Mobile video is great on website . When it works.

Implemented correctly, video or audio *should not* impact the speed that pages load on a mobile device and when the play button is pressed, it needs to start quickly and work well.

Video content is top of the agenda for many brands. It is proving a great way to engage customers and visitors, but when viewed on mobile devices, particularly those on cellular connection, video (and to a lesser extent audio) should come with a health warning.

Users are increasingly impatient with slow-to-load and stalling video and will start to abandon a video after waiting just two seconds, research from UMass and Akamai shows.

This column, the first of three parts, will take a close look at how and why video affects page performance. In the second part, we’ll look at the impact of video autopay and audio on page performance, as well as what makes a poor viewer experience (VX).

Finally, we’ll explore how to detect, avoid and remedy issues to prevent users tuning out.

Video is a massive mobile data hog

The provision and consumption of video on mobile devices via web and apps is growing rapidly. Mobile video is already 60% of total mobile data traffic worldwide and is expected to be 78% by 2021 according to Cisco’s Visual Networking Index (VNI).

All other elements will grow over the next five years, but their proportion of overall traffic will be less. Audio will be 5% compared with 8% today and mobile web will be 14% compared with 30% of traffic today.

Video – and audio – used wrongly or inefficiently will impact mobile user experience (UX) – or should we say: “viewer experience” (VX) or “listener experience” (LX) – massively, but not necessarily in the same way as oversized images and poor or inefficient use of JavaScript.

Images and JavaScript, as seen in previous columns, are the biggest causes of slow loading mobile web pages. As discussed below, video can still contribute to page size and therefore contribute to page load delays, particularly (it seems) where autopay is used, as we will discuss below.

But the biggest impact on VX comes after page load when the video is slow, or fails, to start or stalls.

The two charts below are from HTTP Archive, which twice monthly records the page size and download speed of the homepages of the top 1 million sites to desktop and mobile devices, using the excellent WebPageTest.

The first chart shows the breakdown of content types by bytes – images, JavaScript, video, stylesheets, HTML and fonts – as an average of all homepages recorded on April 15, 2020.

Video is 128kB or 5.5% of the total bytes loaded (2312 kB or 2.3MB). This might appear small, until you realize that 97% of pages monitored by HTTP Archive have no video content (we examine this surprising stat below).

Pages that do have video content will therefore show a higher proportion of video content.

The second chart (captured April 15, 2020) shows the content breakdown for the homepage of the US digital agency Huge. Here video content is 727kB or 14.5% of the total bytes. The total weight of the page is 5MB, which is a homepage worthy of the company name, and, when measured, took 25.8 seconds to load on a mobile device, according to HTTP Archive.

To be fair, many agencies (digital, media, advertising et al) have surprisingly slow loading, heavy weight sites (considering the importance of digital to their businesses), though Huge is exceptionally large. A trimmer example is Young and Rubicam. On the same date the Y&R homepage took three seconds to load 783kB on a mobile device (on other dates it took nine seconds) according to HTTP Archive.

Video shouldn’t affect page load size or download speed

Implemented correctly, video (or audio) should not impact the size of the webpage or the speed that pages load on a mobile device, according to the experts.

Even when video is present on the page, to render the page, the browser only needs to load the video container, teaser image, start button etc. it doesn’t need to download the entire video (as the visitor may not want to watch it at all). Thus video and audio ought not to be a significant proportion of content recorded by HTTP Archive / WebPageTest – as we will see when we look at the most popular sites.