What a SaaS promo video realistically gets you

After fifteen years of making these videos for a living, I went and read the primary sources behind every conversion statistic my own industry quotes. Here is what a promo video can honestly be expected to do for a SaaS product, what it cannot, and the point where the evidence runs out.

Start with the room where those numbers get used. Here is the composite version of a kickoff call I have sat in more times than I can count. Somebody senior asks, quite reasonably, what lift they should put in the deck to justify the video. There is a pause. Then a number arrives from somewhere further down the table — it is almost always eighty percent — and it gets written down.

Nobody asks where it came from. I never asked either, and I was the one being paid to make the thing.

I've spent fifteen years on the production side of this, for Microsoft, Cloudflare and a long tail of startups, and I now build one of the tools that makes these videos, so weigh what follows accordingly. I've sat in a lot of those rooms, and for most of them I nodded at the slide. It was a number in favour of the thing I was being paid to do.

Eventually I went and read the primary sources. So here's what I think you can honestly expect from a promo video, ordered from the best-evidenced thing I can show you down to the thinnest. It gets thin quickly. I'd rather point at where the ground gives way than pretend it's solid all the way across.

The strongest evidence is about budget, not about video

If you want one experiment where somebody controlled the variable and spent real money finding out, it's Wistia's, from 2018. They hired Sandwich Video to make three versions of the same ad for Soapbox — a product Wistia discontinued in September 2024 — at $1,000, $10,000 and $100,000 production budgets, then put a $60,000 test ad budget behind all three over three weeks on Facebook and YouTube. Worth saying before the results: Wistia ran the test, Wistia published it, and it flatters Wistia, Soapbox and Sandwich all at once.

The middle one won. On YouTube, in that 2018 test, the $10,000 version averaged $6.66 cost per install against over $10 for both the $1,000 and the $100,000 ads — "nearly half the Cost Per Install," in Wistia's words. The write-up said the Facebook gap was wider still, but published no Facebook figures.

The rest of the disclosure belongs in the same breath, because it's what lets you trust anything else here: the test has been publicly argued to be a flawed experiment. One test, three weeks, one creative concept, no significance testing and no replication.

There's a detail in the same 2018 campaign that gets skipped. The sequential brand-building format drove nearly 600,000 impressions at just over $10 CPM, which reads well until you see the $281 cost per install beside it; Wistia paused that format mid-campaign. Impressions are not the thing you were buying.

What I take from it matches fifteen years of watching budgets get signed. Past a fairly low threshold, production money stops buying performance and starts buying other things — reuse, longevity, a creative director who tells you no. Those are real, and the cost side of this argument is worth its own read. They're just not cost per install. The most expensive version losing is the opposite of what my industry, me included, has an incentive to tell you.

The number everyone quotes, and the number in the same document

The eighty percent is real in the sense that a document containing it exists — not in the sense that anyone measured it.

Here's the specimen everyone circulates: "a video on your landing page increases conversion by 80%." It comes from exactly one place: EyeView Digital's Making Video Accountable, a sales white paper written in May 2010 by a video vendor that produced and sold the videos it was measuring. It was rehosted on a marketing blog in September 2012, which is the only reason you can still read it: EyeView ran out of money and shut down in December 2019.

And here's the sentence the white paper actually says: "In some cases the boost in conversion was over 80%." That's a best case, drawn from what the paper calls hundreds of tests for over one hundred customers, with no sample sizes, no visitor counts, no confidence intervals and no distribution published anywhere in it. It's the tail of a distribution, and sixteen years later it still gets passed around with the "in some cases" filed off.

The 86% version that also circulates is worse rather than better. That one is the same 2010 paper's TutorVista example, which says the figure — "over 86 percent relative to the original landing page" — arrived only after several rounds of testing launch mechanisms, playback types and post-roll animation: a fully optimised page beating a no-video page, not the effect of adding a video. And the paper never says which configuration won, so nobody quoting it knows whether the winner was autoplay with audio, which browsers no longer permit by default, or autoplay without audio, which they still do.

The useful number is in the same PDF, a few pages away. EyeView ran a split test on eToro's FX-trading landing page around 2010: the existing page against the same page with a video embedded, 50% of new users to each. In its words, "once the test reached statistical significance it demonstrated that the conversion rate for Group B who received the page with the video was almost 32 percent higher than the conversion rate for Group A."

That's the only single-variable video-versus-no-video test in the document with a number attached, and it's about a third of the figure the same document is famous for. Its limits, plainly: EyeView produced and sold the video being tested, so this is sales collateral rather than research; no visitor counts, confidence level or absolute conversion rates were published; and it's roughly sixteen years old, from before mobile-majority traffic.

The evidence I don't have

By now you should be asking why a man with fifteen years of this behind him is quoting a defunct vendor's brochure at you instead of his own work.

The honest answer is the least flattering thing in this post. I almost never got told what happened. Here's the composite: the video ships, the invoice clears, the thread goes quiet, and a year later somebody forwards a dashboard screenshot with no dates on it and no control. Production is downstream of the measuring, and the people doing the measuring have no reason to loop you back in. That isn't a grievance. It's just the shape of the job.

What I have instead is a set of impressions, and I'm labelling them as impressions because that's the entire argument of this post. The expensive one legal made us re-record, which nobody ever embedded. The cheap one still sitting on a homepage six years later. The one a client stopped using inside a month — not because it underperformed, but because the positioning moved underneath it and the video couldn't move with it.

None of that is a measurement and I won't dress it as one. I sell video, so add me to the list of people whose numbers you should discount. The only difference between my impressions and the figures above is that I'm telling you which is which.

Ask what the control actually was

Two more tests get quoted constantly, and both are worth knowing in their real form.

Vidyard co-founder Michael Litt wrote up his own split test from the company's Y Combinator days in an October 2011 guest post: an email-capture page converted at 6.5% with no video and 11% with an explainer embedded, a 69% lift, with a lightbox variant reported at 13%. What that supports is the coarse claim — both video treatments beat no video. The variants are lettered D, H and J, which implies roughly ten variants of which three were reported, and the post never says the pages were otherwise identical. It's also a video-hosting founder writing on a landing-page vendor's blog, where the winning variant happens to be that vendor's own lightbox feature.

Then Crazy Egg's 64%, from an August 2012 case study by Conversion Rate Experts — the agency that was paid for the work, publishing its own result. The sentence is real: the version of the page with the video in it "generated 64% more conversions than the control." What the case study never says is which page was the control for that particular test, and the story around it is a full homepage rewrite that had already beaten the original page. So you can't tell from the document whether 64% is video against no video, or a rewritten page with a video against something else entirely. "Adding a video was worth 64%" is an inference nobody in that case study ever made explicit.

I'll spare you the fourth recital of no sample size, no duration, no significance. Assume it. Every figure in this category was published by somebody with a stake in the answer, and not one of them shows its arithmetic.

Three tests, three vendors, spanning 2010 to 2012. That's the whole published foundation under a claim you've probably seen stated as settled fact this week.

Roughly six in ten people never press play

This is the constraint nobody puts in the deck, and it comes from the one organisation here with no video to sell.

Baymard Institute runs moderated think-aloud usability testing and benchmarks large e-commerce sites. In a May 2019 write-up it reported that 41% of users watched product videos during its product-page testing — leaving 59% who didn't, some of them because they simply dislike the format. Baymard is more optimistic about that 59% than I am, reading most of it as a findability failure rather than a preference; it also states no participant count, says its research isn't designed to produce precise statistical conclusions, and is measuring retail product pages rather than SaaS landing pages.

Take it as a shape rather than a rate. Whatever a video does for conversion, it does only to the fraction of people who start it. Which is why the first three seconds and where the thing sits on the page aren't polish — they're most of the effect size.

For the ones who do press play there's Wistia again, analysing over 36,000 calls-to-action inside videos hosted on its own platform during 2024 and reporting that video CTAs convert at about 16% on average. That figure comes from its 2025 report and wasn't refreshed for the 2026 edition, published in April 2026, whose methodology pairs 13 million videos and 79 million hours of viewing with a survey of 900+ professionals. And none of it compares a page with a video to the same page without one. It's good evidence about how to design the end of a video, and no evidence at all that video lifts signups.

The biggest landing-page dataset in the business says nothing about video

The largest public dataset I could find is Unbounce's Conversion Benchmark Report: 41,000+ landing pages, 464 million unique visitors and 57 million conversions, covering July 2023 to July 2024. It cuts conversion by industry, channel and device, and it goes fine-grained enough on copy to define "difficult words" as words of three or more syllables and report a -24.3% correlation between them and conversion rate.

It reports nothing about video. There's no media or embed variable in its published methodology, in that edition or the 2020 one, and the full report sits behind a form.

A dataset sensitive enough to detect a syllable-count effect has no video finding in it at all. That isn't proof video does nothing. It's a good reason to distrust any precise, general, present-tense number for what video does.

The famous stories are stories, and they're still useful

Dropbox is the one that earned the genre. Drew Houston posted a screencast for the March 2008 private beta, seeded with in-jokes for the Digg audience, and by his own account, quoted by Eric Ries in TechCrunch in October 2011, it "drove hundreds of thousands of people to the website" and the waiting list "went from 5,000 people to 75,000 people literally overnight," with more than 10,000 Diggs inside 24 hours. Those are the founder's own figures, never audited, and the source credits Digg alone. Copying the video without the distribution copies the wrong half.

Dollar Shave Club isn't SaaS, but it's the reference everyone reaches for. The video went live on 6 March 2012, the same day the company announced a $1M seed round. By Michael Dubin's own account in Inc., it went up at 6am and the site was dark by 7:30, where it stayed about a day; around 12,000 people signed up in the first 48 hours, per the New York Times in April 2013. Entrepreneur reported in October 2012 that it cost $4,500, attributed to its director Lucia Aniello, who estimated $50,000 for comparable polish — the same finding as Wistia's, arrived at six years earlier by accident.

And Slack, held up as proof that a great promo launches a company: the Sandwich film went up on Slack's channel on 12 August 2014, six months after Slack's general-availability launch on 12 February 2014. It was commissioned for that launch and arrived long after it. James Sherrett, Slack's ninth employee, has written that in 2014 commissioning a Sandwich video was the hottest thing a tech startup could do — a rite of passage as much as a growth play. It has around 1.47 million views on Slack's own channel as of August 2026, and no signup, conversion or cost figure for it exists in any source I could reach. Anyone quoting you one invented it.

Cheap and on time. Cheap and lucky. Expensive and late. There's no budget pattern in there at all.

What to expect from a SaaS promo video

There's a fair objection to everything above, and it's the one I'd raise. If those sources are as weak as I've just spent two thousand words showing, the 32% can't carry an expectation either, and handing it back at the end is a conjuring trick.

That's right, so here's what I think actually survives. The 32% isn't a forecast, it's a boundary. The only clean measurement anybody published is nowhere near 80%, and it isn't zero. Everything between those two edges is your traffic's business, not mine.

Inside that, expect a video to help the people who watch it understand you faster than your copy does. Expect roughly six in ten never to press play, on 2019 e-commerce usability data rather than anything measured on SaaS. Expect production spend past a few thousand dollars to buy craft, reuse and calmer stakeholders rather than a better cost per acquisition. And expect no video to create demand that wasn't there; every famous case above had distribution, timing or funding news doing at least half the work.

If you want to know what it did for you, the method is duller than the number. Put the video up and leave the copy alone. Then change the copy and leave the video alone. If you swap both at once, as almost everyone does, you've bought a story instead of an answer.

And if your traffic is small enough that a lift like that would be invisible in a fortnight, which is most SaaS at the start, don't pretend otherwise. Make the video for the reasons that don't need a test, and watch the play rate rather than the signup rate — if six in ten aren't starting it, your problem is the thumbnail, the placement and the first line, and none of that is a rendering problem.

The thing that isn't in any of the research: what it costs to find out

Everything above is about what a video does once it exists. None of it helps you if you can't get one made.

That is the real constraint for most SaaS products, and it has nothing to do with conversion rates. Bespoke animated promo work is commonly quoted from around $1,500 into five figures, and the invoice is the smaller half of the cost. The other half is calendar. A brief, a kickoff, a script round, a storyboard round, an animation round, and a revision process with a meter running. Weeks, and almost none of those weeks are animation — they go into deciding, not making.

So the honest question for most products was never "is a video worth eighty percent." It was "is a video worth six weeks and several thousand dollars, on a page I might rewrite next quarter."

For a company-defining brand film, that maths works. For a landing page it usually doesn't, and I think that is the actual reason most SaaS landing pages still have no video on them. Not because anyone concluded video doesn't help. Because the price of finding out was higher than the thing was worth.

That gap is what PromoHyper was built to close, and it's mine, so discount it accordingly. You give it your product URL. It reads your site, writes three script directions from what it finds — three angles on the same product, not three rewordings of one — voices the one you pick, and designs every frame around what that script says, in your brand's colours and using your own product screens. It is finished in about fifteen minutes. The first one is free, watermarked, one per account; plans start at $39 after that.

What that changes isn't the conversion rate. It's the size of the bet. At fifteen minutes and $39 you don't need a forecast to justify trying — which is convenient, because as the last two thousand words have shown, nobody can honestly give you one.

One warning that follows from all of the above, and it cuts against me rather than for me. When a new video costs fifteen minutes, the temptation is to swap it every fortnight, and a page that changes every fortnight tells you nothing about any version of itself. Cheap and fast removes the excuse for not trying. It does not remove the need to leave one version up long enough to learn something. Pick one, sit on your hands, write down the date.

How to decide

Decide the ceiling before you decide the budget. If the video is a company-defining brand moment with a shelf life of years, the expensive tier exists for good reasons and I'd hire a studio. If it's a launch page, a landing page or an ad — which is most of them — spend the smallest amount that gets you something real, and put the saved money into traffic to judge it with.

Then when the first version turns out wrong, which happens constantly, note what you'd change about the sentence before you note what you'd change about the animation. A bigger budget is something you can approve in an afternoon. A clearer first line is something you have to write.