A/B-testing

What is A/B testing?

What it basically is, is to compare an analyze two different scenarios, which worked better? In E-Commerce it is (most of the time) to try small changes to increse the conversion rate, and with that your sales. Doing just that, the most common way is to focus on the steps leading to an order. From the product page to a completed purchase. Often this means that you change the position and color of the price and purchase button and so on. But basically you can try changing anything you want.

In short it is to try small changes to get higher sales.

Why A/B-testing?

A/B testing is done because not all businesses or markets work the same way, something that works for an industrial wholesale company may not work for a jewellery store. Because the visitors are behaving differently.

Would another color of your purchase buttons increse your conversion rate?
There's only one way to know for sure, try! This test has been done many times, with different results, as the same changes does not affect all websites the same way.

How is A/B testing done properly?

Can i simply try?
When you work with A/B testing you want to minimize variables that may affect the outcome. Let's for example say that you try a green purchase button for two weeks and then a red one. What may ruin your results?

• Perhaps your competitor had a huge campaign the first week?
• Perhaps you statistically already get better sales during the last few days of a month?
• Perhaps the items are seasonal and the red button weeks are closer to a holiday or similar?
• Perhaps the weather was awful week two?

When you A/B test, you are trying to minimize these variables, it may be a good idea to not test one color at a time, but to spread it evenly among the visitors, let half of them get red colored buttons and 50% get green? This way the results of the study will be more accurate. (With large number of visitors, you may not need to test with that high percentage and not over several weeks either, the important thing is to gather enough data to anlayze, perhaps you get enough data running the test for an hour for only 1 percent.)