An explainer video should end with a behavior: one spoken sentence that tells the viewer what they can now do, followed by a call to action that matches the page the video lives on. A video that ends on a tagline like "Ship with confidence" throws away its most valuable ten seconds, because the viewer understands your product best at the end, and a slogan gives that understanding nowhere to go.
Most videos in the genre end on a logo card, a tagline, and a swell of music, and the habit survives because endings get written last, when the team is tired and the template is sitting right there.
The running example is a 90-second explainer for a feature-flag platform. On screen, a risky feature ships behind a flag, a rollout percentage climbs from 5% toward 100% while a health chart stays green, and when an error rate spikes, one click rolls the feature back.
The ending is worth more than the opening
A 90-second explainer spends its whole runtime building a causal model in the viewer's head. By second 85 of the flag video, the viewer holds a working machine — flags gate features, rollouts widen gradually, rollback is instant — and that model starts decaying the moment the tab closes. The viewer will never again be this ready to act, so the last ten seconds convert peak understanding into a next motion. A hook only buys the right to keep going.
End on a behavior the viewer could actually do
Every closing line gets held to one rule: it names a behavior or a payoff, never a slogan, and it keeps its verbs. A real pair from our review history, on a video about run logs:
- Rejected: "Every run writes one of these. Nothing about a run is a mystery."
- Accepted: "Every run produces a log automatically. When a workflow does something unexpected, the log is the first place to look."
The rejected line is an aphorism: it asserts a feeling instead of telling the viewer what to do. The accepted line hands the viewer an errand attached to a specific future moment — the next time their workflow misbehaves, they know exactly what to open.
Test your own closer by asking whether a viewer could fail at it. "Ship fearlessly" cannot be failed, so it cannot be acted on, while "the next risky feature you ship, put it behind a flag, and if the health chart dips, roll it back with one click" passes, because a viewer can do that on their next release.
The recap closer also works: one sentence that compresses the whole video with its verbs intact — "the flag gates the feature, the rollout widens it gradually, and rollback is one click." A viewer can repeat that to a coworker, and repeating it is a behavior too.
The closer and the packaging are different devices
The closer is the last narration line. It lives inside the film, it is spoken, and it obeys the teaching register the rest of the narration used: a full sentence, a behavior or a payoff, no ask. The narration never says "sign up today," because a voice that has spent 85 seconds teaching and then pivots to selling burns the trust it just built — the same failure as trailer-voice narration, arriving late.
The packaging CTA is everything around the film: the end card, the button under the player, the link in the description, the page itself. The ask lives here, in text and interface, matched to wherever the video is placed.
Because the closer stays constant while the packaging swaps, one video can be reused across placements, and the swap is cheap: an end card is an edit, never a re-record.
The final frame is part of the ending
The final frame sits on screen while the viewer decides what to do, and a paused embed shows it forever. In our pipeline it is one of exactly two stills rendered for review before any motion exists (the other is the money shot).
A good final frame shows the set piece settled in its resolved state. For the flag video: the rollout at 100%, the health chart green across its whole width, the rollback event still visible in the log — the frame itself says the risky feature shipped safely. A logo on a void erases the causal model the video just built.
- Hold the settled frame for 2 to 3 seconds. The payoff has already happened, so the stillness reads as a breath rather than dead air.
- Give a text end card about 3 seconds of its own. A URL plus one line needs that long to read, and it should be the only new thing entering an otherwise settled frame.
- Never close on a logo card. The name and mark belong in the corner or on the end card next to the claim, not alone on a black frame.
Match the ask to the page
The right ask depends on what the viewer was doing ten seconds before they pressed play, and in our experience placement matters more than anything inside the last scene itself.
- On your homepage, ask for nothing. The signup button is already rendered a hundred pixels below the player, and the video's job is to make that button make sense. End on the recap closer and the settled frame; a spoken "get started at…" narrates the furniture.
- On Product Hunt, end on the name and the one-line claim. The page provides its own voting and visit buttons, so the video competes to be remembered an hour later rather than clicked this second. End on a recap sharp enough to survive being quoted in a comment: "instant rollback for every release" outlives "the future of shipping." The Product Hunt guide covers the rest of that page.
- In docs, end on the next step inside the product. The viewer is mid-task, so the closer names the next concrete action. The best docs ending we've shipped was a plain handoff — "now that we have a high-level understanding of the module, let's try this out inside the product" — with no card and no button, because the CTA is the next paragraph of the docs.
- In ads, put the ask on an end card and keep it out of the voice. The viewer did not seek you out and will never find the link on their own, so this is the one placement for a hard, explicit ask: URL and one imperative line on a card, held a full 3 seconds. Ad viewers bail early, so the product name and payoff have to land in the first half; the end card is for the viewers you kept.
Write the closer before anything else
The closing line is the only narration written before the visuals exist. Everything else waits until the picture is locked, but the whole video is choreographed toward the closer, and the scenes exist to make that final sentence feel obvious rather than asserted. In the flag video, the rollout scene is in the film because the closer needs the viewer to already believe that rollback is safe.
Writing the closer first also works as a planning gate: if you cannot state the behavior a viewer should walk away with, you are not ready to storyboard. The full authoring order lives in the script guide.
FAQ
Should the CTA be spoken or on screen? On screen, almost always. The voice ends on a behavior or a recap, and the ask lives in the end card or the surrounding page. The exception is docs and course content, where a spoken handoff into the product is the natural close.
Is "Sign up free" ever an acceptable ending? As an end-card line in an ad, yes. As the last spoken line of an explainer, no, because it converts the narrator from teacher to seller in the final second, and viewers feel the switch.
How long should the ending hold? Give the settled final frame 2 to 3 seconds, and a text end card about 3 seconds of its own. Longer holds turn into dead air at the tail.
We use one video everywhere. Do we need different endings? You need different packaging around the same film. Keep the closer and the final frame constant, and swap the end card per destination: none for homepage and docs, name plus claim for a launch page, URL plus ask for ads. Cutting variants of the last three seconds is the cheapest edit in the entire production.
The fastest way to test an ending is on your own product: send us the URL, watch twenty short videos of it, and pick the one worth finishing.