Statistics¶

The Statistics page in the dashboard (per project) shows how your videos are watched. It is organized into headline metric tiles, trend charts, and, when you focus on a single video, a second-by-second audience-retention graph. A caption at the top of the page shows the active date range, so it is clear what period the figures cover.

Headline metrics¶

At the top of the page, five tiles summarize the selected period. Each tile shows the value, a small trend sparkline, and a change badge comparing it to the previous equal-length period (for example, the previous 30 days). A green ▲ is an increase, a red ▼ a decrease; the badge is hidden when there is no prior data to compare against.

Metric	What it means
Plays	Number of play sessions (a viewer who pauses and resumes in one session counts once).
Play rate	Plays ÷ loads: the share of times the player loaded that turned into a play.
Watch time	Total time watched across all viewers, shown in its natural unit (hours / minutes).
Completion	Share of plays that reached the end of the video.
Unique viewers	Distinct viewers in the period.

How viewers are counted

Each play session is tagged with a viewer id. Returning viewers are recognized across sessions, and Unique viewers counts them once. Sessions without a viewer id are not included in that count (but still count toward Plays and Watch time).

Plays over time¶

The Plays over time chart breaks Views, Plays, and Watch time down per day, so you can see when your videos got attention across the selected period.

Completion funnel¶

The Completion funnel shows how far viewers get, as a share of plays: Played → Watched 25% → 50% → 75% → Completed. Each bar is the percentage of plays that reached that checkpoint, so you can spot where viewers drop off. The checkpoints are percentage-based per play, so the funnel is meaningful both for a single video and across an entire project (videos of different lengths still line up on the same scale).

Audience retention (per video)¶

Select a single video (from the Top videos table or the Video filter) to reveal the Audience retention graph. It plots, for each moment of the video, the share of plays still watching:

Watched: the percentage of plays that reached that moment.
Rewatched: replays layered on top, so the moments viewers went back to and watched more than once stand out.

The dashed line marks the average engagement across the video. A curve that drops sharply early points to an intro that loses viewers; a bump in Rewatched marks a moment people replay. The graph has several aids for reading it at a glance:

Hover preview: hovering shows a card with the video frame at that moment (from the scrub-preview storyboard), the timestamp, the Watched/Rewatched percentages, and the current chapter, so you can see what is on screen where viewers drop off or replay.
Key-moment callouts: the most-replayed moment (★) and the single biggest drop-off (▼) are marked automatically.
Engagement heatmap: a warmth strip under the graph, aligned to the timeline, colors each moment from green (high retention) to red (low), so you can tell at a glance which parts hold attention.
Chapter markers: if the video has a chapter track, each chapter start is marked on the timeline, so you can read retention against your own sections.
Click to seek: click anywhere on the graph to open the video at that exact moment in a new tab, to watch what's happening where viewers drop off or replay.

The storyboard hover preview appears only when the video has scrub-preview thumbnails (generated for completed videos), and chapter markers only when it has a chapter track.

Playing time by rendition (per video)¶

With a single video selected, the Playing time by rendition card shows how long viewers actually spent on each rung of that video's quality ladder. Time is counted only while the video is playing, so pauses, buffering and seeking do not inflate a rung.

The card has two halves. A stacked chart shows the mix per day over the selected period, one color per rung, which is where a change to your encoding settings shows up as the colors shifting. Below it, each rung is listed largest first with its share of the total playing time and the time itself. The bars are shares of the total, so they add up to 100%.

It answers two questions the ladder itself cannot. Which rungs are worth encoding: a top rung that collects almost no playing time is bytes you are storing and paying to deliver for nobody. And whether your viewers can reach the quality you publish: a video whose watch time sits mostly on the lower rungs is a video most people are seeing at a lower quality than you uploaded, usually because of their network.

Sessions that never reported a rung are not counted here. That is mainly iOS, where the operating system plays HLS itself and does not tell the player which rendition it picked, so those seconds appear in watch time but on no rung.

Playback errors¶

The Playback errors card lists the player errors your viewers hit in the selected period — one row per error code, with a sample message, how many views it affected, and what share of all views that is. It respects the page's date / video / country filters, and a healthy period simply shows a green check.

Each row links straight into the per-view drill-down filtered to that exact error code, so "what is error 3 and who does it hit?" is one click: the list shows every affected session, and opening one shows the device, browser, network, and playback story around the failure.

Individual views¶

Statistics → Views in the sidebar (or the Views button on the Statistics page, which carries over its active filters) opens the per-view drill-down: every playback session in the period as its own row, newest first — when it started, the video, the viewer, how much was watched and how far it got, startup time, and any stalls or errors. It shares the Statistics page's date / video / country filters and adds two of its own:

Viewer search — paste a viewer id to see everything that viewer watched. Viewer ids also appear as links throughout the page, so one click answers "what else did this person watch?". Anonymous viewers (no id supplied) are recognizable by their generated id.
Errors only — a one-click filter to just the sessions that hit a playback error, for support and debugging workflows.

Click any row for the view detail page — the full story of a single playback:

Quality-of-experience scores for this one view (viewer experience, playback success, startup time, smoothness, video quality), each graded Excellent / Good / Fair / Poor on the industry-standard 0-100 scales.
A playback timeline: which milestones the view reached (loaded → started → 25% → 50% → 75% → finished) and when, plus a watched heat strip showing exactly which parts of the video were watched — replayed moments show darker.
Delivery quality: startup broken down into page load / player load / first frame, rebuffering totals, every playback error with its code and message, upscaling and dropped-frame percentages, and which renditions (quality levels) the session actually played.
Viewer context: location, device, browser and OS, connection type, CDN, the page the video was embedded on, and the player/SDK version.

Sessions of deleted videos remain visible (marked as such), so history survives library cleanup. This drill-down is ideal when a customer writes in with "the video didn't play for me" — search their viewer id, open the session, and the error, device, and network story is all on one page.

Filters¶

Use the Filters panel to change what you see:

Date range defaults to the last 30 days; pick any range up to one year.
Video focuses the whole page on a single video (and shows its retention graph).
Country filters the metrics to viewers from one country (click a row in Top countries).

Clearing the Video filter or pressing Reset returns to the project-wide view.

Programmatic access¶

The same numbers are available through the Statistics API.

August 31, 2026