Hook lab

The first two or three seconds decide whether a viewer stays. The hook lab gives each new episode of a channel two or three alternative openings beside its own, posts a different one on each platform of a slot, and measures which kind of opening holds that channel's viewers. Once one kind is ahead with confidence, the channel's next episodes lean to it.

What it does#

Three kinds of opening#

The episode's own opening (the master's) is one of these, or a statement: the premise said plainly.

How an alternative opening is made#

While the episode is written, the writer also writes two or three alternative openings: a title, the first lines and their narration for each, and one cut point where every opening hands over to the rest of the episode. After the master is finished and safe in the library:

1. only each opening's first seconds are rendered again, on the GPU;
2. its narration is voiced, then the master's own from the cut point on;
3. it is joined to the master frame for frame at the cut, so everything after the cut is exactly the master;
4. it is checked like the master — the picture, the first frame, the seam at the cut, flashes, the voice, loudness and, where needed, the 90-second edit for Facebook and Instagram;
5. the critic reviews all the openings at once: each needs 7 out of 10 or more for its hook, must be clear, must deliver what it promises, must join the rest with no visible seam, and must keep to the facts and the platforms' policies.

An opening that fails anything is dropped, with the reason shown. The master always stands, so the hook lab never costs you an episode.

How openings are posted#

When a slot posts the episode, each platform gets a different opening while one is left, and the kinds are balanced across the channel's episodes on each platform, so every kind comes up in turn. The opening a post carried is recorded on it and never changes on a retry.

Instagram Trial Reels#

On Instagram you can post the openings as Trial Reels, which Instagram shows first to people who do not follow the account — so an opening can be tried without every follower seeing each version. If Instagram does not accept a post as a trial reel, it goes out as an ordinary one, and the channel page says so.

What the measurements learn#

About 48 hours after each post goes out (the measurement nearest 48 hours, between 24 and 96), the early hold is read for the opening it carried:

Each kind is compared with the channel's other openings on the same platform, with the same statistics as Insights: medians with 95 % intervals, permutation tests and Benjamini–Hochberg q-values. A kind is tested once it has 5 posts and 10 in all on that platform; before that it says Too little data yet. Results are labelled Strong evidence, Suggestive, No difference detected or Too little data yet.

When a kind is ahead with strong or suggestive evidence, it becomes the channel's preference — the channel's own result first, or, when the channel has none yet, the result across your channels of the same format. The writer of each new episode and the topic planner are told to lean to it. With nothing ahead, each episode keeps the writer's own choice.

Turn it on#

Editors, admins and owners can change it.

1. Open Channels, then the channel.
2. In Hook lab, tick Test alternative openings on this channel's new episodes.
3. Under On Instagram, choose how its openings are posted (below).
4. Press Save the hook lab.

Episodes made from now on get their openings; episodes already made keep theirs. Untick it and save to stop: new episodes then have their one opening again.

Settings#

What you see#

• On the channel page — the hook lab's setting, Which opening went where (the last posts, each with the kind of opening it carried, its own for the master's, and a trial reel badge), and What the openings taught: per platform, the median of the posts with each kind and, from 5 posts, its 95 % interval; the findings; and which kinds New episodes lean to.
• On the episode page — Openings: the master and each alternative side by side, with its kind, title, first lines and the critic's review. A dropped one reads Not kept — the master stands, with why.

Credits and plans#

Making and posting alternative openings is on every plan, the Trial included, within its credits. Measuring them and learning which kind holds viewers comes with Insights, on the Creator plan and above: on the Trial the openings are made and posted, and each episode keeps the writer's own choice.

An episode with alternative openings costs a little more — the openings' few GPU minutes, one more review and their narration — charged what it used, like the rest of the episode: about 10 to 20 credits more, by our estimate. The episode's price is shown before it starts, as always.

Limits and good to know#

• Time on the GPU. The openings are made only when the episode's own run leaves enough time; otherwise the episode keeps its one opening.
• Your numbers only. Each workspace's openings are compared on its own posts; nothing is pooled with another customer's.
• Ending the test. Turning the hook lab off keeps everything already measured; no post carries an opening any more.

Troubleshooting#

Related#

• Insights — the measurements and statistics behind it
• Captions and endings — endings that loop
• Packaging lab — titles and thumbnails
• Retention to script loop
• Scheduling and autopilot