What Is A/A Testing in Marketing?

By Indeed Editorial Team

Published April 5, 2021

Choosing to change your company's website or app design and layout can have a significant impact on user experience and ultimately your company goals. Conducting random testing may help determine which extra features to incorporate into an upgrade. Choosing an accurate testing program and method can help ensure credible data backs your changes. In this article, we discuss what A/A testing is, why it's important and how to run one for your webpages and apps.

What is A/A testing?

A/A testing is a statistical testing method used in web and app design. It uses an algorithm tool to test two identical versions of an app or webpage against one another. An A/A test can be a precursor to an A/B test, or split test. An A/B test compares two different versions of a webpage or app to determine which one performs better with users.

Related: 110 Types of Software Testing

Why is A/A testing important?

One of the primary purposes of A/A testing is to help choose an accurate and well-calibrated tool for A/B testing. When running an A/A test, a reliable program reports a statistically insignificant similar conversion rate between both pages. It is also beneficial to show you the potential margin of conversion error for a particular testing tool.

When should you run an A/A test?

There are certain stages in the web design and development process where running an A/A test may be most beneficial. These situations include:

  • When you've installed a new testing tool in your system

  • When you've updated or changed the setup of your current testing tool

  • When you're starting a new web or app project

  • When you find differences in the data reports of your testing tool and other reported analytics

How to run A/A testing

Use these steps to learn how to conduct an A/A test:

1. Choose your tool

Choose a testing tool to start your A/A testing process. These types of programs are available from multiple different analytics companies. You may choose a tool you've used in the past or a new program that you want to try out and potentially switch to for A/B testing. Review training documents for any new tools to ensure you're calibrating it and inputting the test parameters properly. This helps you determine if the program gives an accurate result.

2. Choose your type of testing

Depending on the tool you choose, decide which method to implement for your A/A test. Options include:


Hypothesis testing requires a pre-determined sample size. The program runs until there are enough samples for each variation. Once it reaches the desired number of samples, you can determine if there's a lack of difference in your key performance indicators and stop the test.

Read more: Defining Hypothesis Testing (With Examples)


Bayesian tests don't require a pre-determined sample size, which you may prefer for an A/B test. Instead, this type shows, out of the two options, which variation is best based on even tiny fractions of difference in the main metric. The more data collected for a Bayesian test, the more sensitive it is to differences in the key performance indicators. This means it may be more likely to choose a "better" version in an A/A test, even though the samples are exactly the same.

3. Set up the user experience

Users won't know that you're collecting data about their browsing habits on your site while performing an A/A test. One reason for that is because the user experience for those who visit the control page and the variable page can be identical. Ensure that there are no changes between the two and then set the key performance indicators for both groups to test the conversion rate. Key performance indicators may include actions like clicking a button, enlarging an image, navigating to a specific other page or making a purchase.

4. Interpret the results

Look at the data you've collected to see if it makes sense within the context of the program and the larger project. Keep in mind that there is always an element of randomness in A/A testing. Your identical pages may have slightly different conversion rates depending on the program and testing method you choose. A conversion rate is the percentage of users who take the desired actions stated in your key performance indicators.

Any conversion rate smaller than 0.05 or 5% is statistically insignificant, and you can view it as a random generation within the test. Bayesian tests may be more likely to choose a statistically insignificant but higher result as a better performing version, but you can discount these during a human data review. You can determine that identical and statistically insignificant A/A test results show that a testing program works properly.

Related: Understanding Different Software Testing

5. Set the baseline conversion rate

After completing an A/A test, you may determine what the conversion rate margin of error is for your specific testing tool. Knowing this number can help you set a baseline conversion rate for your A/B test to determine the most significant areas of change from your control page to your variation.

6. Determine your sample size

Conducting an A/A test may help you determine how many user interactions to expect in your A/B test. For a hypothesis test, this may help you choose a reasonable number for your pre-determined sample size. For Bayesian tests, it may help you understand at which point to end your testing and analyze your results.

A/A testing best practices

Use these tips to help you conduct a thorough and accurate A/A test:

Use a large sample size

No matter which testing method you choose, pick a minimum sample size that is large enough to yield actual results. The number may be statistically proportionate to your conversion rate goal. Use your current analytics data and statistics to choose a number that fits your daily reach or access. Remember that an A/A test follows the scientific method and can repeat multiple times before it gives tangible results.

Related: How To Become a Software Testing Engineer

Watch your timing

A/A tests may take longer than A/B tests to reach a conclusion about performance accuracy. With reliable programs, this is beneficial because it's having a hard time finding discrepancies between the two samples because they don't exist. Letting your test run longer helps increase accuracy and collect a wide variety of data. It also allows for more consistent and accurate results.

Explore more articles