For well-established sites, it’s almost always quicker, cheaper and easier to lift conversion rates by 30% than to bring in 30% more traffic. Then when we do lift the number of visitors, you’ll get more value from each one. The gains are multiplicative.
How do we do that?
There are lots of reasons people might not convert—and a conversion can be a sale, submitting a contact form, making a phone call, downloading a PDF, a social share or any other action with a benefit to you. It might be that the viewer isn’t quite at the right point in the buying cycle yet or they aren’t a perfect match for your product. With SEO and AdWords, we seek to bring you the right visitors, not just greater numbers.
But what about the visitors who are almost there, but not quite converting yet? The browsers, the uncertain and undecided, and the window shoppers—it’s these people that can be tipped over the edge with Conversion Rate Optimization.
The first step is to identify potential points of friction and points where value can be added. Can people navigate every step of the conversion process easily? Are important calls to action well signposted? Are you showing people the benefits of conversion?
Our highest ROI service
Web analytics can be of great help in identifying parts of the user conversion process that are underperforming, and we can also identify potential problems by auditing your site against known best practices. Sometimes it’s also useful to deploy additional data gathering, for example in the form of heatmaps or clickmaps, to help us understand how people are using your site right now and where the weaknesses are.
We consider a full conversion rate optimization audit to be the fastest route to big gains for almost all clients. We’d quote some ROI figures from past projects but some of them are almost unbelievable.
Split testing
You don’t have take our word for it. We almost always recommend split testing to determine if a variation on your page will be more effective. Here, (in the case where we are testing just one variation) a randomly selected 50% of viewers see the original and 50% are shown a fully-functional variation.
The variation might include color and design changes, altered messaging, additional (or removed) material, elements that relate to the users particular geographical location or device type, and more. Performance data is gathered both for the control and the variation/s.
Statistical significance is a concept close to our hearts, and we like to have sufficient data to report results with at least 95% significance. Whether the change we see is 5% or 45%, you can be sure the results are statistically meaningful.
Example test, data + best practices
Fostering trust is a central tenet of conversion rate optimization. This goes double where potential investment and risk are both large—as in most high value B2B sales. One common strategy is to use human faces. A smiling human fitting into a similar demographic as the reader immediately fosters feelings of kinship and trust at a conscious or subconscious level and can be a powerful influence on user behavior.
With this in mind, we tested the addition of a picture of a smiling engineer to a page designed to appeal to engineers looking for an industrial nozzle supplier—the image shown above. Let’s call him Nozzle Guy. This picture ticks all the best-practice boxes for this kind of trust building exercise: a similar appearance to the buyer, displaying pride in a flagship product, in color, smiling and professional. The presence of Nozzle Guy helped us raise conversion rate through to the point of contact by 18% (significant at 99% confidence).
But careful attention to data had showed us something else. In display advertising, it wasn’t the human faces leading click-through rates. It was the image to the right. Let’s call him Spaceman (he’s not in space but Heat Proof Suit Man is much less snappy).
From a best practices point of view, Spaceman isn’t quite right. You can’t see his face. But he does fit with an engineer’s view of a very cool image and a strong signal of technical expertise. Adding Spaceman to the page lifted conversion rate to contact by a massive 56%.
In this case, a combination of best practice testing, diligent data analysis and consideration of the client’s particular niche led us to an excellent result.
Personalization
Different users might be better served by different design, different messaging, or even different features or pricing schemes. We can build a system that will show a tailored variation to a particular group of users, and only that group.
How our software works
Although we have plenty of experience in other packages, our split testing software of choice is Visual Website Optimizer. This creates a kind of overlay that loads along with the page itself, changing the appearance where the user is in the variation group. The viewer doesn’t know they are looking at a test page, and if they see variation A on their first visit, cookies ensure that they’ll see variation A if they return tomorrow.
Tests can be applied across a single page or group of pages, or site-wide. For example, we might like to make a change to a site-wide header. If a user sees variation A on their landing page, they’ll see variation A throughout the site.
Using VWO means that most of our tests can be run without involving your web developers (and therefore without incurring development costs) once the code is installed. Installation is simple. A small snippet of javascript needs to be placed on all pages, much in the same way the Google Analytics snippet is deployed. For many CMS including Magento, Drupal and WordPress, a simple plugin is available to do the hard work for you.
Types of test
Split tests come in many varieties: mobile and desktop, simple and complex, or geographically targeted. We can test page content, images, pop-ups, form fields, page design and more.
Here are some of the other test types to be aware of:
A/B tests
In an A/B test, there is one control page and variations created through VWO. The number of variations depends on what it is we are testing and your traffic—the more variations, the longer a test can take to come to a statistically meaningful conclusion when traffic is limited. We usually recommend between one and five variations.
Multivariate tests
Multivariate tests are a little more complex. Instead of testing version A against version B, we’ll nominate a series of elements to be included in a single test, and with the aid of some mathematical magic, these individual elements will be shown in various combinations until we can draw solid conclusions about what’s useful and what’s not.
Split URL tests
Split URL tests are a type of A/B test, but instead of building variations through our software, an independent live URL is created for each variation. These are excluded from site navigation and have noindex, nofollow tags to keep them out of search engine reach.
Our software is given the URLs, and handles redirecting viewers to a particular variation or letting them see the control.
Split URL tests can be useful when you want to test a completely new page design, or when your developers want to be the ones building the variation.
