Our lives are full of tests. We use products that were consistently tested before hitting the shelves. We drive cars that were test crashed in order to find construction flaws and improve safety. We use scientific advancements that were nothing more than hypotheses tested by their creators.
You test stuff yourself. You go into several restaurants, try their food and pick the one that gives the best bang for the buck. You test a company’s service and if they deliver on their promises, you order more; if not, you move on to the next one.
Testing is a natural part of our lives. We call it "let’s give it a try," "let’s try," "lets see how it goes," etc, but we constantly test everything. Once we find something that passes the test and delivers in line with our criteria, we use it. This is common sense.
Why not do it on the web? Why do we create a design, throw in some content and let it stick, and then complain about the horrible 1.3 percent conversion rate? Well, there’s no magic to increasing it – only well-planned tests, backed up by strong theory (hypothesis).
In the last article I brought up the topic of testing, its role in conversion and how it works together with search engine optimization. In this article we will go deeper into testing and explore A/B tests, mulitivariate tests, and how to know when you’re ready for testing.
Am I Ready to Test?
You are ready to start testing if:
You consistently get a satisfactory amount of traffic (by your standards) from search engines/PPC/referrals with a low conversion rate (the Internet benchmark is ~2.5%, however through constant improvement we can raise that number to 7% – 15%).
Despite your previous efforts to improve your conversion rate — changes in marketing strategies, copy writing, messages, website design, website flow, landing pages, etc – you still cannot increase the conversion rate or get desirable results.
Your competitors are getting more customers from online channels, even though you invest as much or more than they do.
Your investment in search engine optimization brings you traffic, but not profits. (Be careful, since this can mean that you are optimized for the wrong or too broad phrases).
If you drive traffic to your website, but still struggle to make a good profit, it is time for you to do some heavy conversion testing.
There are two primary classifications of tests: the A/B Split Test and the Multivariate Test.
A/B Split Test
The A/B test lets you compare two versions of an element on the page against one another and select the one that performs best. This is the simplest form of testing and it is a great place to start.
Let’s say you want to determine which call to action works better, a link or a button. You create two pages, one that has a call to action as a link and another as a button. You run the test and measure the conversion rate each page delivers.
If the call to action in the form of a link delivers a better CR, then use it rather than a button. In the next phrase, test different wording for the link – the original wording and new wording. Select one that works best and run further tests.
Keep testing until the conversion rate dips, and then simply go back to the variation that gave you the best results.
Multivariate tests are much more complex than A/B tests and require software to successfully run (further information on this below). In multivariate tests you test various page elements with different variations.
For example, you have three calls to action on a page that you want to test: one on the very top, one in the middle and one at the bottom of the copy. Let’s say you want to test six variations on each call to action (colors, wording, font size, graphics, etc). This gives you 6 x 6 x 6 = 216.
The test will determine the best performing combination, but is impossible to run without software and requires a considerable amount of traffic/time.
You can run A/B tests manually, but when it comes to multivariate testing you need software. I have a detailed review of Website Optimizer coming soon, but here is a list of software programs you can use for multivariate testing.
Google Website Optimizer
A/B Split Test Pro
ASP Hosted Applications
Vertster Conversion Rate Testing
Duncan Carver’s Scientific Internet Marketing Assistant
Phil Huff’s Split Test Generator 2
SplitHit Professional Split-Run Testing Software
CGI Programs using SSIs
Marty Foley’s Scientific Web Marketing System
Sales Page Master PRO
Raymond McNally’s Split Test Creator
Ad Tracking Programs with a Split-Testing Feature
Add2it Go-To Pro
Dedicate exclusive time to conversion optimization and testing. Run a search engine optimization campaign first, and when the rankings are there and you get a consistent flow of visitors – start testing. You might want more traffic and more visitors, but work with what you have at the moment. It is better to optimize and streamline conversion before you hit the sweetest search spots, because revenue differences can be dramatic.
For instance, if you get 100 visitors per day that convert at a rate of 2.5% (2.5 visitors per day), with $100 per conversion, then your profits are $250. If you take the time to test and improve conversion for a few months, and say hit 7%, your profits grow proportionally to $700 without changes in traffic and rankings!
When the time comes for more SEO investment, you will be much more prepared and get a lot more money back. Let’s say from 100 visitors per day you increased traffic to 300 visitors. At a 2.5% conversion rate this translates to $750, and at 7% the number is $2100!!! That’s a dramatic difference! This is the difference between you driving a Ford or a Porsche.
Do testing with what you have at the moment. (of course there must some traffic on the website, at the very least 30 – 50 visitors per day).
Things To Watch Out For
Testing Practices – Also known as "they do this, so I must do this as well." What works for others will not necessarily work for you. You have a different business, with different goals and different customers. Superbowl ads work for McDonald’s because it has a brand to maintain and must remind people where they need to eat. Superbowl ads won’t work for you because they will eat up your entire advertising budget with no guarantee of long-term success.
Testing the Wrong Parts of the Page – Sometimes one element may be preventing a higher conversion rate while you focus efforts on another element. For example, a ridiculously long form might be stopping users from finishing the conversion process, while you’re focused on testing calls to action. In the end you might abandon the test, blaming calls to action for not delivering the goods, while it is the form that needs changes. Don’t forget to look at the analytics and exit pages. Identify and fix problems first.
Replacing Page Elements with Weaker Versions -You might replace an image with a weaker version that does not give users the perspective they need. As a result conversions dip and you put the old one back without running further tests. Don’t give up to early. If you’re met with lower conversions, remember what you did and don’t do it again, but continue testing.
Conversion testing needs a plan, a predicted outcome, just like a scientific experiment needs a hypothesis. Look at the test as if it is a scientific experiment. Do not expect much; rather, approach it with curiosity – "I predict this will happen if I do this action." And if it doesn’t happen… well, you were wrong and you need to come up with a new action. So a website test has several elements:
You need to name the variables for the actual test. Variables such button colors, copy writings, font sizes, etc, are equivalent to A chemicals + B chemicals + X chemicals in a scientific test. Look at yourself as a scientist making a compound with the goal of proving a hypothesis.
The second part of the test is the hypothesis, a predicted, or more correctly in our case, a desired outcome. The hypothesis will usually be something like this: "I will increase the conversion rate by 50% if…." or "I will reduce the shopping cart abandonment rate if…"
The third part of the puzzle is simply connecting the hypothesis with an action plan.
In the lab:
Action: I mix X + Z + A chemicals and boil them at 100 C. Hypothesis: The end solution will be able to burn through metal.
On the web:
Action: I change the wording on the call to action button, change the color to blue and increase the font size. Hypothesis: I will be able to increase the click through rate to the form by 100%.
Online tests in the multivariate format can have hundreds of actions at once – all working for one hypothesis.
Putting A/B Testing in Its Place
A/B Split and Multivariate Testing Software Reviews