How to create a VSL hook test

Configure variants, goals, traffic, and decisions for a hook experiment without replacing the embed on your page.

A hook test helps you compare VSL openings by using different complete videos in the same embed. Moviie distributes the variants among eligible visitors, tracks the goal you choose, and brings the results together in one dashboard.

Use it to answer questions such as “does a direct promise generate more plays?” or “does a curiosity-led opening bring more people to the offer?”. To interpret the result confidently, change mainly the hook and keep the rest of the variants as similar as possible.

You do not need to replace the code installed on the page when creating the test, making a decision, or rolling it back. The same embed remains in the same place.

Before you start

Prepare the following:

an active embed that is already on, or will be added to, the sales page;
2 to 10 complete, available videos in the library;
a clear hypothesis with an easy-to-recognize difference between variants;
a CTA configured in the player if your goal is Unique CTA click.

Every selected item is a complete video, not only a short opening file. If you want to test only the hook, replicate the rest of the VSL in every variant.

Create the experiment

In the Moviie dashboard, open Experiments and select New experiment. The wizard organizes setup into four steps: Context, Variants, Goal, and Review.

1. Define the context

Under Internal name, enter a hypothesis your team will recognize later. Prefer a name such as “direct benefit vs. curiosity” over a generic name such as “test 1”.

Under Embed already on the page, choose the embed that will receive the variants. If the page is already live, select the exact embed used there. Its ID and snippet remain the same throughout the experiment.

Context step in the hook test wizard, showing the internal name and embed selection

The first step records the hypothesis and links the test to the embed that will remain on the page. Interface labels in this screenshot are shown in Portuguese.

2. Choose variants and the control

Select 2 to 10 videos. The first selected video becomes the Control, but you can mark another variant as the control.

The control is the reference used to organize the comparison. It receives no automatic distribution advantage. A useful practice is to use the current VSL as the control and name the other videos after the hook change.

Only complete, available videos can join the test. If a video is missing, confirm that processing has finished and that it belongs to the same organization.

3. Choose the primary goal

You select one primary metric. This choice determines which result is emphasized in the dashboard and should match the question in your hypothesis.

Option	What changes	When to choose it	What it does not mean
Unique play rate	Compares people who played with eligible exposures for each variant.	When the hook needs to overcome initial resistance and get the viewer started.	It does not measure retention, clicks, or purchases.
Pitch reach	Compares people who reached the offer point with eligible exposures.	When the hook's promise should hold attention until the pitch.	Manually seeking forward is not the same as watching continuously to that point.
Unique CTA click	Compares people who clicked the CTA with eligible exposures.	When a CTA is already configured and the goal is action in the player.	A click is not a sale, checkout start, or revenue.

The other metrics remain visible in the report. Choosing a primary metric prevents you from deciding only on whichever number looks most favorable after the test starts.

4. Set traffic distribution

With Equal, Moviie splits traffic evenly among variants. This is recommended for most tests because it makes comparisons simpler.

With Custom, enter the percentage for each variant. Every value must be valid and the total must equal exactly 100%. The continue button stays unavailable until the distribution is complete.

Use custom distribution only when you have a clear operational reason, such as limiting exposure to a riskier idea. A variant receiving less traffic will usually take longer to gather enough evidence.

5. Review and publish

In Review, check:

the hypothesis name and selected embed;
the primary goal;
variants, control, and distribution;
messages from the automatic check.

The check confirms that videos are available, there is exactly one control, distribution totals 100%, and the embed is not being used by another active test. Resolve every issue before selecting Create and publish.

Publishing makes the experiment ready for visitors. The analytics clock starts only with the first eligible exposure. Opening a dashboard preview is not the same as starting real traffic.

Monitor the experiment

Open Experiments and select the test. The header shows its current status:

Published · waiting for traffic: the test is live but has not received an eligible exposure;
Running: the test is receiving traffic and collecting results;
Paused: collection is temporarily stopped;
Concluded: a variant has been applied as the winner;
Inconclusive: the test ended without enough evidence to name a winner;
Archived: the test has been removed from the active workspace.

For each variant, the dashboard shows distribution, exposures, unique plays, pitch reaches, unique CTA clicks, and fallbacks. Fallbacks are occasions when the intended variant could not be served and a safe alternative was used. A high fallback volume should be investigated before making a decision.

A person is counted once per phase in unique indicators. These numbers should not be compared directly with total plays or repeated clicks shown elsewhere.

Decide on a winner

The dashboard tells you when a decision has enough support and lists any remaining conditions. Consider runtime, sample size, and the number of goal events. Do not name a winner only because one rate is temporarily higher.

When the test is ready:

choose Selected variant;
confirm with Apply winner;
verify the Concluded status.

New loads of the same embed receive the selected variant. The embed on the page does not change.

If an operational need requires an early decision, the tool asks for a reason and confirmation that the action is not a statistical conclusion. Use this route for incidents or urgent decisions, not as a shortcut for ending tests early.

Roll back a decision

After applying a winner, select Roll back decision to restore the reference version. The rollback opens a new phase and preserves the previous history, so you can distinguish what happened before and after the change.

Check the page as a visitor after applying or rolling back a decision. The same embed should continue loading without changes to the installed code.

Practices for useful results

test one primary hypothesis at a time;
keep the offer, length, CTA, and rest of the script equivalent when the question is about the hook;
choose the metric before publishing;
avoid changing videos while data is being collected;
do not treat a CTA click as a purchase;
wait for the readiness indication whenever there is no operational urgency;
document the learning, including inconclusive results.

Troubleshooting

The Experiments menu is missing: the tool may not be enabled for your account. Contact Moviie support.

There are not enough videos: wait for at least two complete videos to finish processing.

The embed is missing: confirm that it is active and linked to a video.

Publishing was blocked: read the review messages. Another active experiment on the same embed, a distribution that does not total 100%, or an unavailable video will prevent publishing.

The test is waiting for traffic: open the public page containing the embed. Dashboard previews, automation, and visits that do not meet measurement criteria may not be counted.

A rate looks good, but the decision is not ready: keep collecting traffic. A rate without enough time, sample, and goal events can change significantly.

Keep learning

Continue with these guides: