First frame vs last frame testing compares two very specific, literal frames of a video ad: the single still image a viewer sees before any motion starts, and the single still image left on screen when the video ends. They are not the same as hook testing or CTA testing — they are about what the eye lands on at the two moments a viewer is most likely to actually look, rather than scroll or swipe away. Get the first frame wrong and the ad never gets watched; get the last frame wrong and the ad gets watched but forgotten.
What first frame and last frame testing actually mean
A video ad is made of thousands of frames, but only two get seen by every single viewer whether they watch the whole thing or not: the first and the last. The first frame is the still image shown the instant the ad loads, before any cut, zoom or voiceover kicks in. On some feeds this is also used as the static thumbnail shown while the video buffers or before autoplay triggers. The last frame is whatever is on screen the moment the video stops — the frame a viewer's eye rests on if they don't immediately scroll to the next post, and the frame that shows up in a screenshot.
This is different from hook testing, which is about the first few seconds of message and motion, and different from testing the call to action copy, which is about what's said or written, not which exact frame is frozen. If you've already run tests on hooks versus full video structure, frame testing is one level more granular — it isolates a single still image rather than a whole opening sequence.
Why the first frame matters for stopping the scroll
The first frame has one job: make someone stop moving their thumb for long enough to see what happens next. It has to do this with zero context — no voiceover has played, no caption has appeared yet, and the viewer may not even know what category of product this is.
Variables worth testing in a first frame:
- Close-up versus wide shot — a tight crop of a product detail often reads as more intriguing than a full product shot
- Face versus no face — a human expression can out-perform a clean product shot, or distract from it, depending on the category
- Text on frame one versus a clean image with the text arriving a beat later
- Motion-implying stills — a frame that looks mid-action (a hand reaching, a liquid pouring) versus a static, posed shot
- High contrast background versus a background that matches the feed's visual noise
Example: a vitamin gummy brand could test a first frame of a hand holding the bottle against a plain kitchen counter against a first frame that's an extreme close-up of the gummy's texture with the words "tastes like candy" already on screen. Both are valid hypotheses — one leads with trust and familiarity, the other leads with curiosity and a slight promise. Only a side-by-side test tells you which one earns the next three seconds.
Why the last frame matters for recall and retargeting
The last frame carries a different job: make sure whatever a viewer remembers, screenshots, or sees again later (in a retargeting impression, in a feed replay, in a shared clip) still does work for the brand. It's also often the frame seen by anyone who scrolls away a half-second before the video technically ends, which happens more than most people assume.
Variables worth testing in a last frame:
- Packshot with offer text versus a lifestyle shot with no text at all
- Logo visible and legible versus logo absent
- A frozen CTA button or arrow versus a plain final image with only a caption
- Price or promo code baked into the frame versus left out, if the on-screen captions elsewhere already carry it
- A frame that matches the first frame (bookend structure) versus a frame that deliberately contrasts with it
Example: a phone stand brand might end one version on a clean packshot with the price overlaid, and another on a lifestyle shot of the stand holding a phone on a desk with no text at all. The first is built for someone who's ready to act now; the second is built for recall if they're not ready yet. Testing both tells you which stage most of your audience is actually in.
First frame vs last frame testing: a side-by-side comparison
| First frame | Last frame | |
|---|---|---|
| Primary job | Stop the scroll before any message lands | Leave something that sticks after the message has landed |
| Seen by | Every viewer, including those who never press play | Viewers who watch to the end, plus anyone who screenshots or replays |
| Also doubles as | Static thumbnail on some feeds | Image left behind when sound/motion stops |
| Good variables to test | Shot distance, face vs. product, motion cue, text timing | Packshot vs. lifestyle, offer visibility, logo placement, CTA frame |
| Best measured by | Early view-through or 3-second hold rate | Completion rate, outbound clicks, and down-funnel recall |
| Risk if ignored | Strong ad never gets a chance to be watched | Good ad gets watched and forgotten, or screenshot badly |
How to run a clean first frame vs last frame test
Treat this like any other creative test: isolate one variable, keep everything else constant, and measure the metric that frame is responsible for.
- Pick one frame to test at a time. If you change both the first and the last frame in the same version, you won't know which one caused the difference.
- Keep the middle of the video, the voiceover, the music and the CTA copy identical across both versions, so the only variable is the frame you're testing.
- Where the platform allows a thumbnail override separate from the video file itself, set it to match your intended first frame — otherwise the platform may pick its own frame and your test is comparing the wrong thing.
- Run both versions as a genuine split rather than one after another, since feed conditions and audience mood shift over time. If you're unsure which setup fits your situation, this breakdown of split testing versus sequential testing is worth reading first.
- Give the test a fixed amount of spend or time before calling a winner, rather than stopping as soon as one version looks ahead — early leads can flip.
- Log the right metric for the right frame: early hold or view-through rate for first-frame tests, and completion rate plus any outbound action (clicks, add-to-carts) for last-frame tests.
Common mistakes when testing frames
- Changing the first frame and the hook message at the same time — you end up testing two things and learning about neither
- Judging a first-frame test purely on total views, which rewards the algorithm's delivery choices more than the frame itself
- Forgetting that both frames need to work with sound off, since ads are frequently seen on mute by design — a last frame that relies on a spoken price with no text equivalent loses that information for silent viewers
- Leaving the platform's auto-selected thumbnail in place while testing a custom first frame, which quietly cancels out the test
- Ending every version on the same generic logo card instead of testing whether a product or lifestyle last frame performs differently
Where this fits in a wider testing plan
Frame testing is a refinement step, not a starting point. If you haven't yet nailed down which bigger variables to test first — hook angle, offer, format, creator type — it's worth working through creative testing variables to test first before narrowing in on single frames. First and last frame tests tend to pay off most once a winning hook and structure are already in place and you're squeezing extra performance out of a proven ad, similar in spirit to how one hook vs multiple hooks testing works for the opening seconds.
How FrameNotion fits into frame-level testing
FrameNotion builds each 30-second vertical ad from scratch for a specific product, and its AI checks its own rendered frames as part of making the video — including the opening and closing images, not just the message in between. Every ad is also built with word-by-word captions and designed to work on mute, so a first or last frame carries meaning on its own rather than depending on a voiceover a silent viewer never hears.
Because text and colors can be edited after an ad is finished — through change requests rather than rebuilding the ad — it's straightforward to request a variant with a different opening shot or a different closing frame to run your own test, without starting over. Pro and Agency plans also support A/B hook variants for testing the opening more broadly. Every ad comes out in 9:16, 4:5, 1:1 and 16:9 automatically, so whichever frame wins can be reused across every feed you advertise on. The first ad on any new account is free, and plans start from a Starter tier for brands that want to run a handful of frame tests a month — see pricing or browse example ads to see how first and last frames are handled in finished videos.
FAQ
Frequently asked questions
Is first frame testing the same as hook testing?+
No. A hook is the message and motion in the opening few seconds; the first frame is one literal still image shown before any motion begins. You can keep the same hook message and still test two different opening images, or change the hook while keeping the same first frame.
Which matters more for a brand-new product launch?+
First frame usually matters more early on, because a new audience has no reason to stop scrolling yet. Last frame becomes more important once you're running retargeting or once completion rate shows most viewers already watch to the end.
Can I test the first frame and last frame at the same time?+
You can run both tests in the same period, but as separate test pairs — one pair that only changes the first frame, another pair that only changes the last frame. Changing both in a single version at once makes it impossible to tell which frame caused any difference.
Does the last frame still matter if most of my audience watches with sound on?+
Yes, because retargeting impressions, replays, and screenshots often happen without the original sound, and some viewers scroll away a moment before the video technically ends. Designing the last frame to carry its own meaning protects performance either way.
How do I control which frame shows as the static thumbnail?+
Many ad platforms let you choose a custom thumbnail separately from the video file, often by selecting a frame or uploading an image. If you skip this step, the platform may auto-select a frame that has nothing to do with the first frame you intended to test.