How mobile sites performed during Q2 2017
Editor's Note: Below is a quarterly report penned exclusively for Mobile Marketer by Dennis Callaghan, director of industry innovation at Catchpoint that analyzes the performance of mobile sites from leading brands across retail, news, travel and banking.
In Q2 2017, the top five performing and fastest mobile retail sites were Amazon, W.W. Grainger, Apple, Netflix and Toys ‘R’ Us. All five sites delivered webpage load times under two seconds. The average webpage load time for the 33 retailers measured in our index was a very respectable 3.24 seconds — a slight improvement over our last set of measurements (from Q4 2016), when the average was 3.3 seconds.
Amazon’s mobile site performance (average webpage load time in Q2 — 1.47 seconds) has been truly remarkable. Not only did it maintain this performance leadership through Prime Day, the company actually improved, delivering the strongest web page load time (1.14 seconds throughout the 30 hours of Prime Day sales) among the major retailers running concurrent sales. Considering that traffic loads were estimated to be 20 times larger than normal, this was an impressive feat.
There are numerous factors driving Amazon’s success, including a relatively low page weight (.8 MB — 56% lighter than the industry average of 1.8 MB), which plays an important role in site speed. Other factors include a relatively small number of third-party hosts (27 versus the industry average of 44) and items (98 versus the industry average of 129). Multiple hosts can be particularly problematic — for every host supporting a site, there is a series of back-and-forth interactions that can extend webpage load time; and the more hosts present, the bigger the impact.
Another interesting finding is that of the 33 retailers measured, the majority (18) delivered webpage load times above the industry average. There were eight sites that clocked speeds of less than two seconds, and only three in the five-second range. This suggests that websites are increasingly clustering closer to leaders (i.e. Amazon) setting the performance standard. It also suggests a clear understanding of how vital strong performance is to driving conversions and revenues in this sector.
This quarter’s top five performing mobile news sites were NPR, CBS News, CNET, The Financial Times and Christian Science Monitor, all delivering average webpage load times of two seconds or less. NPR had the fastest loading page of all the sites on our entire index, clocking in at 1.08 seconds. The average webpage load time for the 49 news organizations measured in our index was 4.93 seconds — the slowest of all the industries measured, and a degradation from Q4 2016, when the average was 4.5 seconds.
This is consistent with past analyses, where we have found mobile news sites overly bloated due to a heavy reliance on third-party ad hosts. Compared to retail sites, the average number of hosts supporting mobile news sites is much higher (65), as is the number of items (192). The average total downloaded bytes of mobile news sites is also heavier than retail, at 2.67 MB.
Of all the industries in our index, mobile news sites are most susceptible to slowdowns due to an overabundance of third-party hosts. Across industries, page weight (total downloaded bytes) has traditionally been viewed as the biggest determinant of site speed. However, in the mobile news category there are numerous examples of significantly heavier pages delivering faster webpage load times than lighter sites with more hosts. Minimizing the number of third-party hosts therefore represents a major optimization opportunity in this sector.
Additionally, compared to retail — where not a single company had a webpage load time exceeding approximately five seconds — in news, there are more outliers in the six to eight second range, or higher. The performance range is also more evenly spread out — for example, 12 companies in the upper echelon (2-3 seconds), 11 in the lower echelon (7-8 seconds), with the highest concentration right in the middle, in the 4-5 second range. This suggests that compared to retail, there is less of a clear Amazon-like leader exerting competitive pressure.
One mobile news site worth calling out is CBS News, which has a significantly heavier page than the average, along with a higher number of hosts and items. CBS News is still able to deliver exceptional performance (average webpage load time — 1.31 seconds) mainly due to the page structure. The page strategically loads content essential to user interaction ahead of non-critical and third-party host content, which can continue loading in the background if needed. This minimizes the potential negative impact on user perception.
Q2’s top five performers were Trip Advisor, Google Flights, United, Booking and Jet Blue, all delivering webpage load times of 2.7 seconds or less. The average webpage load time for 21 travel organizations measured in our index was 4.23 seconds — slightly slower than the 4.08 seconds recorded in Q4 2016.
Of the four industries, mobile travel and mobile retail are most similar in terms of average total downloaded bytes (mobile travel — 2 MB) as well as the average numbers of hosts and items, although these two numbers were slightly lower in travel. However, the average webpage load time in mobile travel was about a full second slower (4.23 seconds) than mobile retail.
Once again, this may be based in the retail industry’s understanding of the tight correlation between web performance and sales. That is, retail companies may be more likely to implement additional strategies to circumvent the higher number of hosts and items, thus delivering stronger performance than the travel sites. One common strategy is enlisting third-party infrastructure like CDNs, which can bring heavier web content like video closer to geographically distributed end users, reducing wait time.
The top five performers in Q2 were U.S. Bank, Huntington, Wells Fargo, Bank of America and Chase, all delivering webpage load times of 3.1 seconds or less. The average webpage load time for the 10 banks in our index was 3.14 seconds — a drop from Q4 2016’s 2.43 seconds, however, still the fastest of all the industries. Like the mobile news category, individual webpage download times were evenly spread out, with four above the industry average and six below (of these six, one was just milliseconds below).
These findings are in line with past analyses, as banks traditionally keep their number of hosts and items much lower than the other industries (18 and 77 respectively this quarter), contributing significantly to their fast download speeds. However, one interesting finding from this quarter was that of all the industries, mobile banking sites were, on average, the heaviest (approximately 4 MB), when traditionally they tend to be lighter. This was due to two outliers who were quite heavy this period — HSBC at 13 MB, and Citizens Bank at almost 17 MB. All of the remaining sites were under 3 MB.
Recently, HSBC has been adding and removing heavy media files on its site including images and videos, which has caused average page weight to increase and site performance to fluctuate frequently. However, HSBC’s speed average (3.29 seconds) was right in the middle of the industry pack, as HSBC also had the lowest number of hosts in the index which helped counteract the performance impact of the heavy page. Since Q4 2016, Citizens Bank has increased the number of images which has contributed to the page size significantly. Not surprisingly, Citizens Bank had the slowest webpage load time in the index.
The key observation from this quarter is the extent to which the number of third-party hosts deployed on a site influences webpage load time. This challenges our earlier notion that keeping page weights down was the single most effective way organizations can maximize performance. While page weight and the number of items, particularly images, are still extremely important factors, minimizing third-party hosts — and carefully managing those that must be there — can have an equal or even greater net performance benefit. Minimizing the number of third-party hosts is also a strategy to offset the negative impact of heavy pages.
Methodology and Definitions
This benchmark monitors each website’s home page with measurements taken from 27 Catchpoint Systems backbone monitoring nodes in various locations across the United States, at intervals of two minutes, with results collated into a median average.
Webpage load time: the time it takes for enough page elements to load for an end user to begin interacting with a page.
Host: a server or other computer that delivers data, content or services over the internet, often to websites. Modern websites are increasingly complex, pulling from a number of external hosts that provide specialized services.
Items: static web page components (files and images) that a website loads from its own servers. These can include PDFs, PNGs, JPEGs and GIFs.
Backbone monitoring nodes: website performance (speed) measurement software located in servers at central points along the high capacity networks that form the core of the internet. These nodes simulate end user contact with each website and measure the performance.