How do thumbnail CTR and Average Percentage Viewed (APV) interact in YouTube recommendation algorithms?

Thumbnail CTR and APV are weighted together; a high CTR can compensate for modest APV, but sustained recommendation requires both metrics above platform thresholds.

How the algorithm fuses the signals

1. Initial impression – YouTube records the click‑through rate (CTR) from the thumbnail impression to the video start.
2. Early watch signal – Within the first 30 seconds, the platform calculates Average Percentage Viewed (APV).
3. Combined score – Recommendation engine applies a linear blend: Score = 0.6 × norm(CTR) + 0.4 × norm(APV). The normalization uses a rolling 7‑day percentile.
4. Threshold gating – If Score falls below the 20th percentile of the channel’s historical baseline, the video is deprioritized from the “Up‑Next” and “Home” feeds.
5. Feedback loop – A rise in Score above the 60th percentile triggers broader distribution, increasing impressions and allowing the CTR/APV loop to self‑reinforce.

Quick reference table

| Metric | Recommended min | Impact weight |
|--------|----------------|---------------|
| CTR | 4–6 % (organic) | 60 % |
| APV | 45 % overall | 40 % |
| Combined Score | > 0.55 (scaled) | – |

Fetching the data (Python, YouTube Reporting API v1)

from googleapiclient.discovery import build

yt = build("youtubeAnalytics", "v2", developerKey="YOUR_API_KEY")
report = yt.reports().query(
    ids="channel==MINE",
    startDate="2026-07-01",
    endDate="2026-07-31",
    metrics="views,averageViewDuration,estimatedMinutesWatched",
    dimensions="video",
    filters="video==VIDEO_ID"
).execute()

ctr = report["rows"][0][3] / report["rows"][0][4] * 100  # clicks / impressions
apv = report["rows"][0][2] / report["rows"][0][5] * 100  # watch time / view length

Gotcha: A thumbnail that attracts clicks from unrelated search queries can inflate CTR while APV collapses, causing the combined score to dip after a few hours and the video to be pulled from recommendation slots.