6 Tips for Successful A/B & Multivariate Page Optimization Testing

Page optimization and website performance testing often gets overlooked by design, inbound marketing, and advertising tactics. However, it has been shown to be one of the most cost-effective tactics you can employ to increase the effectiveness of your website and marketing/advertising campaigns as a whole.

A/B Testing is testing two different designs against each other, randomly displaying one page for one user and the other for the next. Multivariate testing allows for testing combinations of elements on the same page (for instance, a headline and picture can be tested to see which works best together). Multivariate testing requires more traffic, but both can substantially improve ROI.

It pays to pay attention to the details. Our most recent tests included a 55% increase in form conversions with a home page redesign, a 30% increase in form conversions just by changing the text on a button, and a 91% increase in subscriptions by redesigning a sign-up page. This sign-up page improvement you can see below. Seemingly simple and intuitive, but most websites just don’t do it.

Interested in doing your own page optimization? If you aren’t going to hire a professional, then here are 6 tips to keep in mind when planning your test.

1. Come Up With a Plan

Don’t just “go with the flow” and test whatever you feel like after each test concludes. Come up with a testing plan. Know about how long the tests will take and lay out a schedule of testing page elements. After all, this is supposed to be a scientific method, so try to at least loosely resemble that scientific method process they taught you in 8^th grade. Plus, it will keep you organized and ensure your results are accurate.

2. Big Changes First, Small Changes Second

This fits in pretty well with coming up with a plan. It’s important to change large elements first, then move to smaller elements. It doesn’t make sense to test button text and then completely redesign your page. Those button test results are only viable for that original page, not necessarily the redesign. This may not seem entirely intuitive, but that’s how it works. It’s tough to tell why people act they way they do online, so it’s best to start with a page redesign if that is one of your tests, move to larger page elements like pictures or slight layout modifications, and end with headlines, button texts, and form fields.

3. Never Assume

Don’t assume because your left form worked better on your other design that it will work better this time. Don’t assume because you like a picture better your visitors will, or even that a picture looking better will make people more likely to buy. Don’t assume you can skip steps. And don’t let other people assume either. Practice these words: “you’re probably right, but let’s test it just to be sure.”

4. Don’t Let Anyone Get Emotional

Design is a personal, emotional process. Designers do the best they can, defend their designs, and are proud of them. Many don’t like hearing that you want to change it on them. Page optimization tests are on elements that it’s difficult to predict the results; it isn’t an attack on their personal ability. Why is it important you communicate this? You want designers giving you 110% all the time. If they get emotional about it, they won’t do as good a job on the other variations. Worse yet, if you’re convinced you’ll get a certain result before things have even started, you could harm the results as well.

5. Don’t Test Too Much at Once

This ties in to “don’t assume.” You can’t test a left form against a right form, change the headline and images, and conclude that the left form is better because of it. When testing, you want to know what is affecting the results. If you’re testing several variables, there’s no way you can know for sure what is working better. Obviously you can A/B test completely different pages this way then optimize from there, but the problem comes when you do a test like that then apply the kind of form logic I mentioned to put together another option which you infer is superior.

6. Trust Statistical Significance

When statistics says it’s true, it’s true. I’m sorry, that’s just the way it is. Don’t end tests extremely early because what you thought was going to happen looks like it is happening. Wait for it to REALLY happen. Since you don’t want to second-guess your results, and you certainly don’t want to give anyone else reason to second-guess them, appropriate an appropriate amount of time for each test and come up with defensible results.

If page optimization isn’t currently part of your plan, I strongly suggest you consider its benefits. It just isn’t for large businesses anymore either. Increasing site conversion rates from 3% to 4.5% usually results in a short payback period even for companies getting less than 100 visitors a day. Before you spend a lot of money sending traffic to your website, be sure you’re converting that traffic as effectively as possible.

Author: Mike LaLonde is a consultant at Londes Creative Digital Marketing. He specializes in website optimization, search marketing and PPC, and SEO. Any questions or comments can be directed to mike [at] londes.com.