YouTube thumbnail testing spreadsheet
Decide what the thumbnail comparison should teach you, then record the exact versions and observations. Use the metric definitions and comparison method actually available to your account. A sequential change can be affected by traffic mix, timing and other changes to the video.
Your working document
Spreadsheet columns
| Column | Record | Purpose |
|---|---|---|
| Video and variation | Exact video reference and thumbnail file version | Identify what was compared |
| Window | Start, end and time zone | Keep the period clear |
| Metric and source | Actual metric name and extraction source | Avoid mixing incompatible definitions |
| Numerator and denominator | Both underlying values where available | Interpret ratios with sample size visible |
| Other changes | Title, traffic mix or distribution changes | Record alternative explanations |
| Decision | Chosen version, confidence and next check | Carry the finding into production |
Use the downloadable sheet
The CSV download contains two rows for your observations and two clearly labelled fictional examples. Open it in your spreadsheet application and keep the record_type column visible. Remove the illustrative rows from a client report; they are not account data.
Interpret a worked comparison
A fictional illustration: 40 observed clicks from 1,000 qualifying impressions is 4%; 60 from 2,000 is 3%. The second variation has more clicks in total, while the first has a higher ratio. The useful decision depends on the actual comparison method and audience conditions.
If you replace a thumbnail after the first observation window, describe the result as a sequential comparison. Do not present it as a randomised experiment unless that is what the method actually did. Keep watch-related observations separate from the click measure so a more clickable image is not automatically treated as a better viewing experience.
Before choosing a version
- The thumbnail accurately represents the video.
- Version references and observation windows are complete.
- The metric and denominator definitions match.
- Title and distribution changes are recorded.
- The conclusion does not extend beyond the comparison's evidence.