Our theme for the start of 2009 is – squeeze every bit of performance out of every campaign. The only way to do that is to test, measure, optimize, and test again. Since one of the cool new features in Pinpointe 2.0 is split testing (so you can automatically optimize campaign results), I thought it would be a good time to talk about split testing.
Here's a screen from Pinpointe 2.0's Split test campaign feature:
For example, if you have a list segment of 100 contacts you can send one campaign to 33% of your list, another to the next 33% and another to the final 33%. You can then see which had higher open/click rates and send that to your entire list.
Split tests are a great way to test multiple versions of an email and see which one gets the best response. In the above example, I might leave the email content identical in all 3 campaigns but just change the subject line in each to see which gets me the best open rate.
I wanted to talk about another way to do split testing – with your website using the free website optimizer tool from Google.
What is Google Website Optimizer?
Google website optimizer is a free marketing tool that allows you to perform two types of split tests on your website:
1. A/B split testing – Choose 2 pages on your website to split test. 50% of your visitors will see page A. The other 50% will see page B.
2. Multivariate testing – Allows you to break your web page up into blocks and show different content (in different combinations) for each block.
If you're new to split testing then make sure you watch this presentation from Google. It explains everything you'll want to know before reading the rest of this newsletter.
For most people A/B split testing is easy to understand – you might keep your current home page (which we'll call "page A") and create a new version of it with different header text and colors (which we'll call "page B"). A program such as Google website optimizer will show page A or page B randomly to visitors on your web site and when they buy from you a conversion will be recorded and it will contain a note about which home page the customer saw (page A or page B).
With A/B split testing the same visitor will always see the same page. For example if I came to your website today and saw page A, when I come back tomorrow I will still see page A. This is accomplished using a cookie to remember which page I saw the first time I came to your website.
Multivariate testing seems to confuse a lot of people but it's really just a way to test different combinations of content on a page and see which combination brings the most conversions. If you image a website as being nothing more than a group of blocks, it will start to make sense:
Assume each red block above is a piece of content (such as a paragraph of text, an image, a video, or even all 3 combined), using multivariate testing I can define different versions of content for each block. For example in the small red block on the left one version might have an image. Another version might have a paragraph of text with a thumbnail image instead, etc.
If I create different versions of content for each block on the page, I would then use multivariate testing to randomly show different combinations of content for each block and, based on conversions, determine which combination of blocks drove more sales on my website.
After the test period was over (generally anywhere from 1-8 weeks) Google website optimizer would give me a clear winner (through randomly showing different combinations to my website visitors and tracking which lead to sales). I can then stop the website optimizer campaign and show the winning combination on my website all the time.
Why Split Testing?
In marketing, anything that can be tested should. The layout of your website really is marketing – showing certain things about your products/services in certain ways to people that visit your website. Using split testing (either A/B or multivariate) you can maximize the conversion rate of your website which leads to more sales, newsletter subscribers, free trial sign ups, etc. The only way to find out which version will perform best, is with split testing.
Using Google Website Optimizer
Google Website Optimizer is free and is available when you sign up for a Google AdWords account. After signing up you don't actually have to pay to run any campaigns – just click the "Website Optimizer" link in the "Campaign Management" tab at the top of the page.
You'll then be able to setup a new split testing campaign. You will need to add some code to your website to run and track the split tests but it's only basic JavaScript code, so if you know some HTML and are familiar with an FTP program (such as FileZilla – free) then you should be fine.
Recommendations for an A/B split test
If your website only gets a few hundred visitors a day then I'd recommend starting with a simple A/B split test which would involve creating 2 pages on your website – generally you'd use your existing home page (or landing page) and create a new page containing one or more visual changes.
Here are some things you could change for page B:
Change the color of the main headline on your page. Try changing it from black to red.
If you have a large promotional-style image across the top of your page, try changing the image from a person to a picture of your product/service, or vice-versa.
Try replacing a few paragraphs of text with a few bullet points instead. People prefer to skim content instead of reading everything.
If your web page has a fixed width, try making it fluid (the whole width of the web browser's window) and vice-versa.
Change the font you use on your web page from Verdana to Arial or Times New Roman to Helvetica.
Of course these are only a few things you might change, but you get the idea. To make it easy to determine if the changes make a difference, only try changing 2-3 things at a time. Then give the split test time to do its thing (2-3 weeks).
Recommendations for a multivariate split test
Similar to an A/B split test, don't try testing more than 2-3 things at once. Also try not to create more than 3 variations of any one section of your website, otherwise you'll end up with hundreds of combinations to test. Try and keep the total number of combinations under 20. So you might test 3 different things, each with 2 variations.
Remember that even the slightest change can make a big difference in conversion rate. Instead of completely changing content, try subtle changes such as font size, style, color and spacing. Try re-arranging paragraphs of text and try testing with/without a prominent image on the page.
Finally keep in mind that having a beautiful website doesn't necessarily translate into more sales. I've tried combinations that made me second guess whether or not I should run the split test, and they've produced some really good results.
Remember that marketing is all about experimenting for the best result so the more tests you do, the higher you should be able to get your conversion rate over time.
