Mobile Site Speed — The Business Perspective
Part 3 of our ongoing study on “Mobile Site Speed & The Impact on Web Performance”
This is the third blog post in our ongoing series on “Mobile Site Speed & the Impact on E-Commerce”. So far we have summed up existing research on the user perspective and found that many factors determine user satisfaction and the degree to which poor web performance disrupts the user experience. As the bottom line, however, we concluded that fast page load times are vital for keeping the users’ attention and thus for making them convert. (Details: speedstudy.info)
What This Post is About
In this post, we will assume the business perspective and discuss the connection between fast page load times and key performance indicators (KPIs) of your business. To this end, we first summarize the wealth of studies already out there and condense their results into an overarching comparison. We then conclude with open research challenges and an outline of how we are going to address them later in this series.
Load Time & Revenue
There are many reports about how even minuscule improvements in page load time can yield enormous profits [13]. One of the more famous ones comes from Amazon who observed a revenue decline of 1% for every additional 100 ms that users had to wait [1]. Walmart conversely found that shaving of 100 ms of page load time resulted in 1% increased revenue for them [2]. Similar results have been reported by others, e.g. by Zalando whose session revenue increased by 0.7% revenue after a 100 ms acceleration [3].
Load Time & SEO
Page load time is known to be SEO-critical, since it used as a ranking factor by Google [14]. But an inherent positive effect of faster page loads on visibility is also documented. After redesigning their website in 2015, for instance, GQ experienced not only a page load improvement from 7 to 2 seconds, but also a sustained increase of traffic by 80% [5]. As a more recent example, Pinterest reengineered part of its website in 2017 and achieved (among other minor changes) an acceleration of 40% and 15% more SEO traffic [6]. While these examples show the positive impact of page load time improvements, there is also evidence that deceleration can lead to decreased traffic. Back in 2006, Google famously tried delivering 30 instead of just 10 search results already, because market research had indicated that this was generally desired by users. However, the new functionality was accompanied by half a second of additional page load time which resulted in a 20% traffic drop [4].
Load Time & User Engagement
In our last blog post on user expectations on web performance, we already covered that users are not happy with websites that feel sluggish. But you do not even have to ask users for their opinion on slow page load times: You just have to look at your tracking data. In the context of the Google Mobile Speed Race, the Otto group observed a 25% increase of average session length, after the time until the first contentful paint (FCP) had been reduced by about 40% [7]. There are also studies that combine data from multiple cases. In one of them, Akamai aggregated data from many of their customers to find that adding Two seconds of additional page load time typically increases bounce rate by a whopping 103% [8]. In a number of interviews, Forrester similarly concluded that reducing page load times by 80% correlates with about 60% increased session length on mobile devices [9].
Load Time & User Satisfaction
There are also several studies on user satisfaction and productivity. Radware and the Aberdeen Group conducted studies in 2013 and 2008, respectively, discovering similar results: In the Radware study, increasing load times by 500 ms resulted in increased frustration (26%) [10], while the Aberdeen Group found user satisfaction to decrease by about 16% when increasing load time by 1 second [11]. While it is not entirely clear how frustration and satisfaction were measured in these studies, it becomes clear that waiting time is perceived as a nuisance and keeps users from achieving their goals. And this does not come as news, seeing that a comparable study uncovered the same effect on productivity almost half a century ago: In 1978, the Imperial College London was concerned with waiting times in interactive applications and observed that increasing application response times by 50% reduced productivity by the same amount [12].
The Big Picture
Obviously, all these different reports are hard to compare. Some observations might be dependent on the scale of the use case, so that effects clearly visible at Amazon or Walmart might apply differently at small- or mid-sized businesses. Some of the studies were issued decades ago and might have become outdated. And some results are simply incomparable altogether, since they refer to different verticals — for example, the correlation between Web performance and business success is likely to be different for publishing and e-commerce websites. As another fundamental issue with most studies, results are usually not reliable because they come from before-after comparisons rather than statistically sound A/B tests, because the methodology is not explained well, or simply because the measured acceleration was accompanied by other effects that are also related to business success. For example, it is impossible to attribute increased traffic after a website redesign to faster page loads alone, when functionality and UI have changed as well.
But even though a detailed analysis seems difficult, all of these results underline a single message that is intuitively plausible: As illustrated above, waiting users are unhappy users — and it shows in your business numbers. Even more precise, accelerating your website always seems to result in an improvement that is relevant to your business, irrespective of the concrete metric that you are looking at. And the inverse also seems to hold: Whenever performance deteriorates, so do your success criteria.
The problem with existing approaches for measuring the business impact on Web performance simply is that they are incidental reports after the fact rather than scientific approaches for determining the actual correlation behind it. Next, we will therefore go into more detail on how technical and business performance can be measured reliably.
What Comes Next
To facilitate more precise performance measurements, our upcoming blog post in this series will cover the most relevant notions of performance in the Web and different ways to capture them. We will look at both synthetic tests and real-user monitoring as the two basic options for actually taking measurements, but we will also discuss passive ways of information gathering such as using the Chrome UX Report (CrUX) or plain old log analysis.
Learn More
If you want to receive a notification as soon we published the next post in our series, register for our speedstudy.info newsletter. (No spamming, promise!) Until then, feel free to read the other posts in our series or check out our code.talks 2019 presentation video on our ongoing study on “Mobile Site Speed & the Impact on E-Commerce”.
Stay fast!
References
[1] Greg Linden. Make Data Useful. Stanford Data Mining Class CS345A, 2006
[2] C. Crocker, A. Kulick, B. Ram. Real-User Monitoring at Walmart, SF & SV Web Performance Group, 2012.
[3] Shuhei Kagawa, Jeff Cybulski, David Martin Jones, et. al.. Loading Time Matters. Zalando Engineering Blog, 2018
[4] Marissa Mayer, Conference Keynote, Web 2.0, 2006
[5] Lucia Moses. How GQ Cut Its Webpage Load Time By 80 Percent, Digiday, 2015.
[6] Sam Meder, Vadim Antonov, Jeff Chang. Driving User Growth With Performance Improvements, Pinterest Blog, 2017
[7] Lars Bognar. Mobile Speed Race der Otto Group Verbessert Mobile Ladezeiten, TWG, 2019
[8] Akamai. Akamai Online Retail Performance Report: Milliseconds Are Critical, Akamai Blog, 2017
[9] Forrester. The Total Economic Impact ™ Of Accelerated Mobile Pages, 2017
[10] Tammy Everts. Mobile Web Stress: The Impact of Network Speed on Emotional Engagement and Brand Perception, Radware, 2013
[11] The Performance of Web Applications: Customers Are Won or Lost in One Second, Aberdeen Group, 2008.
[12] T. Goodman, R. Spence. The Effect of System Response Time on Interactive Computer Aided Problem Solving, ICL, 1978
[13] Peter Glynn, Richard Wheaton, et. al.. Milliseconds Make Millions, Deloitte Digital & Fifty-Five, 2020
[14] Addy Osmani, Ilya Grigorik. Speed is now a landing page factor for Google Search and Ads, Google Developers Blog (Web Updates), 2018
