How Do People Perceive A Website’s Speed?

How Do People Perceive A Website’s Speed?
May 01, 2018

Summary

We all know our websites should load fast. But, how fast is “fast”? Every marketer, designer and developer has their own ideas, but we wanted to know what an ordinary web visitor thought about a website’s speed. To find out, we asked over two thousand people to wait for a website load and then asked test participants to describe how fast they thought the website loaded.

How fast is the website?

Most participants in our test described load times of 3-5 seconds as fast or medium speed. That is true for people on mobile devices as well as desktop devices. To say the least, this result surprised us since most recommendations are for websites to load in under 2-3 seconds. It wasn’t until 8-second load times that people finally started to describe a website’s load as slow but even then participants didn’t describe these speeds as very slow. Visitors might be more forgiving of slower speeds than we thought.

How would you rate the speed at which this website loaded?
Mean Perception of Load Time (All Devices)
Perceived Speed Scale: 5 = Very Fast, 1 = Very Slow

How fast should you make your website?

They may be more forgiving but a faster load time is still perceived as faster. We can’t and shouldn’t accept sluggish load times. The question, then is, when improving a website’s speed, how much of an improvement should we aim for? This test shows that smaller gains in speed won’t be all that noticeable to an ordinary visitor. If a goal of a speed optimization project is user experience and making sure visitors notice something has changed for the better, then bigger gains in speed will be needed to make an investment worthwhile. Put another way, a huge investment in improving speed that results in a half-second gain won’t help visitors much, though the speed gains may of course have other benefits.

What if you can’t make your website load faster?

Many companies can’t make their websites load any faster due to budget, time, or technical constraints. One common recommendation is to use a load indicator, but this seemed to make little difference for participants in this test. As a result, when it comes to improving a website’s load time, the only real solution is to find a way to overcome constraints in order to make a website fast enough that it can feel like it loads fast for visitors.

Who gets the blame?

The majority of participants in this test blamed the internet provider for slower speeds, in whole or in part. However, still just over half of participants blamed the company behind the website, in whole or in part, for the slower website speeds. As the speed lengthens, more the blame shifted toward the company behind the website. We hope this information helps you find the right ways to improve your website and avoid being blamed for slow speeds.

Contents

How Do People Perceive a Website’s Total Load Time?
How Do Load Indicators Affect The Perception of Speed?
Who Gets Blamed For Slow Speeds?
Longer Load Times
Discussion: Takeaways & Action Items
Future Questions
Methodology
About Elementive

How Do People Perceive a Website’s Total Load Time?

We asked participants to wait for a website to load from somewhere between .5 to 5 seconds. After the website completed its load, participants were asked how fast the website seemed to load on a scale from one to five. A response of five indicated “Very Fast” and a response of one indicated “Very Slow”, with a three representing “Medium”.

Most recommendations are for websites to load in 2-3 seconds and because of that, we expected participants to rate load times more than 3 seconds as “Slow” or “Very Slow.” However, this wasn’t the case—across mobile and desktop devices. Even at a 5-second load time, participants responded with a median score of 3.75, which would fall between “Medium” and “Fast” on this test’s scale.

All Devices

Mean Perception of Load Time (All Devices)
Perceived Speed Scale: 5 = Very Fast, 1 = Very Slow

Total Load TimeMean Perception of Speed
0.54.60
14.40
1.54.33
24.19
2.54.24
34.14
3.53.98
43.90
4.53.85
53.75

Desktop & Laptop Devices

Mean Perception of Load Time (Desktop & Laptop Devices)
Perceived Speed Scale: 5 = Very Fast, 1 = Very Slow

Total Load TimeMean Perception of Speed
0.54.54
14.37
1.54.40
24.09
2.54.26
34.20
3.53.82
43.85
4.53.94
53.78

Mobile Devices

Mean Perception of Load Time (Mobile Devices)
Perceived Speed Scale: 5 = Very Fast, 1 = Very Slow

Total Load TimeMean Perception of Speed
0.54.69
14.46
1.54.22
24.32
2.54.19
34.04
3.54.16
43.97
4.53.73
53.68

How Do Load Indicators Affect The Perception of Speed?

To help make a website feel faster, one common recommendation is to include a load indicator, such as a status bar, spinning circle, or a simple text message. Knowing this, our hypothesis was that by including a load indicator, participants would rate the website’s load as if felt somewhat faster even at slower speeds. This hypothesis was largely disproven and it was shown that an indicator on initial website load doesn’t affect the overall perception of speed.

Participants were asked to wait for a website to load from somewhere between .5 to 5 seconds and were randomly assigned a load indicator of a spinning circle or the word “Loading…”. About a third of participants saw no load indicator, so while the website loaded participants saw a blank screen. After the website completed its load, participants were asked to rate how fast they thought this website loaded. A response of five indicated “Very Fast” and a response of one indicated “Very Slow”, with a three representing “Medium”.

All Devices

Perception of Total Load Time by Indicator (All Devices)
Perceived Speed Scale: 5 = Very Fast, 1 = Very Slow

TimeCircleTextNoneAll Indicators
0.54.574.564.664.60
14.434.494.294.40
1.54.284.334.384.33
24.044.314.204.19
2.54.254.164.284.24
34.224.174.044.14
3.53.973.864.143.98
44.063.893.753.90
4.53.963.853.743.85
53.793.983.513.75

Desktop & Laptop Devices

Perception of Total Load Time by Indicator (Desktop & Laptop Devices)
Perceived Speed Scale: 5 = Very Fast, 1 = Very Slow

TimeCircleTextNoneAll Indicators
0.54.504.494.634.54
14.304.484.294.37
1.54.294.474.454.40
23.974.194.124.09
2.54.214.294.294.26
34.354.104.144.20
3.53.913.604.003.82
44.033.733.763.85
4.54.003.973.833.94
53.704.033.573.78

Mobile Devices

Perception of Total Load Time by Indicator (Mobile Devices)
Perceived Speed Scale: 5 = Very Fast, 1 = Very Slow

TimeCircleTextNoneAll Indicators
0.54.674.704.704.69
14.574.504.294.46
1.54.254.144.274.22
24.164.504.284.32
2.54.333.954.284.19
34.004.293.934.04
3.54.064.114.334.16
44.114.083.733.97
4.53.903.693.593.73
53.953.753.433.68

Who Gets Blamed For Slow Speeds?

When websites slow down, naturally visitors may feel frustrated. Who gets the blame for the slower speed? To find out, we asked participants who they were more likely to blame for the slower speeds, the internet provider, the company who runs the website or both the internet provider and the company running the website equally. Our findings were conclusive across all devices and all load times: the internet provider gets most of the blame. However, while 16% people only blame the company running the website for slow speeds, 38% of participants gave equal blame to the internet provider and the company running the website.

Who are you more likely to blame for slower speeds?

Blamed PartyDesktopMobileOverall
The internet provider43.02%44.19%43.59%
The company running the website16.51%15.45%15.99%
Both37.78%37.71%37.74%
Neither2.70%2.66%2.68%

Longer Load Times

Perception of Slower Speeds

Many websites take longer than 5 seconds to fully load. Our hypothesis was that any load time beyond 5 seconds has to be perceived as “Slow” or “Very Slow”. To determine this, in a follow up to our main test, we asked participants to wait between 5.5 to 10 seconds for a website to fully load then rate how fast they thought the website loaded. A response of five indicated “Very Fast” and a response of one indicated “Very Slow”, with a three representing “Medium”. Our hypothesis was that a 10-second load would be rated as “Very Slow”, however this was disproven.

Mean Perception of Load Time (Mobile Devices)
Perceived Speed Scale: 5 = Very Fast, 1 = Very Slow

Total Load TimeMean Perception of Speed
5.53.33
63.21
6.53.48
73.36
7.53.08
83.10
8.52.86
92.69
9.52.88
102.59

Load Indicator at Slower Speeds

Participants in the longer load time test were randomly assigned a different load indicator of a spinning circle or the word “Loading…” while about a third of participants saw a blank screen while waiting for the website to load. Our hypothesis was that load indicators would matter greatly at these slower speeds. While this is somewhat true for the text indicator, largely the indicator did not help make the website seem faster to participants, especially at 10 seconds.

Perception of Total Load Time by Indicator (Mobile Devices)
Perceived Speed Scale: 5 = Very Fast, 1 = Very Slow

TimeCircleTextNoneAll Indicators
5.53.383.603.173.34
62.933.403.503.21
6.53.543.623.333.48
73.543.482.753.36
7.53.473.302.503.08
83.132.943.303.1
8.52.763.092.852.86
92.563.112.502.69
9.52.613.103.002.88
102.632.472.692.58

Blame at Slower Speeds

When asked who they would blame for the slower speeds of a website, participants viewing a load time of 5.5 to 10 seconds were more likely to blame the company running the website, either in full or in part, than participants in the main test group.

Who are you more likely to blame for slower speeds?

Blamed PartyDesktopMobileOverall
The internet provider38.43%38.86%38.64%
The company running the website24.02%21.33%22.73%
Both36.24%37.91%37.05%
Neither1.31%1.90%1.59%

Discussion: Takeaways & Action Items

How Fast Does It Feel?

One of the first questions this test can help us answer is how does the feeling of how fast a website loads correlate with how fast the website actually loads?

Before answering this question, we need to define two terms: the technical speed of the website and the perceived speed of the website. The technical speed of the website is how fast a tool like Pingdom, GT Metrix or Web Page Test tells you the website loads. The perceived speed is how fast the website seems to a visitor—that perceived speed is what our test measured.

The perceived website speed doesn’t fall off as dramatically as the technical website speed as we had expected prior to conducting this test. On desktop devices, the breakpoint between the perception of “Fast” and “Medium” happens somewhere between the 3 to 3.5 second mark. However, at 3.5 seconds, people only perceived the website as loading about 9.23% slower than it loaded at 3 seconds. What is interesting, though, is that a 3-second technical speed is 14.3% faster than a 3.5-second technical speed. In other words, the gain in perceived speed wasn’t as great as the gain in technical speed; people didn’t feel the increase in speed.

Similarly, there is a 10% difference in perceived speed between 5 and 3 second load times. This means 3 seconds isn’t just faster than 5 seconds, it also feels faster—but only 10% faster, even though improving the technical speed from 5 seconds to 3 seconds would be a gain of 40%. What this means is that people’s perceptions of the website’s speed do change. People do notice that a website is faster. But, the critical point is that the gain in perceived speed isn’t as great as the gain in the technical speed.

The conclusions are similar for participants using mobile devices. On mobile devices, the difference between the perceived speed of 5 and 3 seconds is somewhat smaller to the perceived speed on desktop devices—people only think the website loads 9% faster at 3 seconds than it does at 5 seconds. This is the same as the difference in the perceived speed between 3 seconds and 1 second.

On mobile devices, the breakpoint between “Fast” and “Medium” happens closer to the 3.5 to 4 second mark, a full half second later than it did on desktop devices. However, the difference at this breakpoint isn’t as pronounced. On desktop devices, the breakpoint between “Fast” and “Medium” represented a 9.23% difference in perceived speed but on mobile devices, the breakpoint between “Fast” and “Medium” only represented a 4.57% difference in the perception of speed.

Gains in Technical Speed Compared To Gains in Perceived Speed

Perceived Speed GainTechnical Speed Gain
2 to 1 second5.26%50%
3 to 2 seconds1.21%33%
4 to 3 seconds6.17%25%
5 to 4 seconds3.73%20%
5 to 3 seconds10.13%40%
5 to 2 seconds11.46%60%

How Fast Should My Website Be?

It might seem like the takeaway, then, is that five seconds might be plenty fast enough and no further work needs to be done improving the website’s speed. That isn’t the right conclusion, though; faster is still better as people do notice a difference. People may not notice the difference as much as we thought. The question, though, is how fast should your website be? Or, the related question is if you were to invest in a speed optimization project, how much of a gain in perceived and actual speed should you try for to make the investment worthwhile for your users?

What this test suggests is that if you are optimizing your website’s speed, you need to go big to make the gains worthwhile—at least from a user experience perspective. At the end of a speed optimization project, you want people to think and feel that your website is faster. After all, half of participants said they blame your company at least partially for the website running slowly. Therefore, we want to avoid the blame and ensure our website’s perceived speed is as fast as possible.

With little exception, gains in the perception of speed across all devices are not significant at half-second increments. This would mean that a half-second gain in speed is not enough for people to notice. For example, if you invest in a speed optimization project, and improve the technical speed from 4 to 3.5 seconds, your visitors won’t notice much, if any, difference in the perceived speed of the website. Instead, if your website currently loads at 4 seconds, you would need to reduce the technical speed of the website to 2.5 to 3 seconds in order for people to think your website was faster. Anything less and your investment in that speed optimization project would not be worthwhile for users.

This shifts somewhat on mobile devices. Here improving from 4 to 3.5 seconds of load time might have a noticeable increase in perceived speed given this is where we see the breakpoint between “Fast” and “Medium”. However, the difference in perceived speed between 4 and 3.5 seconds isn’t statistically significant. So, for visitors to see a difference an optimization project would likely require improving from 4 to 2.5 seconds before visitors notice the improved speed. Mobile isn’t unique—the overall conclusion is that bigger gains in speed are needed for visitors to feel the difference.

Load Indicators

However much we do not like it, there are barriers to improving your website’s speed. For example, the budget or time may not allow for an investment in the development work that would be required to improve the load time. Or, the technical structure is such that improving the speed would essentially mean rebuilding the entirety of the website as no quicker solutions exist.

One thing this test suggests is that a technical load time between 3-5 seconds may not be the end of the world from a user experience standpoint. People won’t think your website is fast at 5 seconds, but they won’t think it is slow either. This test shows at five seconds, people will think your website’s speed is medium—middle of the road. That is somewhat comforting, but we still want to fix the situation. If we can’t fix the technical speed, can we at least improve the website’s perception of speed?

One of the best recommendations to improve the perception of speed is the inclusion of load indicators. Nielsen Norman Group published an article in 2014 offering this conclusion regarding indicators:

Users experience higher satisfaction with a site and are willing to wait longer when the site uses a dynamic progress indicator.

However, the kinds of load indicators discussed in that article are for actions that happen after a visitor has arrived on a website and then initiates a type of task. The article gives an example of visiting United’s website and looking up a flight once there. In these cases, people are already on the website and have asked the website to do something. In these instances, a load indicator can help to alter the perception of speed. Would the same be true if we showed an indicator while waiting for the initial load of the website?

What this test suggests is that at load times under 3.5 to 4 seconds, indicators don’t make much of a difference in the perceived speed. Starting at a 4-second load time, the perception of the load time improves, slightly, when a load indicator is used. For people using smartphones, the impact of the indicator is somewhat more pronounced. However, few of the differences between load indicator types are statistically significant at any load time.

Our takeaway from this, then, is that it may not be a bad idea to use a load indicator if your website does take longer than 3.5-4 seconds to load. It might help your visitors make it seem like your website loads a little faster than its actual technical load speed. However, the difference is very slight so if an indicator presents other challenges or is costly to develop it likely is not worth the investment.

If an indicator is used, the type of indicator does not matter much. In this test, we evaluated a simple text indicator that said “Loading…” and a spinning circle indicator. Generally, there wasn’t much difference between these two indicators until the 5-second mark where it appears that the circle might have more of an impact on the perceived load time. However, this difference is minimal so if an indicator is used, pick one that is most appropriate for your website and brand.

Mobile versus Desktop

Prior to this test, we also thought people using mobile devices were less patient than people using a desktop or laptop computer. We also thought people using a mobile device would think time is moving slower while they stare at their screens waiting for the website to load. As a result, we fully expected participants using mobile devices to have significantly lower perceptions of load time. This simply is not the case.

The data shows quite clearly the way people perceive speed on mobile devices is not very different from the way people perceive speed when using desktop or laptop devices. At .5-5 second load times, where we had more participants allowing for a more robust sample size, the median difference between desktop and mobile perception of speed is -0.01. The biggest difference in the perception of speed comes at 3.5 seconds where participants on mobile devices rated that speed as “Fast” while people on desktop or laptop devices rated that speed as “Medium”. If anything, this difference argues directly against our hypothesis as it indicated mobile participants were more forgiving of the slower speed.

This data strongly suggests something very different from what we previously thought. Slow load times are bad for everybody and in pretty much equal measure. Mobile visitors do not seem to be any more or less impatient. Mobile visitors do not seem to be any more or less likely to think time is moving slower while waiting for a website to load. The conclusion we reach, then, is that our mobile websites should not be any more deserving of an investment in speed than our desktop websites.

Load TimePerceived Speed – MobilePerceived Speed – Desktop
0.54.694.54
14.464.37
1.54.224.4
24.324.09
2.54.194.26
34.044.2
3.54.163.82
43.973.85
4.53.733.94
53.683.78

Longer Load Times

Our initial test’s hypothesis was that a 5-second load time would be considered “Slow” or “Very Slow” and this hypothesis was definitively (and surprisingly) rejected. As a result, we wanted to know where people would begin rating a website’s speed as “Slow” or “Very Slow”. To determine this we ran a follow up test to evaluate load times between 5 to 10 seconds in half-second increments. This test went to fewer participants than the main test, but there are still several interesting conclusions.

Still Not Very Slow

The most intriguing finding is that even at these longer load times, the mean perception of speed still was still not rated as “Very Slow”. For people who waited 10 seconds for the test website to load, they only rated the perceived speed at 2.58, which is in between “Slow” and “Medium”. Prior to 8-second load times, the mean perception of speed was still 3.10, which on this test’s scale is “Medium”. This result surprised us and ran counter to what we had expected to find.

Indicators Still Don’t Shift Perception

At these slower load times, the indicator continued to make only the mildest of differences. Perhaps, a simple text indicator helped improve the perception of speed better than a circle indicator or no indicator at all, especially between 8-9.5 second load times. As with the main test, this suggests using a load indicator may help with the perception of speed and likely will not hurt the perception of speed either. However, to be clear: a load indicator is not going to make a 9-second load time feel like 3-second load time.

Blaming The Company Behind the Website

At the longer load times, participants were more likely to blame the website for the slower speeds than they had been in our primary test. That difference is statistically significant. It is important to note the question asked didn’t relate specifically to who participants blamed for the speed of the test website but instead asked more broadly about who people blame when they encounter a slower website. Still, a 42% increase in people saying they blamed the company running the website exclusively is a concerning sign for any company whose website runs this slowly on mobile or desktop devices.

Mobile and Desktop Continue to Be Similar

A surprising find at these slower speeds is that there continues to be no difference between perception of speed between participants using desktop and mobile devices. For example, at a 10-second load time, participants using a mobile device gave a mean score of 2.47 (“slow” to “medium”) while participants using a desktop or laptop computer gave a mean score of 2.68 (“slow” to “medium”). The biggest difference is at 6-second load times, where participants using desktop or laptop devices rated the speed as “medium” while participants on mobile devices rated it as “slow” to “medium” instead. However, the perception at a 6-second load is the exception—the median difference between desktop and mobile is only 0.085.

Load TimePerceived Speed – MobilePerceived Speed – Desktop
5.53.443.17
62.873.5
6.53.743.26
73.473.24
7.52.793.35
833.24
8.53.152.66
92.572.76
9.52.812.94
102.472.68

Overall Conclusions for Longer Load Times

Obviously, you don’t ever want your website to load this slowly. However, these longer load times are not perceived as badly as we had initially expected. Of course, even if people do not say 10-second load times are “Very Slow”, there is still a low perception of speed from participants at these speeds. People do not say 10-seconds is “Very fast” either. At these load times, you have to speed up your website if for no other reason than the good of your users.

Future Questions

While we are excited and surprised by this test’s results, it is important to note some limitations in this test. First, this test involved loading a webpage that contained some basic branding, an image and a few lines of text. It is possible that the way people perceive the speed of a website could change based on the type of content contained. Second, this test involved people loading a website directly but it is possible that people loading a website from a search result, a social media feed, or an email might perceive time differently as each of these sources could shift user intent. Future tests evaluating perceived speed by content type or visitor source would be interesting and beneficial.

An open question is what load time will finally encourage people to say a website loads “Very Slow”? We are frankly stunned that people rated 10-second load times as only “Slow”. It would be interesting to explore the perception of even longer load times.

Finally, this test looked solely at the visitor’s perception of speed. The load time of your website affects more than the user’s perception of speed though—how fast your website loads affects SEO, CRO, server management, and more. Other tests have been conducted and articles have been written discussing those impacts and we encourage you to seek those out as you embark on a speed optimization project.

Methodology

Participant Size

This test was conducted with 1,607 participants in the primary test (0.5 to 5 seconds) and 444 participants in our longer load times test. All tests were conducted between January and April 2018. Participants were paid 20 to 50 cents for loading the test website and were located throughout the world, with a heavier concentration in the United States. Of participants in the primary test, 60.98% used a desktop device and 39.02% used a mobile device and in the longer load times test, 52.25% used a desktop device while 47.74% used a mobile device.

Test Setup

Participants were asked to review a website but did not know in advance the test was about a website’s load time or perceptions of speed. After reading the instruction page, participants were asked to press a start button to begin the website review. While reading the instruction page, the test website’s files were fetched in the background, allowing us to load the website as soon as the participant initiated the test.

Once participants pressed the start button to begin the website review, a timer controlled the participant’s total load time of the website. This allowed us to control exactly what the participants total load time was and also track what that time was.

Total load times within the primary test ranged from .5 to 5 seconds, in half-second increments and participants were randomly assigned a load time. In the follow up test, load times ranged from 5-10 seconds and, again, load times were randomly assigned. As well, all participants were randomly assigned a load indicator of text, circle, or no load indicator.

Questions Asked

The following questions were asked of participants. No default answers were pre-selected.

How would you rate the speed at which this website loaded?

  • • 5 – Very fast
  • • 4 – Fast
  • • 3 – Medium
  • • 2 – Slow
  • • 1 – Very Slow

When you encounter a slow website, who are you more likely to blame for the slower speeds?

  • • 1 – Your internet provider
  • • 2 – The company running the website
  • • 3 – Both
  • • 4 – Neither

Requests for Raw Data

We are happy to share the raw data upon request. Please direct requests for raw data to info@elementive.com with the subject line “Speed Perception Test Raw Data Request”.

About Elementive

Founded in 2014, Elementive is a Colorado-based consulting firm that specializes in helping clients optimize their website to increase conversions (CRO) and improve their website’s user experience (UX). Clients include small business, startups, non-profits and Fortune 500 companies.
Through data analysis, research, measurement, audits, and continual experimentation and testing, Elementive helps our clients find the best ways to get more clicks, calls, leads, sales, or whatever conversions they are hoping to achieve. Services include ongoing and one-time consultation, as well as education and guidance. Learn more about Elementive.

Matthew Edgar is a partner and consultant at Elementive. Author of the book Elements of a Successful Website, Matthew has worked in the web performance and analytics fields since 2001. He has spoken at numerous events about CRO, UX, and analytics including IA Summit and MozCon. He regularly teaches workshops about analytics, including teaching courses for O’Reilly Media. He holds a Master’s of Applied Science in Information and Communications Technology from the University of Denver. You can connect on Twitter @MatthewEdgarCO and read his blog at MatthewEdgar.net.

Andrea Streff is a partner and consultant at Elementive. She specializes in mining marketing data and analytics to help clients learn who their customers are and what customers want from their business. A former “jack of all trades” marketer, Andrea uses her background and expertise in direct, online and inbound marketing to help businesses bridge the gap between their marketing analytics and marketing strategy and tactics.

You may also like

Using Log File Analysis To Improve Your SEO Performance

You have to understand how Google’s bots are understanding and crawling through your website. Learn what log files are and how you can start analyzing your website’s log files to improve your SEO performance.

Measuring How Speed Impacts Visitors

Are you getting the full story about page speed on your website? Along with load time metrics, we also need to know how speed affects our visitors.

How To Find & Fix Duplicated Content

Put simply: Duplicate content confuses both human and robot visitors. Let’s walk through how we deal with duplicated content: locating, evaluating, and resolving.