If you’re not A/B testing your website chances are you’re losing revenue. Here’s how it works.
For the next month a simple web-based tool sends 50 per cent of your visitors to your usual landing page and the other 50 per cent to a slightly different landing page. The results are analysed for conversions and the winning page becomes the site’s new landing page. Then the process starts over again with another new page containing another small change as the second option.
As the test continues all year long, particularly with vital pages of your website such as landing pages and shopping carts/checkouts, the site’s performance increases incrementally. Rather than standing still, the site changes and adapts as market forces and web trends continue to alter the way people interact with e-commerce sites.
“Sometimes when you deal with very small changes you can get an increase in conversion rate of three to four per cent,” says Costantino Marotta, General Manager of Digital Services and Performance Marketing from The Gruden Group. “It seems so little but it may result in thousands of dollars of extra revenue each week, and all it took was a very small change in the layout – the colour of a button or a new testimonial.”
The practice is known as A/B testing. It is simply about testing – in real time and at the same time – two (or sometimes more) versions of a website page to figure out which earns the best results. The versions should only be mildly different from each other, Marotta says. If they differ drastically, it becomes impossible to identify why a certain page performed better.
“The idea is to isolate small changes and identify which component is resulting in that increased conversion rate. Then you build from there,” he says.
How A/B testing works
There are several options around A/B testing. Organisations such as The Gruden Group can be contracted to manage the testing and analysis process on behalf of a business.
Otherwise, Marotta says, several web-based tools are available for those who prefer the DIY option. In simple terms, these tools create a clone of your landing page or checkout and allow you to make a small change to the cloned page. It’s up to you then to decide which percentage of your traffic to direct to one version or the other, with results tallied at the end of the chosen time period.
“You need to have a statistically significant sample of users,” Marotta says. “That is the only way to truly discover which one is the winning version. Normally the amount of traffic will define how long a test needs to run. It can be anything from one week to two months or more.”
Results of A/B testing
The change injected into the cloned page might simply be a different colour, a larger button, a new location for a button or an offer of free shipping.
Some of the most dramatic results, Marotta says, come from simplifying and easing navigation through the checkout process.
“We work quite a bit with our clients on the different steps of a checkout process,” he says. “When you go to your shopping cart and take the journey to purchase, there are a certain number of actions. We try to reduce the number of actions or steps. That is when we see the biggest impact.”
He continues, “There is research that shows that the higher number of fields to fill out, the lower the conversion rate. So sometimes success is as simple as getting rid of some of the optional fields or using auto completion for a shipping address. Within one field you might use the Google API for the address, and that can save the user from having to fill in seven fields.”
Never stop testing
A great e-commerce business, Marotta says, will make testing a part of their year-round practice. It means the business’s shopfront is constantly improving and always responding to user preference.
“Never stop testing,” he says. “There are always trends and the web is evolving continuously. And the way we utilise the web is evolving continuously, too. For that reason we should be constantly testing and striving to improve our web pages. When you do it well it is rewarding on a number of levels.”
The difference between A/B testing and multivariate testing
While an A/B test is testing two different versions of a web page (or you can A/B/C test with three pages etc.), multivariate testing is a subtler test of multiple elements within one web page and how they interact with each other. The point of this is to give you a sense of which elements play the biggest role in helping you to achieve the objectives of the page. Multivariate testing is like multiple A/B tests being conducted at once and is a more complicated test to manage.
Ben Thompson, CEO of HR and employment cloud software developer Employment Hero, has enormous respect for the A/B testing process.
“A/B testing is one of our most effective weapons in the marketing team’s armoury. It helps us perfect our display campaigns, EDMs and our web pages to resonate with the target audience. We have had great success with our website pages in the way they are laid out on desktop and mobile browsers. In one case, when the boardroom was divided on one very important aspect of the web page, our marketing team went with both options, tested it over two weeks and came back with heat maps and conversion tracking to showcase how the web pages worked with the audience.”