15secLearnings methodology

Experiment methodology

A controlled test before the public drop.

Compare different edits with real viewers, choose a winner with confidence, and turn every result into evidence for the next video.

Variants
2–5 edits
Audience entry
One test link
Routing window
Up to 72 hours
Define the test

Variants

A variant is a different edit of the same video: a different hook, pacing, structure, or payoff. Every variant uses the same title and thumbnail, keeping the video edit as the primary intentional difference.

Recommended2–5 comparable edits
Prepare privately

Unlisted videos

15sec uploads each variant to your YouTube channel as an unlisted video. They stay off your public Videos tab and out of search, so test viewers reach them through the experiment only.

VisibilityUnlisted on YouTube
Route the audience

The test link

Share one test link with your audience. Each new browser is randomly assigned to an eligible variant, and repeat visits from that browser return to the same one. Every variant has the same view cap.

Why randomization matters

Separate links shared at different times can expose each edit to a different traffic spike, audience mix, or viewing window. Assigning arrivals from one shared entry point distributes those conditions across variants, making the edit the primary intentional difference.

Random first assignment
New browsers are assigned among variants that still have capacity.
Concurrent traffic
Every eligible variant is open to assignment during the same test window.
Consistent repeat visits
The same browser returns to its original variant instead of being counted across edits.
Equal assignment targets
Routing stops when every variant reaches the shared cap, after 72 hours, or when you stop it.
Scope of control

The view cap counts viewers routed through the smart link; performance results come from YouTube Analytics. 15sec does not create traffic or control who receives the link. A different device, private browser, or cleared browser data can be treated as a new viewer, and randomization does not guarantee identical viewer demographics in a small sample.

AllocationRandomized, concurrent traffic
Compare performance

Results

Once routing stops, compare hook retention, the full retention curve, watch time, engagement, and subscriber impact for every variant. If enabled, auto-publish can move the winner live for you.

DecisionEvidence tied to each cut
Build an advantage

Learnings

Each completed experiment adds evidence to your channel's Hook model. It surfaces hook patterns that repeatedly help or hurt retention across your videos. Later video segments are not part of Learnings for now.

Learn more
Compounding valueOne channel-specific Hook model

Ready to test?

Run your first controlled experiment.

Choose your variants, set the view cap, and share one test link.

Go to dashboard