YouTube SEO: review your video and its evidence
Fictional manual review examples. No live channel connection, metadata publishing, expert-service verification or ranking prediction.
| Inspection step | Example observation | Evidence to record | Action if it differs |
|---|---|---|---|
| Viewer promise | Vague title: My camera tips!!! | What practical question does the video answer? | Propose an accurate title only if the lesson delivers it |
| Title and thumbnail | A shutter-speed lesson with an unrelated image | Check whether both describe the same subject | Record one coherent proposal, not a guaranteed click uplift |
| CTR denominator | 10000 impressions; 500 attributed views; 200 external views | Use 500 ÷ 10000, not 700 ÷ 10000 | 5% is a fictional ratio, not a universal target |
| Retention observation | A dip near 00:12 during the introduction | Inspect the moment; skipping and leaving are different possibilities | Write the uncertainty and one hypothesis before changing the opening |
| Next review | Later snapshot: 20000 impressions and 800 attributed views | 4% with more views; compare period, source and audience | Record changes and context without asserting causation |
How to use the matrix
Consider a fictional camera tutorial called “My camera tips!!!”. A clearer proposal is “Reduce motion blur indoors: shutter-speed setup”. The lesson must actually show that task, and the thumbnail should communicate the same subject without promising a result it never demonstrates. Our invented report contains 10000 registered thumbnail impressions and 500 views from those impressions: CTR is 5%. An additional 200 external views must not be added to that numerator.
In authorized YouTube Studio, open the relevant video’s Analytics and record the reporting period and traffic-source filter. Registered thumbnail impressions do not include every exposure outside YouTube. Compare like scopes and allow for reporting delays. A later fictional snapshot of 20000 impressions and 800 attributable views has 4% CTR but more views from impressions. That change alone proves neither a worse title nor successful optimization; audience and source mix may differ.
Review checklist
- Define the viewer’s question and actual demonstration
- Review title and thumbnail as one promise
- Record period, source filter and matching counts
- Describe retention observations and alternative explanations
- Choose one justified change and a later review date
Review the audience-retention report at the video level. A dip around 00:12 could mean viewers skipped or left; inspect what happens there and record a hypothesis, not a diagnosis. A spike can involve rewatching and may deserve investigation rather than automatic praise. For our tutorial, check whether a long opening delays the promised demonstration. Record one proposed improvement, why it addresses the viewer’s task, who will make it and when to review the evidence again.