| A/B testing is simply comparing one page design (A) | | | | 5. Measure the results |
| against another page design (B) on your website. | | | | Using your web analytics software, measure which |
| Below are the steps to conduct an A/B test. | | | | variation had the highest conversion rate. This is the |
| 1. Identify metrics, what is success to you? | | | | winning element. |
| Typically this is the conversion rate of the page but it | | | | 6. Repeat with different elements |
| can be any relevant metric such as the average visit | | | | Now, you want to repeat the process with different |
| duration, number of items in a shopping cart, or | | | | page elements. Eventually you'll build a highly optimized |
| average order value. | | | | page. |
| 2. Choose an element on the page to change | | | | I bet you're thinking, isn't this is a lot of work? Well, it is. |
| This could include a 'Contact Us' form, submit button, | | | | That's why there's commercial testing solutions that |
| text, the headline, images, or colours. However, only | | | | automate much of this process. My favorite is Google |
| choose one element to test in each series. | | | | Website Optimizer because it offers a robust solution, |
| 3. Change one element in each new page | | | | and it's free to Google Analytics users. Other paid |
| Create a number of pages with only one change on | | | | solutions include Test & Target from Omniture |
| each page. Change only one element for each test | | | | and Personalization from Sitebrand. |
| series. On a Contact Us page, try changing the number | | | | These solutions use both multivariate testing and A/B |
| of fields in the form, the layout of the form, the size of | | | | testing methods as well as handling the random page |
| the boxes, or pre-populating data, and even the style | | | | serving process. The multivariate testing method |
| and size of the submit button. | | | | allows you to test multiple page elements at once to |
| 4. Deploy the test pages | | | | figure out which combination is the most effective. |
| Launch the pages and direct traffic from the site | | | | Testing important pages on your website can have a |
| randomly to each test page. You'll need some help | | | | significant impact on profitability without spending any |
| from your IT department here to ensure that the user | | | | marketing dollars. Be sure to consider it in your product |
| experience isn't affected and you get a statistically | | | | planning & design strategy. |
| relevant sample on each test page. | | | | |