How a Test and Learn Strategy Can Help Drive Business Results

Facebook Business

August 08, 2019

Ads testing is one of the most effective ways to use measurement to make better, data-driven decisions and improve the performance of your ad campaigns. By choosing the right kind of test for your business question, using testing to vet new strategies and addressing common issues early on, you can get more valuable results.

Choose the right test

In order to use testing most effectively, think of lift studies as your tool to help make longer-term strategic decisions. For example, you may decide whether to invest in long-term campaigns for a particular audience segment, and use A/B testing to find opportunities for faster optimization (like headline or image choice).

If you’re wondering whether to allocate more budget to a campaign or how many conversions your Facebook ads drive, lift studies allow you to be more confident in your decisions, since they’re based on proven experimental methods. Whirlpool used a lift study to test different creatives for a digital-first millennial audience. Whirlpool's lift study helped the team to determine which creative performed best before investing in new creative resources, and build the right strategy to reach this audience.

Use testing to vet new strategies and tools

Testing can help you evaluate new strategies and marketing tools without fully committing your team’s time and resources. For example, say you want to understand how machine-learning driven tools impact your ad performance. You can use Facebook’s Test and Learn tool to measure and prove the incremental value of Campaign Budget Optimization (CBO). CBO uses machine learning to deliver your ads to the people most likely to take action. To test CBO with A/B tests, copy your campaign, divide your budget between them and run your test. When the test is finished, you can combine the audiences and allow the campaign manager to pause the underperforming group, letting the more successful campaign continue running.

Here’s how it looks in practice: Retailer Glue Store used split testing to compare the performance of automatic placements for dynamic Facebook ads—which use Facebook’s machine learning algorithms to determine the most efficient placements—against dynamic ads that ran solely in News Feed. With the test, the team was able to confirm that using automatic placements to determine where dynamic ads would appear resulted in better performance—an 88% increase in ad spend—than dynamic ads that ran solely in News Feed.

Learn more about how machine learning can drive better advertising results in our recent blog post.

Address potential issues early on

It’s important to identify and address factors that could impact the viability of your test results in the planning stage to ensure you get actionable results. Low-confidence list studies and variables that are not properly isolated are two common issues in setting up tests.

• Confidence measures the likelihood that a test’s results can be recreated, and weren’t a one-time result. Low-confidence lift studies can happen if your test didn’t gather enough data either because a long purchase cycle led to few conversions during the study or because the audience was too narrow. You can also get low-confidence results in your lift studies if your test and control treatments are too similar, or if there wasn’t a measurable difference in people’s responses.
• In order to ensure variables are properly isolated for your test, consider whether people in your test audience are being exposed to multiple ad creative concepts or a marketing message outside of the digital channel you’re testing. While it may not be necessary (or possible!) to stop other marketing efforts, you should make an effort to ensure every group in your test is comparably exposed. You can learn more about how to resolve potential testing issues faster .

Check out our measurement foundations course to learn more about Facebook’s approach to measurement.

Tags

Best Practices Small Business