Video Advertising

Formats and What They Buy

Skippable in-stream — viewer can skip after five seconds. You typically pay only when they watch a meaningful portion. Low risk, and the five-second window is the entire creative challenge.

Non-skippable in-stream — usually fifteen to twenty seconds, guaranteed full view, higher CPM. Suitable for short single-message brand work.

Bumper — six seconds, non-skippable, built for frequency and recall rather than explanation.

In-feed and outstream — appears in feeds and content pages, usually muted by default, requiring the creative to work without sound.

Connected TV — full-screen, non-skippable, lean-back viewing, no click available.

Creative Structure

For skippable formats the opening seconds carry disproportionate weight. Lead with the hook, not the brand logo. Showing the product or the problem immediately outperforms building to a reveal that most viewers never reach.

Brand early and repeatedly. A significant share of viewers who skip will still have seen the first five seconds — make those seconds carry the brand.

Design for sound-off in feed placements. Captions are not an accessibility afterthought; they are the primary delivery mechanism for most social video.

Platform Differences

YouTube viewers arrive with intent, often searching. TikTok and Reels viewers are in passive discovery and reject anything that feels like advertising. CTV viewers are captive but cannot click.

Recutting the same asset for each placement outperforms running one master edit everywhere, and costs little once the footage exists.

Measurement

View counts are a delivery metric with different definitions per platform — comparing them across platforms is meaningless.

More useful: view-through rate, watch-time curves showing where viewers drop, brand lift studies for awareness objectives, and search lift measuring whether video exposure increased branded queries.

For direct response video, measure conversions with a realistic lookback window. Attributing only same-session conversions will drastically understate performance.

Making One Asset Work Across Placements

Producing a bespoke cut per placement is the correct answer and rarely the affordable one. The workable compromise is to shoot with the constraints in mind so one production yields several usable cuts.

Frame for the crop. Shoot 16:9 with the action and any on-screen text inside a 9:16 safe area, and a vertical cut becomes a crop rather than a reshoot. Getting this wrong is the single most common reason a campaign has no vertical asset.

Structure for the skip. Build the opening so the message survives being cut at five or six seconds, which means brand and proposition early, not a slow build to a reveal.

Design for sound-off. Burned-in captions, not a subtitle track, because many placements ignore the track. If the video only makes sense with audio, a large share of the impressions are wasted.

Produce the short cut deliberately rather than trimming the long one. A six-second asset edited from a thirty-second narrative is almost always a fragment of an argument rather than an argument.

What the Standard Metrics Do Not Tell You

Video reporting is dense and most of it describes the placement rather than the advertising.

Completion rate is largely a property of skippability. A high rate on non-skippable inventory is arithmetic, not performance, and comparing completion across skippable and non-skippable placements compares two different things.

Viewability thresholds are minimums, not goals. The standard for video is a defined portion of pixels in view for a defined duration — a low bar deliberately, so that it can be measured consistently. An impression that clears it has not necessarily been watched.

View-through conversions are the most generous number in the report and the one most often used to justify budget. Segment them from click-through conversions and look at them separately.

The measures worth building a decision on are audible-and-visible-on-complete where the platform offers it, brand lift where the budget supports a study, and a geographic holdout for anything substantial. All three are harder than reading the dashboard, which is why the dashboard remains popular.

Sources

What each claim on this page rests on. Entries are typed so you can see which are primary.