube Studio under how viewers find your videos, with the full table in Advanced mode. For any period and any set of videos it lists the sources, the views each produced, and the watch time, average view duration and — for sources where a thumbnail is shown — the impressions and impressions click-through rate (CTRShare of impressions that result in a click.) that accompanied them. Most sources can be expanded: YouTube search opens to the search terms viewers typed, suggested videos opens to the videos that were being watched when yours was recommended, external opens to the referring sites and apps, and playlists opens to the playlists that carried the view.
Three properties of the report shape how it is read. It attributes a view to the surface the viewer was on when they started it, so a viewer who searched, watched, and then followed a suggestion produced one search view and one suggested view. It reports views rather than people, so a loyal viewer who arrives from the subscriptions feed every week is counted every week. And it reports only what YouTube can see: a viewer who arrived through a link in a messaging app with no referrer lands in a direct-or-unknown bucket, and YouTube does not document everything that falls into it.
The report is one of three that a weekly review opens together; the YouTube Analytics hub describes the routine, and the retention curves guide covers the report that explains what viewers did once they arrived.
Each source and what drives it
| Source | What it is | What drives it | What it rewards |
|---|---|---|---|
| YouTube search | Views that began with a query typed into YouTube | Titles, descriptions and chapters that match how people search; the video answering the query | Precise metadata; evergreen topics with steady demand |
| Suggested videos | Views from the recommendations beside or after another video | Thumbnails and titles that win against neighbours; viewers who tend to watch yours after a related video | Packaging; closeness to videos your audience already watches; session continuation |
| Browse features | Views from the home feed, subscriptions feed, watch later and similar surfaces | The relationship with existing viewers; YouTube's prediction of what each person will open next | Consistency, cadence and retention among returning viewers |
| Shorts feed | Views that began in the vertical Shorts feed | The first second; the loop; the viewed-against-swiped split | Hooks and loops rather than thumbnails |
| Channel pages | Views from your own channel page or another channel's | Channel page layout, featured sections, playlists on the page | A browsable catalogue |
| Playlists and playlist pages | Views from inside a playlist, yours or someone else's | Series design; being added to other people's playlists | Sequencing videos that belong together |
| Notifications | Views from push and email notifications about new uploads | Subscribers who turned notifications on; upload timing | A loyal, alerted audience |
| External | Views from websites and apps outside YouTube, including embeds, social platforms and Google Search | Links and embeds elsewhere; video results in web search | Distribution work off the platform |
| Direct or unknown | Views with no identifiable source: typed URLs, some apps, some messaging links | Sharing that YouTube cannot trace | Nothing directly; a large bucket is a prompt to tag what you can |
| End screens and cards | Views from end-screen elements and cards on other videos | Which videos you point viewers to, and where in the video | Deliberate paths between videos |
| YouTube advertising | Views from video ads bought in Google Ads | Campaign budgets and targeting | Spend rather than organic work |
| Hashtag pages, sound pages and other features | Smaller sources YouTube reports separately or groups as other YouTube features: hashtag pages, sound pages, remixes, live redirects | Hashtags, sounds, participation in formats | Minor; varies by format |
YouTube adds, renames and regroups sources as the product changes — Shorts-specific sources, podcast surfaces and live redirects have all appeared in recent years — so the help centre in the sources is the current list. The grouping above is stable enough to reason with, and the first four rows carry most of the views on most channels.
Reading the mix per video and per channel
For a single video the mix changes with age, and the reading changes with it. In the first days a video is shown mostly to the channel's own audience, so browse and notifications lead; if the video holds those viewers, suggested begins to grow as YouTube offers it beside related videos; search arrives last and, for videos that answer a durable query, keeps arriving for years. A video whose mix stays dominated by browse after its first weeks was not picked up by suggested; a video whose search share keeps growing is an asset worth refreshing. Open the per-video mix at a few ages and compare it with the channel's other videos of the same type at the same age, which is the only fair comparison.
For the channel, the mix is a description of the business. A channel built on search is a library: steady, slow to grow, resilient to a single weak upload, and dependent on search engine optimisation (SEOImproving visibility in unpaid search results.) in the YouTube sense that the video SEO hub describes. A channel built on suggested and browse is a broadcaster: faster growth, bigger swings, dependent on retention and cadence. A channel built on the Shorts feed reaches many people briefly and must build its own bridge to long-form. None is better; each has a different weak point, and the mix tells you which weak point is yours. Watch the mix as a trend rather than a snapshot, and treat a shift in it as leading: suggested and browse usually move before total views do.
Impressions and click-through are reported per source and should be read per source, because the same thumbnail earns very different rates on search, where the viewer has a query, and on home, where they are browsing. A rate blended across sources moves whenever the mix moves, which is one reason never to make it a KPIA metric tied directly to an objective.. The thumbnails and click-through guide covers reading it properly.
External traffic and tracking it
External is the source for views that began off the platform: an embed on a website, a link from a social post, a video result in Google Search, a share in a newsletter. The report expands it to the referring domain, which is enough to see which placements send viewers and which do not, and Google Search appears there as its own line when a video is shown in web results. What the report does not do is read campaign parameters. YouTube Analytics has no dimension for UTM tags (UTMURL tags identifying traffic source.) on links pointing to a video; a tagged link still works, but the tag is not reported, so the referring domain is the finest grain available on the way in.
Tracking in the other direction — from YouTube to your own site — is where tagging matters. Links in descriptions, pinned comments, channel page links and, where the channel is eligible, end-screen links to a website should carry UTM parameters so that web analytics can separate YouTube from everything else and separate one video from another. The UTM and naming conventions guide sets out a convention; the practical rule for YouTube is one source value for the platform, a medium that says the link was organic rather than paid, a campaign that names the video or series, and a content value for the placement, so that a description link and a pinned comment can be told apart. Which videos should carry those links at all, and what the page they lead to should say, is covered in the guide to sending YouTube viewers to your website.
Expect disagreement between the two sides. Untagged clicks from YouTube arrive in web analytics as referrals from the YouTube domain on the web and often as direct trafficVisits recorded as direct because they arrived with no campaign parameters and no referrer, whether typed, bookmarked or from a lost source. from the apps, where no referrer is passed, and the counts on either side are measured differently. The guide to why platforms and analytics disagree explains the general causes; for YouTube the fix is almost always to tag every link you control and accept that the rest is approximate.
What each source tells you to do next
The mix is diagnostic only if each source maps to an action. The mapping below is the one most channels end up with.
- Search is low on videos that answer questions. The metadata is not matching the query. Rewrite the title to carry the words people type, restructure the description, add chapters; the titles, descriptions and chapters guide is the method. Check the search terms the report lists against the ones you expected.
- Search is high and subscribers are flat. The video answers and ends. Add a reason to return: a related video pointed to from the end screen, a series the query leads into, a channel-level promise stated early. The tutorial on planning a YouTube video series builds that series, with its playlist and end screens read as sources at the review.
- Suggested is low on videos with good retention. The packaging loses against neighbours. Test the thumbnail; look at which videos the suggested report says yours appears beside, and make the title and thumbnail compete with those.
- Suggested is high and retention is poor. YouTube is offering the video widely and viewers are not staying. The intro or the structure is the issue; open the retention curve filtered to suggested.
- Browse is falling. The existing audience is engaging less. Look at cadence, at the last several uploads' retention among returning viewers, and at whether the content promise drifted.
- Shorts feed is large and nothing else is. Build the bridge: related long-form videos linked from the Short, a channel page that leads to them, Shorts that act as trailers for the catalogue.
- External is a large share. Find out which placements, keep them fed, and tag everything that leads back out.
- Direct or unknown is a large share. Something is being shared where you cannot see it. Ask the audience, check community posts and messaging groups, and add tagged links wherever you do have control.
A home-repair channel reads two videos' mixes
A home-repair channel compares two videos at thirty days. The first, a guide to fixing a dripping tap, has 96,000 views: 61% from YouTube search, 18% from suggested, 9% from browse, 7% from external (mostly Google Search) and the rest scattered. Its retention is strong, its search terms list is dominated by variations of the title, and it gained 140 subscribers. The second, a bathroom renovation on a £900 budget, has 310,000 views: 58% from suggested, 27% from browse, 6% from search and 5% from the Shorts feed via a trailer. It gained 2,900 subscribers. The channel reads the first as a library asset — refresh its description and chapters yearly, point its end screen at the renovation series — and the second as a broadcaster asset whose job is to convert attention into subscribers, which it did. The action is to make more renovation-style videos for growth and to keep publishing one search-led repair video a week so the catalogue keeps earning.
When channel views fall and the source mix is the first suspect, the diagnosis for YouTube views that dropped suddenly compares each surface against a typical period and works through the causes in order.
Common mistakes
- Reading the mix as a snapshot. A video's sources change with age; compare at the same age and watch the trend.
- Blending click-through across sources. Search and home produce different rates for the same thumbnail; read per source.
- Expecting UTM tags to appear in YouTube Analytics. They do not; the referring domain is the finest grain on the way in.
- Leaving outbound links untagged. Every untagged description link makes YouTube invisible in web analytics.
- Treating direct or unknown as noise. A large bucket means sharing is happening where you cannot see it.
- Optimising every video for search. Broadcaster videos earn through suggested and browse; forcing search metadata onto them wastes the title.
- Fixing the thumbnail when suggested is high and retention is low. The video is being offered; it is not being watched.
- Ignoring the suggested report's video list. It names the videos yours sits beside, which is the clearest picture of the competition.
