The ask is where the whole video cashes out#
A demo video has exactly one job downstream of being watched: to produce a next action. Everything before the final beat, the hook and the walkthrough and the proof, is setup for the single moment when the viewer decides to do the thing you built the video to make them do. That moment is the call to action, and it is where the whole video cashes out. Get it wrong and a demo that held attention for ninety seconds converts nobody, which is the most expensive kind of failure, because you paid for the attention and then declined to spend it.
Two failures account for most of it. The first is the menu: an end card offering five things at once, start a trial, book a call, read the docs, follow us, watch another video, on the theory that more doors mean more exits. The second is misplacement: a single clean ask welded to the last frame of a four-minute video that two-thirds of the audience already left. Both are fixable, and both reduce to two decisions you make once, which ask and where it goes. This piece is only about the close; the end-to-end production playbook covers the rest, and the opening is the bookend at the other end of the same timeline.
Pick one ask, keyed to the funnel stage#
The right ask is not a matter of taste. It is a function of how warm the viewer is and how much friction the action costs them. A cold viewer who found a README clip will star a repo but will not book a sales call, and a buyer three weeks into an evaluation wants the call, not another docs link. Match the ask to the temperature of the traffic.
| The ask | Funnel stage | Friction | Fits which product | Where it belongs |
|---|---|---|---|---|
| Star the repo | Awareness (open source) | Lowest, one click | An open-source tool | A spoken line plus the description |
| Read the docs | Consideration | Low | A self-serve, technical product | End card and description |
| Start free | Activation (PLG) | Medium, an account | Self-serve SaaS | End card, plus a button beside the embed |
| Book a call | Sales | Highest, a calendar hold | Sales-led, higher-ACV B2B | A sound-on follow-up and the sales email |
| Join the waitlist | Pre-launch | Low | An unreleased product | End card and description |
Read down the friction column and the rule falls out: the colder the traffic, the lower-friction the ask has to be. A repo star or a docs link is the right first ask for an awareness clip, because it costs one click and no commitment. Reserve "book a call" for the warm, sound-on, post-discovery send, where the viewer has already decided you might be worth an hour; the async sales demo is built around exactly that ask. Whichever you choose, Wistia's read on the 36,000 CTAs it analyzed in 2024 holds: the ones that convert are built from action-oriented verbs, not vague invitations, and video CTAs on its platform convert at about 16% on average1, a number worth holding onto when we price placement below.
Why one ask beats a menu of five#
The instinct to offer five asks is the instinct that more options cannot hurt: if someone will not buy, maybe they will subscribe. The evidence runs the other way. In Iyengar and Lepper's 2000 field experiment, a supermarket display of 24 jams drew 60% of passing shoppers to stop, against 40% for a display of six, so the big menu won the glance. But only 3% of the shoppers who met the 24-jam display bought a jar, versus 30% of those who met the six3. The extensive display won attention and lost the sale by a factor of ten.
An end card of five asks is the 24-jam table. It wins a half-second of notice and then hands the viewer a decision, which of these do I want, at the exact moment you needed them to act rather than deliberate. Hick's law names the cost plainly: the time to choose grows with the number of options, and a viewer weighing five links past the end of a video mostly resolves the choice by closing the tab. YouTube quietly encourages the pile-up, letting you place "up to four elements" on an end screen for a standard 16:9 video2, which reads as permission to use all four. Use one. A single ask is not a constraint you settle for; it is the version that converts, because it replaces a decision with an instruction.
Placement: the end card only reaches the finishers#
A CTA is seen only by the viewers still present when it appears, which makes its real reach a product, not a constant: effective reach equals the people who started, times the fraction still watching at the ask's timestamp. The end card sits at the one point in the video where that fraction is smallest. Vidyard's 2025 benchmark, drawn from roughly 943,000 videos, puts a sub-minute video's completion at 65% and a past-twenty-minute video's at 20%4; everything longer than a tight cut sheds audience before the end card ever renders.
Price it out. The figures below are illustrative, anchored to Vidyard's endpoints and the plain fact that a retention curve only falls, but the ratios are the point, not the decimals.
| Video and CTA placement | Retention at the ask | Reach per 1,000 plays, at ~16% CTA conversion |
|---|---|---|
| 55-second cut, end card at 0:50 | ~65% | ~104 act |
| 4-minute walkthrough, end card at 3:55 | ~30% | ~48 act |
| 4-minute walkthrough, ask stated at the aha (~1:20) | ~70% | ~112 act |
Same ask, same 16% who act once they see it; placement times length is what moves the outcome from 48 to 112. The lesson is not "always put the CTA in the middle." It is that the end card is a good home for the one ask only when the video is short enough that most people reach it, and a poor one the moment the video runs long. Wistia's placement guidance encodes the same curve from the other side: it recommends the CTA in the last quarter for a one-to-three-minute video, but at the halfway mark once a video runs three to five minutes, and earlier still beyond that1, because the longer the video, the fewer finishers there are to catch. Two hard platform limits sharpen the point: a YouTube end screen requires the video to be at least 25 seconds long and lives only in the last 5 to 20 seconds2, so a fifteen-second social cut cannot carry one at all, and its ask has to be burned into the frame or parked in the description.
This is where "one ask" and "place it where viewers are" stop fighting. You do not solve the retention problem by adding more asks; you solve it by placing the same ask where the audience actually is. State it once out loud at the aha moment, when intent peaks and most of the audience is still there; show it again on the end card for the finishers; and leave it a third time in the description or a pinned comment, the durable link that outlives the end card and catches the viewer who scrolls after the clip stops. One destination, three touchpoints, not three destinations.
The closing ask is a re-render, not a reshoot#
The CTA is the line you will change most, because it is the one you can only grade against real behavior. You will learn that "start free" out-pulls "book a call" for your traffic, or that the ask lands better spoken at 0:40 than shown at 0:55, and each of those lessons is a rewrite. In a hand-recorded demo the ask is welded to the take: the end card is baked into the export, the spoken line sits in a voice track cut to the old timing, and moving it means recording the whole thing again to keep the audio and the cursor in sync. So most teams ship the first CTA they captured and never test a second.
Defining the demo as a script and an action list makes the ask a variable instead of a weld. Our engine, aidemo, works this way: the closing line and the end card are a few fields in an agent-authored storyboard, so swapping the ask or moving it earlier is an edit and a fresh render rather than a reshoot. The honest limits come with the approach: aidemo captures a browser only, a coding agent writes the storyboard as text with no visual timeline to assemble it on, and there is no drag-to-trim editor for nudging the end card by eye. The principle outlives whatever renders the pixels. Decide on one ask, place it where the viewers still are, and keep it cheap enough to change that you actually test it, then instrument the result, because which ask converts is a question only your own click data answers, and where the video lives decides which placement you needed in the first place.
Sources#
- Wistia — How to Add a Video CTA to Increase Conversions (2024 analysis of 36,000+ CTAs)
- YouTube Help — Add end screens to videos (25-second minimum, last 5-20 seconds, up to four elements)
- Iyengar & Lepper — When Choice is Demotivating (Journal of Personality and Social Psychology, 2000)
- Vidyard — Video in Business Benchmark Report (completion by length, ~943,000 videos)
FAQ#
Where should a demo video's call to action be placed?#
Wherever the most viewers still are, which is rarely only the last frame. On a tight sub-minute cut the end card works, because roughly two-thirds of viewers reach it; on anything longer, retention has thinned, so state the ask out loud at the aha moment and reinforce it on the end card and in the description. Wistia's placement data follows the same logic, moving the recommended CTA earlier as a video runs longer, precisely because fewer people finish it.
How many calls to action should a demo video have?#
One ask, though you can state that one ask more than once. Offering a menu of five competing links reproduces the choice-overload effect Iyengar and Lepper measured with jam, where a 24-option display converted at 3% against 30% for six. Pick a single destination and place it in up to three spots, spoken at the aha, shown on the end card, and left in the description, rather than sending viewers to three different places.
What makes a good call to action for a software demo?#
An ask matched to how warm the viewer is and how much the action costs them. A cold viewer on a README clip will star a repo or open the docs but will not book a sales call, so lead with the low-friction ask and save "book a call" for warm, post-discovery sends. Wistia found across more than 36,000 CTAs that the ones that convert use action-oriented verbs, not vague invitations like "learn more."
