Articles in The 'split-testing' Tag

If you are not testing, then you are wrong!

by MoreVisibility

Back in High School, I was a Lieutenant in the Junior Reserve Officers Training Corps (Go Eagle Battalion!). Class leaders had a lot of fun during uniform inspection time, as well as during drill and ceremony time. We would routinely shout out several commands and instructions to our particular platoons during each class hour, which also included several popular sayings within our corps. These included (but definitely were not limited to):

“NINE to the front, and SIX to the rear!” – A reference to the length and distance of your arm swing during marching;
“Get in STEP [Cadet’s Rank and Last Name]!!” – During marching, this would be sounded off to ensure that each cadet’s step would precisely match every other cadet’s step. If every cadet stepped with their right foot, the cadet that stepped with their left foot would stick out like a sore thumb;
“Move the RIFLE around your head, not your HEAD around the rifle!” – My personal favorite during drill and ceremony with our Springfield M-1903’s;
“If you are not double-timing it, you are WRONG!” – You can replace “double-timing it” with any number of different instructions or tidbits of information to convey the message that this was a team effort, and you were wrong if you weren’t participating like everyone else.

We can take that last saying and apply it to testing and experimentation on the web. If you have a website and a marketing plan of any kind, it is imperative that you implement a testing and experimentation plan. Why? Because if you are not testing, you are wrong. In today’s internet world, you absolutely need to have some kind of testing strategy where the ultimate goal is to improve your website’s functionality, your lead acquisition process, and your shopping cart, so that you can have even happier customers, create some more returning shoppers, and ultimately make more money.

For starters, it doesn’t matter what you test – just get your feet wet!

If you’ve been thinking about testing, or if this blog post is the first you’ve ever heard of it, know that for right now, it doesn’t really matter what you test. The mere fact that you going to start testing something – anything – is good enough for now. Get your feet wet and get comfortable and familiar with the idea first, before worrying about what types of testing strategies exist or what standard deviation stands for. Pick anything on your homepage to test for a week or two – that picture of a palm tree, that blue “click here” button, or that first paragraph of text. Pick one of those items (only one for now), and make a change to it, upload it live, and see what effect that has on your traffic and your conversion rate over a week or two. Congratulations – you have just tested something!

This testing idea sounds great, but I wish there was a free tool out there that can help me set-up tests or experiments on my website…

Have no fear – Google Website Optimizer is here! Google Website Optimizer (or GWO for short) just recently came out of Beta, and is now available to everyone on the planet for free. GWO affords you the opportunity to create an unlimited amount of experiments, completely controllable and customizable. GWO goes as far as to offer your technical or website programming team a unique set-up page per each experiment, so that they have every piece of code and every instruction necessary to set GWO up for any page on your website.

What types of Experiments can I conduct with Google Website Optimizer?

There are two different types of experiments:

A/B Experiments – Sometimes also referred to as “A/B Split Testing”, this tests one page on your website up against a different version of that same page, to see which page gives you the best possible chance for an increased conversion rate. Rotating your Ads on Google AdWords evenly is a form of A/B testing in the marketing world. This is the same concept, but for a page on your website.

Multivariate Experiments – Sometimes also referred to as “MVT Testing”, this tests different areas of a page on your website (for example, different headers, footers, or product images), to see which combination gives you the best possible chance for an increased conversion rate. This is actually quite an advanced type of test, but Google Website Optimizer makes it easy for all of us.

How long should I run a test for, and what results will Google Website Optimizer show me?

I like the 15-day rule. With 15 days, you get two full weeks, plus that additional day’s worth of information. This could be longer or shorter, depending on the volume of traffic to your website. However, something in the neighborhood of two weeks should be enough time for a proper experiment.

Google Website Optimizer gives you a “Page Sections” report and a “Combinations” report (specifically for your Multivariate Tests) for you to look at. You’ll be able to view the estimated conversion rate range, in both a numerical form and a sliding bar graph, as well as other fancy statistically-oriented metrics, such as “Observed Improvement”, and “Chance to beat Original / Chance to beat All”, allowing you to very quickly see which page version or which page combination is doing the best job of bringing you more conversions.

What if I run a test between my homepage and a new version of my homepage, but the original homepage beats the new homepage – is it back to the drawing board?

Yes, and no. First of all, you’re going to have to become comfortable with the idea that an original page / original combination beating a newer page or newer combination doesn’t equate to an unsuccessful experiment. If you’re able to conduct a fair and unbiased experiment, then the experiment itself is successful, regardless of the outcome of the experiment. Google Website Optimizer runs fair and unbiased experiments, so rest assured that your experiment will be a successful one.

Now, just because your original homepage beat your new homepage, doesn’t mean that you can’t learn something that you can use in your next experiment. Keep track of what changes were made on the new homepage, and what was different on the new homepage versus the original homepage. If you only make one or two changes, you’ll have a much easier time in keeping track of exactly what’s making the visitors tick and what’s making them leave your site than you would if you completely re-invented the wheel and made several dozen changes.

Other than the homepage, what other types of pages can I experiment with?

The question should really be “what can’t I experiment with?”. You can and you should experiment with all different types of pages – homepages, about us pages, thank you pages, shopping cart pages, order confirmation pages, and so on. GWO lets you run an unlimited amount of different experiments, and you can also run multiple experiments simultaneously with different parts of your website.

Stay Committed!

After you’ve started testing, don’t let the novelty of it wear off. Find a way to make testing and experimentation a part of your job. I know, I know – you’re very busy and you have a lot of work to do, and you can’t possibly imagine putting on yet another hat on. But you JUST have to! Otherwise, your competitors will begin to fly right past you and take your customers away from you. You wouldn’t want that, would you?

Try this: every month, pick 1 thing to test. The “Add to Cart” button, the homepage text, the links on your “Thank You” page…anything. In a few months, you will thank yourself, as you will (hopefully) work towards making your website more attractive to your visitors, which should in turn increase every marketers metric, the conversion rate. Even if your conversion rate doesn’t increase, you will at least have started to learn about your visitors – what they like, what they don’t like, and what they react positively or negatively to – which can only help your business.