Video Marketing

How to A/B Test SaaS Video Ads Without Paying for Two Productions

Lock the core. Swap the edges: A/B testing SaaS video ads with a three-zone storyboard

Why it matters

Two different ads teach you nothing

Here's how most SaaS teams A/B test SaaS video ads. They make two completely different videos and see which one wins.

But when everything changes, you learn nothing. You can't tell if it was the hook, the demo or the ending that won. And you paid for two productions to find out.

Part of the adTwo different videosOne change per test
Hook, the first 6 secondsDifferentSwapped: the only change in test one
Core, the product demoDifferentLocked, never touched
CTA, the last 10 secondsDifferentKept the same until test two
What you learnWhich whole video won, not whyWhich hook, then which ending, won
What you pay forTwo full productionsOne production, plus a new scene and a re-edit

The short answer: split every ad into three zones: the hook, the core and the call to action. Lock the core. Test the hook first, and pick the metric before you test. Then test the ending, one change at a time. Run each test as a real split test, so nobody sees both versions. Four of the sections below include an ad we made for a software or B2B client.

Planning video ads you can test?

Talk to the Advids team about a modular storyboard: one core demo, with hooks and endings you can swap.

Let’s talk

Three zones

Split every ad into hook, core and CTA

Lock the expensive middle. Swap the cheap edges.

Start with the storyboard. Every ad gets three zones.

ZoneLengthIn a test
HookThe first 6 secondsSwappable: a new scene and a re-edit
CoreThe 2D or 3D product demonstrationLocked: built once, never touched
CTAThe last 10 secondsSwappable: a new end card and a re-edit

The core is the most expensive part to animate, and it's the part that explains your product. So you build it once and never touch it during a test. The hook and the CTA are short, so a new version is a new scene and a re-edit, not a new production. Think of it like a sandwich: keep the filling, try different bread.

Here's an example we wrote for an invoicing tool, not client work. Hook A opens on a "payment overdue" email. Hook B opens on a cash-flow chart dropping off a cliff. Hook C opens on a CFO asking, "Where's the money?" All three cut into the same forty-second demo.

UrSpayce AssetX: three zones you can see

UrSpayce's explainer for AssetX opens on Kevin, "A Dedicated Operations Manager!", at a desk buried in boxes, chat alerts and a flipping calendar. Then the drawings stop and real AssetX screens take over: ownership, status, maintenance alerts and barcode labels. It closes on "#HappyWorking", then "Get Started" and the UrSpayce web address. See how we made it.

What works here:

  • The problem lives in 2D drawings and the product in real screens, so each block has its own look.
  • A new opening could replace Kevin's scenes without touching a single product screen.

The move: Give the hook, the core and the ending their own look, so each zone can be cut out and replaced on its own.

So which zone do you test first? The hook.

Hook first

Test the hook first, and pick the metric before you test

If nobody watches past the opening, the demo and the CTA never get a chance.

Test the hook first. If nobody watches past second six, your demo and your CTA never get a chance.

Google's Unskippable Labs ran a test that worked this way. It made three versions of a YouTube ad for Maruti Suzuki, the Indian carmaker. Same ad, three different first five seconds: one opened with a brand watermark, one with an intriguing message, and the original with no branding at all. It was a 2017 consumer test, not B2B lead generation, but it's the clearest published test of different openings on the same ad body.

Opening (first 5 seconds)View-through rateAd recall lift
Brand watermark30%40%
Message-led29%75%
Original, no branding27%71%

The branded opening kept the most viewers. The message-led opening was remembered far better. So the winning hook depended on the metric. Pick what you're testing for before you test. That's the first of most A/B testing best practices, and it matters more when each variant costs a scene.

Video ad producer Fraser Cottrell says the same thing in plainer words. Whatever else you do, test a couple of different hooks.

HID TripTick: a self-contained problem opener

HID's explainer for its TripTick readers opens on a crowded bus stop. On board, a phone ticket fails at the validator, which shows a red cross, while "!" and "?" rise over the queue. Only then does the HID logo appear, followed by the reader range and the ticket types it reads. See how we made it.

What works here:

  • The failed scan is a complete scene on its own, before any product appears.
  • Another opener, such as a long queue or a missed bus, could replace it without touching the reader demo.

The move: Write your hook as a scene that ends where the product begins. Then a second and third hook can plug into the same cut point.

Once a hook wins, lock it too, and move to the end.

CTA test

Then test the ending, one change at a time

For B2B, the CTA test is usually about commitment.

"Book a demo" asks for a lot. "Watch a two-minute tour" asks for less. Same video, two endings, and the results show which one fills your pipeline with qualified people. The pair below is an example we wrote, not client data.

End cardThe askCommitment
A"Book a demo"High: a meeting with sales
B"Watch a two-minute tour"Low: two more minutes of video

Keep one change per test. Hook first, CTA second. If you change both at once, you're back to two different videos. For wording ideas, these call to action examples show how much the ask itself can shift. Most software ad examples end on a single card, which makes this the cheapest test you can run.

Talentia HCM: a high-commitment ending

Talentia's product video follows the real Talentia HCM interface from a list of reports to an appraisal report, a team dashboard, and that dashboard shared in a discussion where a colleague replies. It ends on one card: "To learn more about our solutions." and a "BOOK A DEMO" button. See how we made it.

What works here:

  • The ending is a separate card, so it can change without touching the walkthrough.
  • "Book a demo" is the high-commitment ask you'd test against a lighter one.

The move: Keep your ending on its own card. Then a test of two asks is one new card and a re-export.

Now, here's where budgets get wasted: how the test itself is set up.

Split test

Split the audience, not the ad set

Nobody should see both versions.

Most teams drop both versions into the same ad set and wait. Meta says not to A/B test that way. Its A/B test tool shows each version to a separate part of your audience, so nobody sees both. It also recommends the same budget for each version, so the comparison is fair.

Both versions in one ad setMeta's A/B test tool
Who sees whatAudiences can overlapEach version goes to a separate group
BudgetThe platform shifts spendThe same budget for each version
ResultUnreliable, in Meta's wordsA fair comparison

There's a second reason. Alessandro Cutri, who runs Meta ads for SaaS companies, has found that Meta now groups ads that look alike. That's his reading of how the system behaves, not Meta's wording. So two hook versions in one ad set may not even be treated as two ads. It's one more reason the hook zone should look clearly different in each variant, which is easier when your motion graphics ads are built in zones.

Papershift Plan: an opener that looks nothing like the body

Papershift's sales demo opens on one 2D frame that labels six features around a roster and a phone: "Shift planning", "Time tracking", "Absence planning", "Real-Time tracking", "Weekly hours" and "Employee data". Then live-action shots put the Papershift roster on an iMac, laptops and a phone, and it closes on "That's Papershift". See how we made it.

What works here:

  • The 2D opener and the live-action body look completely different.
  • Swap the opener for another and the two versions still read as different ads.

The move: Make each hook variant visibly different from the others, not a new font on the same frame, so the platform treats them as two ads.

Lock the core, swap the edges, split the audience. Here's where to start.

Where to start

Where to start

The short answer: test the zone where your ad is losing people.

If your problem isTest this zoneDo thisExample to copy
Every test means a new productionThree zonesLock the core demo, swap only the edgesUrSpayce AssetX
People drop off in the first secondsHook firstThree hooks on one body; pick the metric firstHID TripTick
People watch but don't actCTA testTwo end cards, high and low commitmentTalentia HCM
Results don't hold upSplit testSeparate audiences, same budgetPapershift Plan

If you are briefing a new ad, a SaaS promo video built in zones from day one is the cheapest to test. The same applies to short video ads. We've made video for teams such as UrSpayce, HID, Talentia and Papershift. See the software teams we've made video for.

FAQ

Questions about A/B testing video ads

What should I A/B test first in a SaaS video ad?
The hook, the first six seconds. If viewers leave before the demo, nothing after it matters. Lock the rest of the ad and test two or three openings.
How long should the hook zone of a video ad be?
About six seconds. Google's Maruti Suzuki test varied only the first five. Keep the hook short enough that a new version is one new scene.
Why shouldn't I test two completely different videos?
Because when everything changes, you can't tell what won: the hook, the demo or the ending. You also pay for two full productions to learn one thing.
How do I run a fair video ad split test on Meta?
Use Meta's A/B test tool. It shows each version to a separate audience, so nobody sees both, and gives each version the same budget for a fair comparison.
Which metric should decide the winning hook?
Decide before you test. In Google's Maruti Suzuki test, the branded opening held the most viewers, while the message-led opening won ad recall by far.
How do I test CTAs without making a new video?
Keep the ending on its own end card in a layered master file. Then a second ask, such as a product tour instead of a demo, is a new card.

About the author

About Jai Ghosh

Video Producer at Advids · LinkedIn

Jai Ghosh is a Video Producer at Advids, where he has planned and produced videos for B2B software and technology companies for 7 years. He works on product demos, explainer videos and brand videos, from the first script to the final cut, for clients from startups to enterprises.

Before Advids, Jai spent over 11 years in freelance media consulting, after starting out in digital journalism. He writes the Advids blog's breakdowns of product and demo videos, sorting real examples by the question each one answers for a buyer.

  • Focus: SaaS product demos, explainer videos and brand videos
  • Experience: 7 years at Advids, over 11 years in media consulting
  • Background: digital journalism

Connect with Jai Ghosh on LinkedIn

Sources

Sources