Go to Top

How to become an A/B testing master

3 simple rules that will revolutionize your performances! 

Ad testing is a crucial part of account optimization. Ads that perfectly fit the search query lead to better Quality Score, thus can reach higher Ad Positions for lower Cost Per Click.

And here the crucial question: how should the perfect ad look like?

It depends.

Unfortunately we don´t have a magic wand, but we can give you 3 basic tips that will help you to boost your performances!

#1: Goals come first

Before setting up your test, ask yourself: what do I want to know?

In this article we take into consideration ad messages, thus we want to compare two different texts and verify which one brings more results.

In order to have an objective evaluation, it is necessary to change only partially the message content, hence, one line at a time. In fact, if you test 2 completely different ads, you discover for sure which one works better, but not why.

For example, you test the following two ads:



Let´s say that ad B has a higher CTR and a higher CR: why did it happen?

Are users attracted by “cheap shoes”, or by “global shipping”?  Or is “shoes online” the key reason?

As you can see, these kinds of AB testing are completely wrong! What you can do, instead, is to change one line at a time and verify which sentences perform better.

Here is an example of how your AB testing should look like:

Ad A:


Ad B:


We have only changed description line 1, while we kept all the other parts unchanged.

By doing that, you can easily verify if a neutral message, such as “shoes in all variations” performs better than a message that aggressively promotes a current sale.

#2: Interpret the results                                                            

After a couple of days, you get these results:


Which ad is the winning one?

At a first look, Ad A as a higher CTR, thus we should keep it. However, by observing the CR, Ad B leads to better performances.

What to do?

If you analyze the results deeper, you will notice that the good-looking conversion rate is based only on 8 and 10 conversions respectively. From a statistical point of view the difference in conversion rate is not significant.

If you look at CTR, instead, the quantity of data collected is enough to conclude that the text displayed in ad A leads to higher performances.

#3: Never trust what you see

And here is our last tip: always use one of the free statistical significance testing tools available online to check if your test acquired enough data to take a final decision.

Are you scared by statistic and online tool? Don´t worry: you simply need to enter the data your test has accumulated and the tool will automatically report if the difference between ads is significant.

Have an awSEM A/B testing weekend!