Mobile used to take the backseat to desktop, but in 2014 this is no longer the case. One out of four people worldwide own a smartphone, and at least 55% of all time spent on retail sites takes place on mobile devices. These numbers are why we've decided to release our annual state of the union for mobile web performance on the same day as our quarterly desktop state of the union.
In this post, I'll walk through a high-level summary of our key findings, including statistics around mobile load times, website fragmentation across devices, and mobile performance challenges.
Today, 90% of mobile users use multiple screens to complete a purchase. The average shopper makes 6.2 visits to a website - using 2.6 different devices - before they complete a transaction. This is why, regardless of whether they are on a smartphone, tablet, laptop, or desktop, online consumers expect a fast, consistent user experience.
In fact, 85% of mobile users expect pages to load as fast - or faster - on their phones as they do on their desktops, and two out of three say they expect pages to load in 4 seconds or less on mobile devices. One of the aims of our annual mobile performance study is to see if site owners are able to deliver on this expectation.
Background
[caption id="attachment_3992" align="alignright" width="196"]
Click to enlarge infographic[/caption]
Every year, we study the real-world performance of the top 100 retail sites, as ranked by Alexa.com. Using a range of smartphones and tablets, we subject the home page of each site - both the full-site version and the m-dot version - to a series of page speed tests over 4G/LTE and Wifi networks.
Note that these tests are not intended to serve as performance reviews for the individual devices, and the results shouldn't be interpreted as such. Rather, they serve as a reminder of how broadly performance can vary across devices and a strong suggestion that site owners need to test the performance of their sites across a range of devices and connection types.
Key findings
1. Both m-dot and full sites are failing to meet smartphone user expectations
At the top of this post, I cited the statistic that the average mobile user expects pages to load in 4 seconds or less. We found that the median m-dot page took 4.8 seconds to load on the iPhone 5s and 7 seconds to load on the iPhone 4s.
Because many shoppers will choose to view the full-site version of a page over the m-dot version, we tested those as well. The performance numbers were even more alarming: the median page took 11 seconds to load on the iPhone 5s and 15.2 seconds to load on the iPhone 4s. (All tests were conducted over 4G/LTE.)
2. Tablets were also served sub-optimal load times
Median load times varied across the tablets we tested, ranging from 5.7 seconds for the Galaxy Note to 8.1 seconds for the Nexus 7. The iPad 2 (6.9 seconds) and iPad Mini (8 seconds) fell between this range.
3. Mobile poses unique performance challenges, many of which can't be mitigated just by paring down pages
The median m-dot page contained 34 resources (e.g. images, CSS files, JavaScript). The median full-site page was roughly three times larger, containing 106 resource. Each resource request requires an individual round trip from its host server, and each of these round trips can incur 150-350 milliseconds of latency. For this reason, reducing the number of resources is an excellent way to improve performance. But it's complacent to believe that serving an m-dot page is a cure-all for mobile performance. M-dot pages can create new challenges and ignore others, such as:
- Redirects - When a user types in the full-site URL, he or she is redirected from the full site to the m-dot site. Redirecting delays load time.
- User agent sniffing - This is the practice of displaying content customized to serve specific mobile browsers and devices. It also adds to load time.
- Page complexity - Third-party scripts, custom fonts, stylesheets, and JavaScript all have the power to delay or block a page from rendering.
4. Website fragmentation continues, but is it sustainable?
Site owners are making a brave attempt to serve optimized versions of their pages to smartphone and tablet users, but these efforts are fraught with challenge. We found that 81% of sites automatically serve an m-dot version of their site to smartphones, and 8% serve a t-dot version of their site to tablets.
However, we also found that 20% of m-dot sites do not allow shoppers to access the full site in any way. This is frustrating for the large cohort of shoppers who prefer to view the full site on their smartphones.
We also found that sites struggle to show the appropriate version to tablets. Consumers shopping via tablet expect to get a desktop, not mobile, experience, yet we found that 5% of retailers served their m-dot site to iPad 2 owners and a surprising 30% served their m-dot site to Nexus 7 owners.
Website fragmentation is a controversial topic, and it has equally passionate advocates and detractors. Our report goes into some of the advantages and disadvantages of maintaining and serving multiple versions of your site.
Takeaway
The challenges of serving faster pages to desktop shoppers are huge, but they pale alongside the challenges of serving faster pages to increasingly demanding mobile shoppers. But while there is no magic bullet for mobile performance, there are many practices that site owners can follow - and others they can avoid - to ensure that their customers receive the best possible online experience.