Metrics
| You want to know | Go to |
|---|---|
| What did I earn | Overview, then Earnings |
| Who is on my site right now | Live |
| Which titles are performing | Content |
| Where my visitors came from | Audience |
| Which countries watch, day by day | Countries |
| Whether playback is any good | Video |
| Which pages people land on and leave from | Reports, Pages |
| Whether people come back | Reports, Retention |
| What one visitor did | Audience, then click a visitor |
Click any row in any table to filter every screen by it. Clear it in the top bar. Every screen except Live follows the date range and filters.
Audience
Section titled “Audience”| Metric | Definition |
|---|---|
| Pageviews | Every page load and every route change. One visitor reloading a page five times is five pageviews |
| Sessions | One visitor’s activity in one sitting. A session ends after 30 minutes with no events, and a later visit starts a new one |
| Visits | The same number as Sessions, under the label the visitors table uses |
| Visitors | Distinct browsers in the range, by the _ws_uid token, kept 365 days. Same person on a phone and a laptop is two |
| Visit length | Time between the first and last event of a session. A session with one pageview has no measurable length and counts as zero |
| Entry page | The first page of a session |
| Exit page | The last page of a session |
Referrer, campaign, country, and device are session-level facts. That is why one visitor can appear under two different sources over a month.
| Metric | Definition |
|---|---|
| Plays | Playback started. One per watch, not one per visitor. Several plays can sit in one session |
| Validated play | The stricter count that earns. How it is measured depends on your pixel’s site type. See below |
| Completes | Reached the end of the video |
| Completion | Completes divided by plays |
| Avg watched | Seconds actually watched per play. Paused and backgrounded time does not count |
| Audience retention | Share of plays still watching at each point through the video |
| Rebuffer ratio | Stall time as a share of watch time. Under 1% is healthy |
| Time to first frame | Median time from play to the first frame appearing |
| Playback errors | Share of plays that raised an error |
| Delivered in HD | Share of plays served at 720p or above |
Watch time and completion settle after the viewing session ends. Play counts appear immediately.
Rebuffer ratio, time to first frame, playback errors, and HD share describe your delivery, not your content. High completion with a bad rebuffer ratio means people are putting up with your player.
The shape of the retention curve tells you more than any single number:
| Shape | Usually means |
|---|---|
| A cliff in the first few seconds | Loading or autoplay, not content. Check time to first frame |
| A steady slope | Normal |
| A step at the same point every play | An ad break, a slow section, or a bad edit |
Live streams (type: 'sport') have no fixed duration, so completion, retention, and the
25/50/75/100% milestones are skipped for them.
What counts as a validated play
Section titled “What counts as a validated play”On a video site type, read from the player:
- Playback started, and 30 seconds watched or 20% of the runtime, whichever comes first.
- At most 3 per visitor, per title, per day.
On a dwell site type there is no player to read, so it is measured from time on the page. Either of these clears the bar:
- 5 minutes of engaged time on the page, or
- 5 minutes between the first and last event of a session, with at least 3 events in it. The event minimum is what stops an idle tab counting.
At most 1 per visitor, per page, per day, and the page must be one vidstats can identify. The 5 minutes is the default; the vidstats team can set it per account, so check with support if your numbers imply a different figure.
Dwell is the default for a new pixel, which is why a site with a player you control should switch its site type: 5 minutes on the page is a much higher bar than 30 seconds of playback.
Channels
Section titled “Channels”A session is sorted into exactly one channel, decided in this order:
- Direct, no referrer and no campaign tags
- Paid,
utm_mediumiscpc,ppc,paid,display, or similar - AI, the referrer is an AI assistant
- Search, the referrer is a search engine
- Email,
utm_mediumisemailornewsletter - Social, the referrer is a social network, or
utm_mediumissocial - Referral, any other site linked to you
Order is why a paid link from Google counts as Paid rather than Search, and why
gemini.google.com counts as AI rather than Search. A campaign tag beats the referrer, which
is the reason to tag your links.
Direct is usually bigger than expected: links from apps, messengers, PDFs, and any site that strips its referrer all land there.
Retention
Section titled “Retention”| Metric | Definition |
|---|---|
| Cohort retention | Of the visitors first seen in a given week, the share that came back in each later week |
| Visitor lifecycle | New, returning, and lapsed visitors over time |
| Stickiness | Daily visitors divided by monthly visitors. Higher means people come back more often |
These need weeks of data before they mean anything. A new pixel shows empty cohorts, which is correct rather than broken.
Earnings
Section titled “Earnings”| Metric | Definition |
|---|---|
| Latest statement | Your last settled weekly statement. Not tied to the date range |
| Revenue by site | Earnings split across your pixels |
| Next payout | The date and amount of your next scheduled payment. Runs weekly. See Get paid |
Why totals do not add up across screens
Section titled “Why totals do not add up across screens”Pageviews, sessions, visitors, and plays run on separate tracks, so they are not subsets of each other. 100 pageviews, 40 sessions, 25 visitors, and 60 plays is consistent.
Plays and validated plays differ for the same reason: Plays counts every start, validated plays applies the thresholds above. Your Plays number is normally the higher one.
Earnings ignore the date range entirely, because they settle weekly. See Get paid.