Split testing video length is one of the most overlooked levers in content optimisation. Marketers spend months debating aspect ratios, thumbnail designs, and caption styles, then publish a two-minute video because it "feels right." Retention data almost never confirms the feeling. Running a deliberate test against different durations gives you something more useful: a signal grounded in actual viewing behaviour rather than instinct.
Why video length is worth testing specifically
Length isn't a production afterthought. It's a structural choice that controls pacing, information density, and where the audience's patience runs out. A video that performs well at 90 seconds may haemorrhage viewers at the two-minute mark not because the content worsens, but because the format has already asked for more attention than the platform trained the audience to give.
Platform context sharpens this further. A 60-second cut that dominates Instagram Reels can feel sluggish on YouTube, where watch-time signals reward longer content. YouTube's algorithm, for instance, weighs absolute watch time alongside retention percentage, so a three-minute video watched to completion often outperforms a 90-second clip abandoned at 70 percent. That relationship between platform mechanics and duration is exactly why guessing doesn't work: the right length depends on where the video lives, not just what it says.
West Melbourne Studios approaches every production with this in mind. Before recommending a finished duration, the team evaluates distribution context, audience scroll behaviour, and content type, because those three factors together predict drop-off far better than editorial instinct alone.
How to structure a split test for duration
A clean video length test requires one variable: duration. Everything else stays identical. Same hook, same core message, same visual style. If you change the thumbnail while changing the length, you can't attribute a retention lift to either factor with confidence. This sounds obvious. Most teams violate it within the first week of testing.
The practical setup involves producing two or three cuts of the same content: a shorter version that covers the core argument, a standard version at your current default, and optionally a longer version that adds supporting detail or a deeper example. For a typical B2B explainer, that might mean a 60-second cut, a 90-second cut, and a 2.5-minute cut.
Distribute each version to a comparable audience segment, whether through A/B testing tools built into your video platform, paid promotion to split audiences, or controlled organic posting across separate posting slots with matching algorithmic conditions. Collect at least 500 views per variant before drawing conclusions. Anything below that introduces too much variance to act on.
The metrics that matter most in a duration test aren't views. They're average percentage viewed, the exact drop-off timestamp, and, where available, click-through rate to a next action. Attention metrics are gradually replacing view counts as the benchmark for video performance, and a duration test is one of the clearest contexts in which that shift makes practical sense.
Reading the retention curve correctly
Every platform that gives you retention data shows the same shape: a steep drop in the first few seconds, a plateau through the middle, and a second drop near the end. What changes between your length variants is where that final drop occurs and how steep it is.
If your 90-second version shows viewers leaving at the 70-second mark while your 60-second cut holds to 55 seconds, the longer version is losing people before it finishes. That's not evidence that shorter is always better. It's evidence that the last 20 seconds of your 90-second cut aren't earning their place. The fix might be cutting those seconds, rewriting them, or moving your call to action earlier so it lands before the audience exits.
The most actionable finding from a duration test is often not "use this length instead." It's a precise timestamp telling you exactly where editorial tightening is needed. That specificity is something no amount of pre-production guesswork can replicate.
This precision also matters for how A/B testing frameworks apply to video more broadly. Thumbnail tests and duration tests share the same logic: isolate one variable, run it against a comparable audience, and read the data before you act on it. Studios that build this habit for thumbnails and neglect it for length are leaving a significant optimisation gap open.
Common mistakes that invalidate a duration test
Three mistakes appear repeatedly in duration tests run without a clear framework.
- Testing across platforms simultaneously. Running the 60-second cut on TikTok and the 90-second cut on LinkedIn and comparing retention curves directly is meaningless. Platform audiences, algorithms, and viewing contexts differ too substantially for cross-platform comparisons to produce clean data.
- Changing the edit, not just the length. If your shorter version uses a tighter edit with different b-roll pacing, you've changed two variables. Shorten by removing content at the end or trimming pauses, not by restructuring the narrative.
- Acting on a single test. One test cycle tells you about one piece of content. Patterns emerge across four or five tests, and those patterns are what should inform your production defaults going forward.
What to do with the results
Once you have retention data across at least two content cycles, the pattern usually resolves into one of three findings. First, a consistent drop-off point that appears regardless of total length, which suggests a structural issue in how you're building the middle section. Second, a clear length advantage on a specific platform, which informs your production brief for that channel. Third, no meaningful difference between variants, which most often signals that the hook, not the duration, is the bigger retention driver.
Each finding directs a different next action. The first points toward script and pacing work. The second becomes a platform-specific content rule. The third redirects testing attention to the opening five seconds, where most audience decisions are made before length even becomes a factor.
West Melbourne Studios integrates duration testing into content strategy reviews for clients running ongoing video programmes. The results inform not just how long individual videos run, but how content calendars are structured across a quarter. For studios managing a high volume of output, that compound effect is where the real retention gains accumulate. The goal isn't a single perfectly-lengthed video. It's a library where every piece has been tested against its audience and trimmed accordingly.