Our new Ask an Optimizer column addresses questions we’ve received through our website and from members of the MarketingExperiments Optimization LinkedIn Group. The following question was submitted by Nikolay M.
Q: Does anybody have experience with load speed impact on conversion or bounce rates?
This is a very good question Nikolay, one that many people don’t think about. Because this area is something we address with our tests and treatment versions as a best practice, I don’t think we have case studies specific to this issue. However, I can offer insights on factors to consider and resources to use.
The first thing you should do is assess two items:
- Traffic sources
- Technology resources for your audience
Know your top traffic sources
More important than a list of tips or statistics is understanding how the majority of your visitors arrive at your website.
If you aren’t keeping an eye on this, it’s possible to spin your wheels trying everything when stepping back and understanding your audience could save you serious headaches.
Where users are coming from and how they arrive at your site or funnel pages will often dictate how long they will put up with the nuances of your site. For example, visitors arriving from searches (PPC and organic) are likely to be among the least patient site users. Think about it: they have just seen a sizable list of alternatives, so if your pages are confusing, overwhelming, or take too long to load, those visitors are out of there. I typically find these users (especially from Google) to be back-button happy.
Know the technology that the bulk of your audience is using
Through analytic tools you can get a good idea of what technology resources users have at their disposal. For instance, with Google Analytics you can see connection speed, operating system used, and a whole host of other items.
Use these insights to see how much wiggle room you have with how heavy your pages and website can be with content elements. If a large portion of your users are on dial-up or even DSL, or older operating systems, then you might want to increase efforts on CSS-based quick loading pages.
Now, some people will be stuck with a heavy website due to factors outside their control. If that’s the case, look to edit the order in which your website loads. Have some of the lighter elements load first, so there are at least some items on the page while the heavier elements are being compiled. This gives users a glimmer of hope that the greatest part is on its way. However, make sure the initial loading items aren’t information collection or other friction points, because users can be scared away before the incentives or explanations load.
Another element to try in this situation is an animated loading bar. This again shows the user that the site isn’t broken, just taking a while to load. I think people are a little too quick to dismiss this method, but feel that users have a greater propensity to stay if they know items are happening in the background.
Best practices and resources
As a general rule, we like to keep page load times under 8-9 seconds for 56k users. Shorter than that is even better if you can do so without sacrificing site quality or functionality. This covers a wide audience, and seems to fit the typical short attention span that internet users have.
For more on this area, I’d recommend WebsiteOptimization.com, which has several helpful resources, including an excellent Web Page Analyzer Tool that will scan a page, let you know load times, and give you tips to how to improve your page load times.
There is a number of case studies from aptimize.com here
http://www.aptimize.com/customers
and we have our own stats about general bounce rate and conversion (for http://code.google.com/p/web-optimizator/ ) — but there is no numbers for PPC pages. We are looking for any collaboration to run split testing in this area and get real numbers to show them to the industry (I forecast about 5-10% influence).
Sunnybear,
Thanks for the response. I was wondering if you have any more specific data on the case studies that you listed? For example, load rate before was, “X” and after our services it was, “Y”.
Also, what type of pages were these? Were they site-wide? How was testing done? Were there outside marketing efforts going on during testing that might have affected the results on these pages? Please do not take this as me trying to diminish or nitpick your work. I only ask because you are one of the few that has published statistics on this that I am aware of, so I am interested in learning more.
I will also discuss with our team internally the possibility of teaming up on a case study regarding PPC Page Testing. I will say however, if your results were because of optimization efforts on content style pages, the gains on PPC could be larger. This is mainly due to the fact that PPC traffic tends to be less patient than other channels. So if a process is truly slow and leaking users, having the ability to serve pages more quickly could drive significant results. If you want to send your contact information to corey.t (at) marketingexperiments (dot) com, we can connect about teaming up for a possible partnership.
Corey Trent
Twitter: @ctrentmarketing
Some time ago recall reading some states that suggested that the best sites are delivering load times averaging around 2-4 seconds. So I went out and looked for some reporting and sure enough http://www.webmetrics.com shows 100 sites for this past week and the top fifty all load faster then 5 seconds. I’d make that the goal….
Yahoo offers some terrific optimization recommendations at:
http://developer.yahoo.com/performance/rules.html
Also I really like the report that websiteoptimization.com offers this would be a great place to go to identify changes you’d do in an AB test.
Tom,
Thanks for the great response. I would caution in putting so much weight in the top 100 sites. I think it is a good indicator, but with these sites being so brand heavy users are going to put up with more. Many times the top sites are optimized not for user experience, but rather for IT and available resources issues.
The Yahoo link you provided is great resource as well.
I am glad that you found value in the resources at http://www.websiteoptimization.com. You are right the report does give a nice action item list for testing.
Thanks again for the great contribution.
Regarding load time, recent research from Akamai has finally put some specific numbers to the question of how fast is fast enough.
http://www.akamai.com/html/about/press/releases/2009/press_091409.html
The important points are that customers become impatient after just 2 seconds waiting for a page to load, while a full 40% of consumers won’t wait 3 seconds and will simply abandon the site for a competitor.
The really interesting thing is that these numbers are significantly lower than a few years ago which means that consumers are getting less patient with website load times.
@WebsitePerformance
Thanks for the link to the study summary. I believe this puts further weight behind making site performance a priority. There are a couple points in the study that I would like to know more about:
1) On the summary they mention this data was collected in survey form. Depending on how this survey was conducted and to whom, I would say that what panelist communicate might be slighted disjointed from expectations or actions in a real world environment. When you are being asked by a moderator, in a group, or removed from the process in a web survey two seconds sounds reasonable. I expect my stuff and NOW, because I am ready to buy and this is 2010. It would be time consuming, but what would be interesting is observing real browsing/shopping to see how those figures differ.
2) What would also be interesting to see is if users relax on this alleged two second rule, if they see some aspects of the website populating. So testing if our page does not or cannot load in two seconds, would re-arranging code so some web elements appear while the heavy lifting is still going on keep more people in the process. My assumption is yes, by the way.
3) This study and data does not take into account entrances sources. If I am typing in an address because my friend told me, “You have to see this product because it will change you life”. Then I am going to wait longer with much anticipation. If I am comparison shopping for a product and arriving from search results filled with alternative sites, I am much less likely to put up with loading shenanigans.
In my mind if this site is not working, then to the back button and a competitor I go. So does this mean for sites that get a lot of direct or type in traffic loading is not important, absolutely not. But there is a difference in the types of users coming to your site and you need to be aware of their expectations on your business.
It is good to see that the internet is maturing over time, and users are demanding more. It can make our job a bit more difficult, but it also presents an opportunity for use to shine over competitors as well.
Thanks again for the article.
Good job! Can’t wait to start my own blog.
Hi there
I’m just wondering, my site loads in about 7 seconds, I’m running Jshop, I’m not sure if the slow speed is effecting conversion, any ideas?
I’m thinking that I might dump it out and put it on another server, see if the slow speed is a result of the shop configuration or the server space I am using.
The site is: http://www.digidave.co.uk
I’m pretty sure the config is good, but no harm in trying, right?
Luke