The above graph (let's call it Exhibit A, because it's fun to feel like a detective) is the most consolidated display of the data, with one data point for each video length range, on the x-axis, with the average % viewed for videos of that length on the y-axis. Basically, it's representative of the video engagement number within Wistia (so you could compare your number with the number for that length on this graph to see how well you're doing compared to the average!). This graph tells us that shorter videos are better for getting people to watch the whole thing. After all, most business video is created to serve up a pre-packaged message, so the longer the video, the less people will watch. It's also noticeable that after a certain point the engagement average flattens out -- so there's not a major difference in engagement for a 4-minute versus a 10-minute video.
On to Exhibit B, where the x-axis represents the percentage of a video viewed (think of each line as the average engagement graph for a video of that length range in Wistia, one of the bars of Exhibit A over time, with each line representing the average video for that bin, with the lengths normalized), and the y-axis represents audience engagement. In this case, you could compare the engagement graph line of your own video to the appropriate line of this graph to compare yourself to the average.
A possible takeaway from this graph would be to organize the content of your videos journalistically, placing the most important, essential information first, then following with supporting details. For longer videos, notice that the dropoff at the beginning is extremely steep; it seems that most viewers decide quickly whether or not to watch, and once that decision is made, they tend to stick around until the end of the video, when they detect that the video is wrapping up and another drop off occurs. For this reason, if you're using a post-roll call-to-action, you might want to consider a harder stop to your video, rather than a meandering wrap-up -- this will ensure that more viewers stick around to see your CTA.
Exhibit C takes things one step deeper: this is the raw data that went into creating the above graphs. Each frame in this animated graph represents one of the time ranges from the above graph. Each of the faded lines is the engagement graph for an actual video, while the average line for that video length is in orange. Again, the x-axis represents percent viewed and the y-axis represents audience engagement.
The interesting thing to notice here is the wide variation even for videos of the same length. The variation tends to be more wide at the beginning, tightening toward the end (again hinting that people decide whether or not to watch pretty quickly). There are definitely outliers, but all in all, longer videos see a tighter overall distribution, where it's safe to say that if you're doing 30% versus 25% engagement on average, for example, you're doing pretty well.
The main takeaways from our first "Does Length Matter?" post still hold true: overall, shorter videos are more engaging than longer videos. You should strive to make your content as concise as possible to achieve the highest engagement. If your message is more complex, feel free to give it the time it deserves, but understand that a major chunk your audience won’t make it to the end of the video and consider front-loading your video with the most important information at the beginning.