Case Study: Increasing Traffic By 37% Using Split Testing
About The Author
Hi, my name is Mark. I am the owner of Think Traffic and I just love internet marketing and web dev.
I started my career by learning to write HTML and PHP, since then I have fallen in love data and traffic analysis and learning to understand how and why users interact with the web. I enjoy working with small businesses too.
Latest Blog Posts
Think Conversions On Twitter
Contact Us Now
How much time do you spend on generating traffic for your website? Ok, and how much of that traffic actually stays on your website for more than 30 seconds? How much of it actually explores your website?
This post is a case study to show you the power of conversion rate optimization and how we have used it to make the ThinkTraffic website better.
Why Bounce Rate?
There are a number of things that you can test using conversion rate optimization. We decided to test for bounce rate in this experiment because it is one of the easiest ways to get a big win.
One of the advantages to testing bounce rate is that your “conversion rate” will be relatively high (in this instance a conversion would be any action taken), which means that you don’t need to run the test for as long to get a usable result. As you will see below, we got some big improvements very quickly.
What Did We Test?
Before we started our test we had a look at our Analytics data so that we could make some informed decisions. This is important for two reasons:
- You should test pages which are under performing
- You want to test high traffic areas of your site
We discovered that approximately 38% of our traffic was coming into the site via our home page and another 58% was coming in via our blog posts collectively.
So we decided to run two tests. The first test was on the home page and the second, we ran on our blog post page template. By doing two simple tests we were able to test pages responsible for 96% of our total traffic. This meant that any improvements would have a significant impact.
Testing The Home Page
When you design your first conversion rate test, it may be difficult to know what to actually change. You can test small tweaks, or you can test complete site redesigns.
Our home page had been pretty much the same for a while and we felt we had a good feel for what was and wasn’t working, so we tested a couple of complete redesigns with some smaller variations just for good measure. Here’s how our home page looked before the test:
We had a few criticisms of this design. Firstly, there are three call to actions; the first is the free consultation form, the other two are the “Get more traffic” and “Link building” sections lower on the page.
We know from experience that the consultation form is not that effective; most of our serious leads come via our full contact form. So we decided to scrap the consultation form and focus on getting the user to click through to one of our services pages.
We came up with a couple of different designs, you can see the eventual winner below:
We tested having the images both ways around and we also tried with and without the testimonials box. Surprisingly the testimonials had a relatively small impact – we think this is probably because the home page isn’t really a sales page. If you click either image you are taken to a page explaining that service, and the testimonials on that page do appear to be much more effective.
Anyway, as you can see, the focus is now very clearly on getting the user to click on one of two boxes. We have also kept the recent blog posts below the fold. These add some content and keep the home page fresh, which is important.
The funny thing is, the original version probably looks more professional (in our opinion at least), but the new version is simpler and more focused on getting the right action, and that seems to make it a winner.
Testing The Blog Posts
For the blog posts the challenge was slightly different. Most traffic coming onto our blog is not looking for SEO services, so a direct sell will obviously be less effective. We had to have a think about what we want to achieve with our blog.
The main purposes of our blog are
- To attract links
- To build credibility
- To generate long-tail traffic
- To communicate with clients
In the end we decided that our priority for the blog is not to get people to contact us, but simply to keep visitors on our site. Even if those visitors are unlikely to become clients, keeping a visitor for longer maximizes the chance that they will link to us or tell someone else about us.
The critical part in any blog post (assuming the content is good enough) is the end of the post, because this is where a reader will either click further into your site or leave never to return. Here’s how the end of our blog posts looked before the test:
Again we tried several different tweaks, the single most effective change though, was removing the consultation form at the bottom of the blog posts and replacing it with a recent posts widget.
We also tested some other ideas, such as including a Twitter widget at the bottom of the posts Vs A generic social sharing widget, installing Comment Luv to reward our contributors etc… Here’s how our most effective variation looked:
Of course, we have done some more tweaks since then. As you can see, all of those tests have resulted in the page that you are currently looking at!
When testing the blog posts our best performing variation showed 37% better engagement than our original version. But we actually combined a couple of the successful variations to create what you now see. All in all, our engagement has increased by around 43% as a result.
On the home page the results were almost as good, engagement on our best variation was 35% better than the original!
How Does That Translate?
Engagement is effectively the inverse of bounce rate. But to give you an answer in plain English:
Before the testing, the average bounce rate across all of our site was 62.7% over the 30 days before the test. Since implementing all of the changes our bounce rate overall is now 48.9%
What Does This Mean Really?
Before the test, for every 1000 visitors we got, 627 were leaving. We were only actually engaging 373 of those 1000 readers. Now we are engaging 511 of them!
Our bounce rate has decreased from 62.7% to 48.9% that’s a 20% drop. The result is effectively a 37% increase in traffic!
Give Us Your Thoughts!
The great thing about these results is that the changes are permanent. We will get that extra 37% boost every month from now on, and we will of course keep testing and improving our site.