Reference

Glossary

What the metrics, traffic surfaces, and engine tiers on this dashboard actually mean — described, not prescribed.

RegisterObservationalEach term defines what is measured, never what to do about the number.

  1. Metrics

    CTR
    Click-through rate — the share of impressions that became a click. On YouTube it depends on thumbnail, title, and which audience YouTube is showing the video to — rarely a single cause.
    Click rate
    The plain-English gloss for CTR — the share of impressions that became a click. Open Channel Stats labels charts and tables as “CTR”, YouTube’s own name for it.
    Impressions
    The count of times a thumbnail was shown to a viewer in YouTube’s UI — in the Browse feed, the Suggested column, search results, and similar surfaces. Impression counts come from YouTube’s reports, which reach back about thirty days from when a channel connects — the window is per day, so earlier days show views without impressions. Each channel’s own first covered day is named on its data page.
    Watch time
    The total minutes viewers spent watching a video, summed across all viewers and sessions.
    AVD (average view duration)
    The mean number of minutes viewers watched per view. Surfaced as “Avg view duration” (“Avg duration” in tight columns) in the dashboard.
    Retention
    The share of a video’s runtime viewers watched on average — avg view duration ÷ video length. Often correlates with topic-audience fit and pacing.

  2. Traffic surfaces

    Where YouTube placed the thumbnail before the click.

    Browse / Homepage
    A traffic source meaning the video was shown on a viewer’s YouTube homepage or subscription feed — often the first surface where a channel reaches non-subscribers beyond search.
    Suggested videos
    A traffic source meaning the video was shown next to or after another video the viewer was already watching. Typically the largest source for established channels.

  3. Video state

    Era
    A continuous span where a video’s title and thumbnail stay stable. When either changes, a new era begins. Used for before/after comparisons on the /changes page.
    Reporting cutoff
    The latest date with complete reporting data. YouTube’s reporting has a 2–3 day lag, so charts and detectors filter to dates on or before the cutoff — this keeps empty days from looking like zero days.
    Pre-publish stub
    A row in video_daily or traffic_daily whose date is earlier than the video’s publish date. YouTube’s Reporting API sometimes emits these zero-impression placeholders for the seven days before publish; Open Channel Stats flags them with pre_publish_stub = 1 and excludes them from every aggregate.

  4. Engine tiers

    The confidence and maturity labels used across the dashboard.

    Maturity stages

    A four-stage classification based on video count, days active, and total views. Each stage gates which detectors are allowed to surface insights, so very small samples don’t drive statistical claims.

    • Seedling— the smallest stage. Only encouragement and milestone signals fire.
    • Sprouting— at least 10 videos, 7 days active, 500 views.
    • Growing— at least 30 videos, 30 days active, 5,000 views.
    • Established— at least 100 videos, 90 days active, 50,000 views.
    Certainty tiers (Signal / Trend / Fact)

    How confident the insight engine is in a given observation. Certainty rises with the channel’s maturity stage and the magnitude of the underlying signal.

    • Signal— the most cautious tier, typical at the seedling stage.
    • Trend— medium confidence, sprouting stage or moderate signal.
    • Fact— the most confident tier, growing stage and above.
    Confidence tiers (noise / low / medium / high)

    A grade that applies to click-rate values specifically. Click-rate values computed on small impression samples are statistically noisy, so each value gets a tier based on the impression count behind it; the dashboard styles the noisy tiers differently so the noise is visible at a glance.

    • Noise— not enough impressions to trust the rate.
    • Low— some signal, large error bar.
    • Medium— the impression count supports a usable estimate.
    • High— the impression count is large enough that the rate is stable.

  5. The shared pool

    Where a reading labeled “derived from the shared pool” comes from.

    Channels like this one

    The channels that made the separate, deliberate choice to share their numbers into a shared pool and happen to broadly resemble this one — similar subscriber size, category, channel age, and video format. Once two or more of them line up, the dashboard can describe what a typical stretch looked like across them, always labeled derived from the shared pool and set beside this channel’s own numbers — this channel is one of the channels in that pool, not measured against an outside group. It reflects only the channels that are actually sharing, not every similar channel out there. When no comparable channels are sharing, nothing is shown — the line simply doesn’t appear rather than inventing a benchmark. The count shown in the details is a minimum, never an exact tally.