A high-converting VSL follows the sales-letter order: hook, problem, story, mechanism, proof, offer, price, guarantee, close, and then a post-pitch loop for the viewers who are still watching. Each section has exactly one job, and each one has a number on the retention curve that says whether it did that job. Structure does not sell on its own. What it does is make the video measurable, so when it underperforms you know which minute to rewrite instead of rewriting all of it.
What follows is the outline as a builder uses it. For each section: what it has to achieve, how to write toward that, what judges it, and which player setting attaches to it.
One rule before the outline
Write for one viewer, cold. They came from an ad, they do not know you, and they have a thumb on the scroll. Everything in the structure below exists to move that one person from "who is this" to "where do I pay" in a sequence of beliefs, each one earned before the next is asked for. If a section does not change what the viewer believes, it is not a section. It is a pause.
One offer. One call to action. If you are tempted to sell two things, write two videos.
The hook: buy the next thirty seconds
The hook does not sell. It earns the right to keep talking. Open on the viewer and their situation, never on you, your company, or your credentials. Name the thing they already suspect is true, or show them a result before you explain it, or ask the question they typed into their own head last night. Then promise that the next few minutes will resolve it.
Write the first sentence assuming the sound is off. Every modern browser starts a video muted until the visitor touches the page, so the opening frame has to hook visually too: a caption, a face, a number on screen. Smart Autoplay turns the sound on at the first touch anywhere on the page without restarting, which is why the first spoken line should still work for someone who joins it two seconds in.
Judged byHook Rate, the share still watching past five seconds, and Unmute Rate. If either is low, nothing further into the video matters yet.
The problem and the story: make it cost something
Now the viewer is listening. Spend this section making the problem expensive. Not with adjectives, with specifics: what it costs them each morning, what they have already tried, what they told themselves the last time it failed. A story does this better than a list, because a story has a before, and the viewer recognises themselves in it.
Keep the story on the viewer's side of the screen. A founder story belongs here only if the founder had the viewer's problem. Resist the urge to foreshadow the solution. The moment you name it, the problem stops hurting.
Judged bythe slope from the hook to the midpoint. A steady slide here means the story is longer than the pain it is describing. Hold Rate is the checkpoint.
The mechanism: why this works when the others did not
The viewer has tried things. The mechanism explains why those things failed and why this one is different, and it is the section most likely to break belief, because it is where you make your biggest claim. Explain it the way you would to a sceptical friend: the cause nobody addressed, the reason the usual fixes miss it, the thing you do instead. Make it simple enough to repeat, because the viewer will repeat it to someone before they buy.
Judged bya cliff right after the claim on the retention curve. Switch to buyers against non-buyers: if both lines drop at the same second, the claim landed badly; if only the non-buyers leave, you have found the people who were never going to believe you.
The proof: pay for the claims you made
Every claim in the mechanism is a debt, and this section pays it. Demonstrations before testimonials, evidence before adjectives, and your own results only if you can show them. Do not invent numbers, customer results, or studies. A cold viewer has heard every fake statistic there is, and a curve does not forgive the second they stop believing you.
Order the proof by the objection it answers. "Does it work" first, "will it work for me" second, "is this person real" third. Cut any proof that answers a question nobody at this point is asking.
Judged bytwo shapes. A long flat stretch that nobody needed is proof to trim. Viewers rewinding to watch a demonstration again is proof that is carrying weight, and deserves more time, not less.
The offer and the pitch: say the price on a known second
This is the timestamp the whole video is built around. Reveal what the viewer gets, in the order they will use it, then the price, then why the price is smaller than the problem. Say the number plainly. A price that arrives wrapped in qualifiers sounds like an apology.
Put the pitch on a known second. The retention graph finds it in your transcript and draws the Pitch marker there, with no setup. Put a CTA card on the same second so the buy button appears the moment the offer lands, not on page load. If the order box, the guarantee, and the FAQ live on the page under the video, hide them behind a watch gate until this moment. A viewer who scrolls down at minute two and finds a price has been given permission to skip the argument.
Judged bypitch reach, the share still watching at the marker, and the CTA card's shown-versus-clicked count at that exact second. High reach with low clicks is an offer problem, not a video problem.
The guarantee, the bonuses, the close
The viewer is interested and looking for a reason not to. The guarantee removes it. State the terms in one sentence, then say what happens in the worst case in plain words. Bonuses come next, and each one should answer an objection rather than pad the stack. Then ask. Plainly, once, with the exact action: click the button below, choose your package, enter your details. Do not soften the ask with a joke.
Judged byClick Rate after the pitch and, from the cart postback, the conversion rate of the people who reached the pitch. Clicks that do not become orders are a checkout problem, and no rewrite of the close will fix them.
The post-pitch loop: the video keeps selling after the ask
Some viewers reach the pitch, hear the price, and keep watching. They are not lost. They are deciding. The section after the close exists for them: restate the offer in different words, answer the three objections you know they have, tell the story of one person who hesitated, and restate the guarantee. Then ask again with a second CTA card.
This is the only part of a VSL where Completion Rate means something. Everywhere else, the viewer who clicks at the pitch and never sees the ending is your best viewer.
Judged bythe second CTA's clicks, and the shape of the curve after the pitch. A viewer who stays through the loop and still does not click has an objection you did not answer.
The page around the video
The structure does not stop at the edge of the player. Above the video, a headline that repeats the promise of the ad, so the viewer knows they landed in the right place, and a one-line prompt to turn the sound on. Nothing else above the fold. Below it, the order box, the guarantee, and the FAQ, hidden by timed events until the pitch and revealed together with the video's own reveal.
Two more settings belong to the page rather than the script. Exit intent shows one designed message to the viewer heading for the back button, once, with a cooldown so it never nags. And for a long video, Continue Watching offers a returning viewer Continue or Restart, on any device, for 90 days, because for long-form offers the second session is often the one that buys. If the funnel captures leads before the pitch, a lead form at first play or at a chosen second does that inside the player, and the lead arrives with their watch depth attached.
The outline on one screen
| Section | Its one job | Judged by | Player setting |
|---|---|---|---|
| Hook | Earn the next thirty seconds | Hook Rate, Unmute Rate | Smart Autoplay |
| Problem and story | Make the problem expensive | Slope to midpoint, Hold Rate | Smart progress bar |
| Mechanism | Explain why this works when others did not | Cliff after the claim, buyers vs non-buyers | Retention cohorts |
| Proof | Pay for the claims | Flat stretches, rewinds | Retention graph |
| Offer and pitch | Say the price on a known second | Pitch reach, CTA shown vs clicked | Pitch marker (automatic), CTA card, watch gate |
| Guarantee and close | Remove the last reason not to | Click Rate, conversion rate from the cart | Cart postback |
| Post-pitch loop | Convert the ones still deciding | Second CTA clicks, curve after the pitch | Second CTA card |
| The page | Match the ad, hide the price until the pitch | Play Rate, exit-intent clicks | Timed events, exit intent, Continue Watching |
Structure is a hypothesis
The outline above is where the first version starts, not where the last one ends. Publish it, send real traffic, and read the curve with buyers and non-buyers separated. Once the video has passed roughly 300 plays, AI Optimize reads the transcript against the retention, clicks, and conversions and points at the dead zones and the CTA placements. When two versions of a section disagree, clone the video, change one section, and run both under one embed as a split test decided on revenue per play.
That is what "high-converting" means in practice. Not a template that converts on day one, but a structure where every section can be judged, and the weakest one can be found and fixed in an afternoon. If you want to read your own video that way, the free plan gives you 3,000 plays a month, free forever, with no card.