YouTube Packaging Test Loop
Let audience behavior choose titles and thumbnails after launch
- Difficulty
- Easy
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 98%
Treat a video's title and thumbnail as measurable packaging rather than a matter of taste. Create at least two credible options. Before publishing, poll the YouTube community and launch with the option viewers prefer, giving the initial package an informed starting point. After roughly a week, use a split-testing tool to alternate the second option on the same live video, preserving its views and engagement while exposing comparable audiences to each version over time. Test one variable at a time, allow enough data for the tool to identify a winner, and retain the package with stronger click performance. Repeat the loop across the catalog. Carmichael emphasizes that the objective is not what the creator likes but what works. Titles and thumbnails deserve priority because their effect is material; descriptions and tags can be tested, but he has not seen evidence that tags drive comparable growth.
Origin
Carmichael describes using YouTube's community tab before launch and TubeBuddy afterward. His team had run more than 1,800 split tests and planned thousands more, primarily comparing titles and thumbnails.
Core principles
- 01Packaging should optimize for qualified clicks, not creator preference
- 02The audience can improve the first test before publication
- 03One live video can test alternating titles or thumbnails
- 04Tests need enough data before a winner is declared
How to run it
- 1
Create two credible options
Draft two titles or thumbnails that accurately represent the same video and could each plausibly win.
Pro tip Apply proven headline principles while preserving the video's promise.
Watch out A strong option against a knowingly weak option teaches little.
- 2
Poll before launch
Ask your community which option they prefer and publish with the winner.
Pro tip Include the video's subject so voters understand the promise.
Watch out A poll measures stated preference, not actual clicks.
- 3
Establish the baseline
Let the published package run for about a week before starting the formal split test.
Watch out Testing too early may mix launch effects with packaging effects.
- 4
Alternate one variable
Use a tool to rotate the alternative title or thumbnail on the same video across comparable periods.
Pro tip Keep the video itself unchanged.
Watch out Changing title and thumbnail together hides which caused the result.
- 5
Choose by behavior
Wait for sufficient data, retain the higher-performing package, and record the lesson for future drafts.
Pro tip Look for patterns across many tests rather than universal rules from one video.
Watch out Low-volume tests may never produce a reliable answer.
In the wild
An interview channel compares thumbnails showing the host alone, the guest alone, and both together. It polls subscribers to choose the launch image, then formally tests the best alternative on the same live video after one week.
→ The channel keeps the image with the stronger click-through rate instead of relying on the producer's preference.
Common mistakes
Testing multiple variables together
Changing both headline and image makes the source of any lift unknowable. Isolate one variable.
Choosing by personal taste
The creator is not the whole audience. Use observed behavior to settle close calls.
Overvaluing tags
Carmichael considers tags a tiny contributor compared with titles and thumbnails. Allocate testing effort accordingly.
Is it for you?
Best for
YouTube channels with enough viewers to produce meaningful community polls and split-test data.
Not ideal for
Very small channels where low impressions make test results unstable or inconclusive.
From the transcript
“it's not about what i like is about what works”
“use your community tab there's a community tab on youtube where you can pull your audience that most people never use”
“what tubebuddy is going to do is take that same video and every day change the title”
From the episode
Evan Carmichael: Surviving Entrepreneurship
Evan Carmichael