How Long Should an Ecommerce Video Ad Be

What each placement actually accepts

Duration options vary meaningfully by platform. In the US and Canada, Prime Video accepts 15, 30, 45 or 60 seconds; Twitch accepts 15 or 30; Fire TV Channels accepts 15 through 90; and third-party apps accept the widest range, from 6 seconds up to 90. That range exists because different placements suit different message lengths — a device-level Fire TV placement like Feature Rotator has room for a longer, more cinematic open than a Twitch video unit competing with an active live stream and chat window.

Amazon's own guidance for both streaming TV and online video placements generally points toward 15 or 30 seconds as the practical sweet spot — long enough to develop a message, short enough to fit most platforms' accepted range without a separate cut for every placement.

International markets complicate this further. Most European markets shift the accepted set toward 10, 15, 20, 30, 40 and 60-second options across Prime Video, Twitch and Fire TV Channels alike, rather than the US and Canada's 15/30/45/60 set. A brand running a single global creative package needs to check the accepted durations market by market rather than assuming the US spec travels everywhere unchanged — a 45-second US cut, for instance, has no direct equivalent in the common European duration set.

What the completion data actually shows

The instinct that shorter always completes better holds on skippable formats, where a viewer who's losing interest can act on it at any second. On non-skippable CTV, the data tells a more nuanced story: reported completion rates run about 95.9% for 30-second ads, 93.9% for 15-second ads, and 90.4% for ads of 10 seconds or under — genuinely close together, and not in the direction the "shorter is always better" rule of thumb would predict. The mechanism is straightforward: a non-skippable viewer was never going to leave early regardless of length, so completion rate on this format measures whether the ad finished playing more than whether the viewer chose to keep watching.

None of this benchmark data is Amazon-specific — it's aggregated across the CTV category broadly, which is exactly why it's worth treating as a planning reference rather than a guarantee for any one placement. An account's own completion-rate history, once it exists, is a better guide for that specific account's future length decisions than an industry average ever will be.

A worked example: choosing length by placement

For a plan spanning Prime Video, Fire TV Channels and Twitch, cut a primary 30-second asset — accepted everywhere in that mix, and supported by completion data showing it holds up nearly as well as 15 seconds on non-skippable inventory. Hold a 15-second cut in reserve specifically for Twitch, where the shorter option may better suit a more actively-engaged, distraction-prone viewer, and for any OLV placement where skippability makes the shorter length's completion advantage genuinely matter. Building both from the start, rather than cutting a 15-second version down from a single 30-second master after the fact, generally produces a stronger short cut than an afterthought edit does.

What to do when a length isn't working

If a 30-second CTV cut is completing well below the ~95% benchmark, the length probably isn't the problem — check the opening seconds and overall pacing before shortening it, since non-skippable completion data doesn't support "it's too long" as a first diagnosis. If a 15-second OLV cut has a weak completion rate against the ~62% skippable-format average, the opening is almost certainly the issue — a skippable viewer decides within the first few seconds, and no amount of trimming the back half of the ad fixes a weak hook at the front.

The common mistake

The mistake is applying one length rule — usually "always go shorter" — across every placement in a plan, without checking whether that placement is skippable or not. The rule is directionally right for skippable formats and much weaker evidence for non-skippable CTV, where the data doesn't show the same sharp completion drop-off by length. reMKTR runs 109 live Amazon DSP advertiser seats and cuts length by placement type, not by a single house rule, specifically because the data genuinely differs between skippable and non-skippable inventory.

The second version of the mistake is choosing length based on production convenience rather than the message. A 30-second script padded to fill the runtime, or a genuinely 45-second story compressed into 15 seconds, both perform worse than a length chosen to fit the actual content — length should follow the message, with the platform's accepted options as the outer constraint, not the starting point.

A second, quieter mistake is assuming length decisions only affect completion rate. Length also affects production cost and, more subtly, how much creative testing a budget can support — a 60-second asset costs meaningfully more to produce well than a 15-second one, and a brand that commits its whole production budget to one long-form hero asset has less left to fund the kind of opening-seconds testing that reliably improves performance more than length alone does.