"Keep it short" is the advice everyone gives about video ad length. It's not wrong, but it's not specific enough to act on, and most of what's published as a "15-second rule" is repeated folklore, not a number anyone tested against real ad accounts.
Where the 15-second number actually comes from
Short-form video ads on Meta and TikTok get judged on two different things that often get conflated: whether someone watches, and whether someone remembers. Length affects both differently. A shorter ad has a structurally higher completion rate almost by definition, less time for someone to drop off. But completion rate isn't the same as attention quality, and attention quality is what actually drives recall and, eventually, conversion.
The honest answer is that there's no single correct length, there's a length that matches the job the ad is doing. A cold-prospecting ad trying to earn a scroll-stopping first impression behaves differently from a retargeting ad reinforcing a decision someone's already leaning toward.
What actually determines whether a short ad works
The first 2 to 3 seconds decide almost everything. On both platforms, the algorithm and the viewer make a keep-watching decision almost immediately. An ad that doesn't establish a reason to stay within that window loses most of its remaining runtime's value regardless of how good the rest of the creative is. This is why hook quality matters more than total length in isolation, a strong hook justifies 15, 20, even 30 seconds. A weak hook makes even 6 seconds feel long.
Completion rate and attention quality aren't the same metric, and conflating them misleads creative decisions. A 6-second ad with a 90% completion rate isn't automatically outperforming a 20-second ad with a 60% completion rate, since the shorter ad's ceiling on message delivery is structurally lower. What matters is whether the viewer retained and acted on the message, not just whether they sat through it. Platform-reported completion rate is a proxy, not the actual thing you're trying to optimize.
Different placements reward different lengths. Feed placements (Reels, TikTok's main feed) reward a tight, front-loaded hook since competition for attention is highest there. Story and in-stream placements, where the viewer has already committed some attention before the ad appears, tolerate a slightly longer build before the payoff.
A practical length framework by funnel stage
Cold prospecting: 6 to 15 seconds. The job is earning attention and communicating one single idea clearly enough to justify a click or a follow-up impression later. Anything longer risks losing the viewer before the message lands, and cold audiences have the least patience of any segment.
Warm retargeting: 15 to 30 seconds. The viewer already has some context, so there's room to build a slightly more complete case, address an objection, or show the product in more detail, without needing to win attention from zero.
Consideration and education content (often run as an ad but built for a warmer or existing audience): 30 seconds or more. This is where genuine product demonstration or a founder story earns its runtime, but it's rarely the right format for a cold-audience prospecting campaign.
Why "attentive seconds" matters more than raw watch time
A related but distinct concept worth understanding: attentive seconds versus raw viewability. Viewability just measures whether the ad was technically on-screen. Attentive seconds tries to capture whether the viewer was actually looking at it, sound on, engaged, not just scrolling past with the ad technically visible for a fraction of a second. Meta and TikTok both increasingly weight engagement signals beyond simple view-through in how they optimize delivery, which means an ad that holds genuine attention for 8 real seconds can outperform one that's technically viewable for 15 seconds but loses the viewer's actual focus after 2.
This is also why creative testing needs to look past completion rate alone. Thumb-stop rate (did they pause scrolling), 3-second view rate, and hold rate through the first hook segment are earlier, more diagnostic signals than full completion, and they tell you faster whether a concept is working before you've spent enough to get a clean read on downstream conversion.
What this means for testing new creative
Don't test length as an isolated variable, since a 6-second cut of a 20-second concept and a 6-second concept built from scratch to work in 6 seconds are not the same asset, and comparing them tells you more about editing quality than about length itself. Test genuinely different concepts built natively for their target length, the same discipline covered in the creative fatigue post on this blog about needing real new concepts, not superficial variations, to get a clean read.
Build the hook first, decide length second. Write and shoot the first 3 seconds as if that's the whole ad, since for a meaningful share of the audience, it effectively is. Everything after that point should feel earned by the hook, not padding needed to hit a target runtime.
FAQ
Is there an official 15-second rule from Meta or TikTok? Neither platform publishes a single hard rule. Both platforms' own creative guidance leans toward shorter being generally safer for cold prospecting, but frames it as a starting point, not a strict requirement, and both explicitly support longer formats for other funnel stages.
Does a shorter ad always cost less to run? Not directly, CPMs are driven by auction dynamics and audience competition, not creative length. Shorter ads can indirectly lower cost per result if they drive better engagement signals, which the algorithm rewards with more efficient delivery.
Should every ad in an account be short-form now? No. The right length depends on funnel stage and objective. A prospecting-heavy account skewing short is reasonable; a retargeting or nurture sequence built entirely from 6-second cuts is likely leaving persuasion on the table.
How do I know if my hook is actually working? Watch 3-second view rate and thumb-stop rate relative to your account's baseline, not an external benchmark. A concept underperforming your own historical average on these early signals is worth cutting before it burns meaningful spend on a full test.
Does this framework apply the same way to TikTok and Meta? The underlying principle, hook quality determining most of an ad's value, holds on both. The specific placement dynamics and typical audience patience differ slightly, TikTok's native-feeling content generally tolerates a rawer, less polished opening than Meta's feed, which is worth factoring into how a hook is shot, not just how long it runs.
Want your creative tested against the hooks and lengths that actually move performance in your account, not a generic rule? Talk to us at growth@adtitudemedia.com.