The numbers stopped being a trend line and became a floor. Most video is now watched on a phone, most of that phone-watching is short and vertical, and the completion and engagement data all point the same way: a nine-by-sixteen clip that earns attention in the first three seconds outperforms almost everything else in the feed. VEED's widely-cited video-marketing analytics roundup pulls the platform and behavioral data into one place, and read together it settles an argument creators used to have — mobile-first short-form is no longer one format among several, it is the primary distribution surface, and everything else adapts to it. That reframes the production question. It is no longer "should we make short-form for mobile," it is "how do we produce enough on-brand vertical video, fast enough, to feed a distribution surface that rewards volume and punishes anything shot for a different screen." This guide reads the analytics honestly — what is solid, what is soft, and what the ranges actually mean — then turns them into concrete production decisions: the specs the data implies, the hook discipline the retention numbers demand, and how to hit the cadence a vertical-first audience needs without your output collapsing into interchangeable AI filler.
For years the case for mobile short-form video was a projection — a trend line pointing up that you were told to get ahead of. In 2026 it stopped being a projection and became the baseline everything else is measured against. The analytics that video tools and platforms publish now agree on the shape of the thing even when they disagree on the second decimal place: most video is watched on a phone, most of that watching is short and vertical, and the engagement and completion data reward a nine-by-sixteen clip that earns attention immediately over almost anything else in the feed. VEED's widely-cited [video-marketing statistics roundup](https://www.veed.io/learn/video-marketing-statistics) is a useful single place to see the platform and behavioral figures pulled together, and read as a whole it settles an old argument: mobile-first short-form is not one format among several, it is the primary distribution surface, and long-form, horizontal, and desktop now adapt to it rather than the other way around.
That settled question opens a harder one. If the mobile vertical feed is where distribution happens, the constraint is no longer strategy — everyone already knows to make short-form — it is production. A surface that rewards cadence and punishes anything shot for a different screen needs a steady supply of on-brand vertical video, and hand-producing that at volume across many platforms is where most operations stall. This guide does two things. First it reads the analytics honestly — separating what is solid from what is soft, and explaining what the wide ranges actually mean so you do not build a plan on a shaky number. Then it turns the durable findings into concrete production decisions: the specs the data implies, the hook discipline the retention numbers demand, and how to hit a vertical-first cadence without your output flattening into the interchangeable AI filler covered in [why AI content stopped working and your metrics changed](/guides/ai-content-didnt-stop-working-your-metrics-did). It is the production companion to the [glossary definition of short-form video](/glossary/short-form-video) and to [how short-form video features evolved in 2026](/guides/short-form-video-music-captions-evolution).
Before turning numbers into decisions, it is worth being clear-eyed about which numbers to trust. Aggregated statistics roundups — VEED's included — are genuinely useful for seeing the landscape, but they blend sources with different methodologies, dates, and definitions, so the honest move is to treat the direction as solid and the exact figure as a range. A stat like "three-quarters of video is watched on mobile" is directionally rock-solid and repeated across independent sources; whether the precise number is 72 or 78 percent depends on whose panel and what counts as a "view," and no production decision should hinge on that gap. Use the aggregates to orient, then trust your own platform analytics for anything you are optimizing.
Four findings show up everywhere, survive scrutiny, and should anchor your planning. One: the large majority of video views happen on mobile — figures around three-quarters are the common citation, and video is the dominant share of mobile data traffic, so the phone is unambiguously the default video screen. Two: vertical video completes at a higher rate than horizontal on those mobile feeds, because a landscape clip in a vertical feed wastes most of the screen and reads as imported. Three: short-form is the in-feed content type viewers most often rate as the most engaging — survey after survey lands near two-thirds naming it their preferred format. Four: attention is front-loaded to an extreme degree, with a large share of mobile viewers — commonly cited around two-thirds — swiping past within roughly the first three seconds. Those four are the load-bearing facts; everything below builds on them.
Other figures are useful but should be held loosely. Exact optimal-length windows — VEED's roundup notes TikTok clips performing best around 21-to-34 seconds, and a majority finishing videos under a minute — are real signals but vary by platform, niche, and the specific audience, so treat them as starting points, not laws. Platform market-share splits (short-form views concentrated across TikTok, Reels, and Shorts) shift quarter to quarter. And any single "X% of views are vertical" or "X% completion" number is methodology-dependent enough that you should anchor to the pattern, not the digit. The mistake is quoting a soft number as if it were a hard one; the discipline is knowing which is which.
Turn the solid findings into production reality and they stop being interesting facts and become a spec sheet. The most-video-is-mobile finding means the default aspect ratio is vertical nine-by-sixteen, full stop — not "we also make a vertical cut," but vertical as the thing you shoot and design for, with any horizontal version treated as the adaptation. A clip framed for a landscape player and reposted to a vertical feed loses the screen and the completion rate that comes with it, which is exactly what the vertical-versus-horizontal completion gap measures. If your capture, your framing, and your on-screen text are not built for a phone held upright, you are producing for a screen your audience is no longer using.
The vertical-completion finding pushes further than aspect ratio into legibility. A video that fills a phone screen is watched with the sound sometimes off, one-handed, in a scroll — so captions are not an accessibility nicety, they are a completion lever, and text has to be large enough and placed high enough to survive the platform UI overlapping the bottom of the frame. The features that used to be optional polish — burned-in captions, bold on-screen text, localized subtitle tracks — became reach levers precisely because the viewing context is mobile-first, a shift traced in [how music, captions, and localization became reach levers](/guides/short-form-video-music-captions-evolution). Designing for a muted, vertical, thumb-scrolled screen is not a constraint on the format; it is the format.
No finding changes production more than the swipe-away number. If a large share of viewers decide within three seconds, then the opening frames do almost all the work of earning a view, and the practical consequence is blunt: the hook is not the first line of your script, it is the first second of the video — the visual, the motion, the on-screen text that appear before anyone has chosen to stay. A clip that opens with a logo animation, a slow establishing shot, or a "hey guys, welcome back" spends its most valuable three seconds on the thing the audience is fastest to skip. On an infinite feed where the next video is one flick away, a weak open is not a slow start you recover from; it is the whole result.
This reframes where effort goes. The disproportionate payoff is in the first second and the caption hook, so that is where a mobile-first operation concentrates — leading with the payoff, opening on motion or a striking frame, putting the promise in on-screen text immediately. The mechanics of writing openers that survive the swipe are laid out in [how to write viral hooks for short-form video](/how-to/write-viral-hooks); the analytic point is why they matter so much more than they used to. And it is why testing variants of the first three seconds — different opening frames, different text hooks over the same body — is the highest-leverage experiment in short-form, a discipline that connects directly to [A/B testing social creatives](/guides/ab-testing-social-creatives). When the data says the open decides the outcome, the open is where the iteration belongs.
Here is the tension the numbers set up and rarely resolve. A vertical-first audience rewards consistent cadence — the feeds are volume machines, and a channel that posts several strong shorts a week compounds while an occasional one stalls. But the same analytics that demand that cadence also raise the bar for each clip: vertical-native framing, burned-in captions, a hook engineered for the first second, sized and formatted correctly for each mobile surface. Producing that by hand — film, edit, caption, resize per platform, schedule — does not scale to the frequency the distribution surface wants, and it does not scale across the several platforms a serious operation publishes to. The strategy is settled and the production is the bottleneck, which is the exact predicament the [YouTube Shorts versus long-form strategy guide](/guides/youtube-shorts-vs-long-form-strategy) runs into from the format-choice side.
The failure modes at this bottleneck are two, and they are opposite. One is undershooting: producing too little because hand-made vertical video is slow, and losing the cadence the feed rewards. The other is overshooting into sameness: cranking volume by leaning on a single template until every clip looks like every other AI-generated short, which the audience has learned to scroll past — the flattening documented in [the AI design aesthetic](/guides/the-ai-design-aesthetic). Solving the mobile-first production problem means clearing both: enough throughput to feed the cadence, and enough brand distinctiveness that the throughput does not read as filler. That is a systems problem, not a shooting-technique problem, and it is where a generation-and-distribution engine earns its place.
[Kompozy](/) is built for exactly the gap the analytics expose: the distance between "we know the audience is mobile-first short-form" and "we produce enough on-brand vertical video to feed that surface at cadence." It attacks the throughput side directly by generating vertical-native video rather than asking you to shoot and cut it. [Persona Shorts](/glossary/persona-shorts) render a talking-head avatar clip with captions burned in; Clipped Shorts turn one long video into multiple vertical cuts; Listicle and Marketing video assemble nine-by-sixteen clips from a prompt — all produced at the aspect ratio and with the caption legibility the completion data demands, not adapted to it afterward. That is the difference between an engine that makes mobile-first video and an editor that helps you reformat video shot for another screen.
The distribution side is where the cadence problem gets solved without multiplying the work. One finished short fans from a single queue to the mobile feeds that matter — Instagram, TikTok, YouTube, Facebook, and more — in each platform's correct format, with [autopilot](/glossary/autopilot) scheduling behind a per-post review gate, so hitting several posts a week across every surface is one pass instead of a per-platform re-export marathon. The multi-account, multi-platform version of that logistics problem is the whole subject of [managing multiple social accounts at scale](/guides/managing-multiple-social-media-accounts-at-scale); Kompozy's answer is to make the fanout a byproduct of generation rather than a second job. And because the same [Persona Brief](/glossary/persona-brief) governs voice, Gemini face-lock holds one identity across every clip, and HyperFrames renders pixel-exact brand styling, the volume does not decay into the interchangeable-filler failure mode — the hundredth short still looks like you, which is the half of the production problem raw output tools ignore.
The honest boundary: an engine produces the vertical video and ships it, but it does not invent the idea or guarantee the hook. The three-second open still has to be genuinely good, the concept still has to be worth watching, and Kompozy gives you fast on-brand production and correct mobile-first distribution — not a substitute for having something to say. Used well, that is the point: it removes the throughput and formatting tax so your effort goes where the analytics say it counts, into the hook and the idea, instead of into resizing and re-uploading the same clip nine times.
The analytics are no longer arguing a case; they are describing a settled reality. Most video is watched on a phone, vertical short-form is the format that screen rewards, short clips win on engagement, and the first three seconds decide the outcome — those findings are solid enough to build on, whatever the exact percentages turn out to be in any one source. What that means for a creator is not "start making short-form," which everyone already knows, but "produce enough on-brand vertical video, fast enough, to feed a distribution surface that runs on cadence and punishes anything shot for a different screen." The strategy is the easy part now. The production is the real work — and the operations that win the mobile-first feed are the ones that solve throughput and brand consistency together, so their volume reads as a recognizable channel rather than more filler to scroll past.
The consistent finding across analytics roundups is that the large majority of video views now happen on mobile devices — figures around three-quarters of all video views are widely reported, and video makes up the bulk of mobile data traffic. VEED's roundup adds platform-level detail, including that the overwhelming share of video views on X happen on mobile. The exact headline number varies by source and definition, but the direction is unambiguous: the phone is the default video screen, and vertical short-form is the format that screen rewards.
For mobile-native feeds, yes, and the completion data is the clearest evidence. Vertical clips fill the phone screen and consistently post higher completion rates than horizontal ones, because a landscape video shown in a vertical feed wastes most of the screen and reads as imported rather than native. Short-form specifically wins on engagement too — surveys repeatedly find short clips are the in-feed format viewers rate most engaging. Horizontal and long-form still matter for other surfaces; they are just no longer the shape the primary mobile feed is built around.
It is the finding that a large share of mobile viewers — commonly cited around two-thirds — will swipe past a video if it does not engage them within roughly the first three seconds. On an infinite vertical feed, the next clip is always one flick away, so the opening frames do almost all the work of earning a view. Practically it means the hook is not the first line of your script, it is the first second of the video: the visual, the motion, and the on-screen text that appear before anyone has decided to stay.
Shorter than instinct suggests, with the exact window varying by platform. Completion rates climb as videos get shorter, and platform data points to sub-minute clips as the sweet spot — VEED's roundup notes TikTok clips performing best in roughly the 21-to-34-second range, and a majority of viewers finishing videos under a minute. The safe default is to make the shortest version that fully delivers the idea, then let each platform's own analytics tell you where your specific audience's attention drops.
The bottleneck is production throughput, not strategy. A vertical-first audience rewards consistent cadence, but filming, editing, captioning, and sizing a nine-by-sixteen clip for every platform by hand does not scale to several posts a week across many surfaces. The practical answer is to generate vertical-native video and fan it — an engine like Kompozy produces avatar shorts, clipped verticals, and listicle video already sized nine-by-sixteen with captions burned in, then publishes the platform-correct version to each mobile feed from one queue.
The analytics converge on one conclusion: mobile is the default video screen and vertical short-form is the format it rewards. Roughly three-quarters of video views now happen on phones, vertical clips post higher completion rates than horizontal, short-form is the in-feed format viewers rate most engaging, and about two-thirds of mobile viewers swipe away within the first three seconds. VEED's video-marketing analytics roundup aggregates the platform and behavioral data behind this. For creators it reframes production from "should we make mobile short-form" to "how do we produce enough on-brand vertical video, fast, to feed a surface that rewards volume."
Get started → · ← All guides · Compare Kompozy vs other tools