Creating kickass ads is equal parts art and science, and today we’re diving heavy into the science behind testing Facebook ads. There are three primary methods for testing Facebook ads. Each comes with different limitations and strengths, but all require a few key steps in the process.

Step One – Set aside part of your budget to test your ads. (If you’re not sure where to start, 5-10% of your total ad spend is generally an appropriate range.) Realize that this money is doing valuable work to lower your cost per result for the life of your campaign. If you test your ads effectively, this investment more than pays for itself.

Step Two – Form a hypothesis. From what you know about your target audience, what do you infer will be the most effective ad experience to garner engagement from them? Remember, a good hypothesis has a dependent variable (what you think will be affected), an independent variable (what you predict will affect the dependent variable), and a prediction of the effect. 

For example: 

If I run a video as the visual for this ad, it will have a higher click-through rate than if I ran a still image. 

     – The video is the independent variable
     – The click-through rate is the dependent variable

Step Three – Use one of the following methods to test your hypothesis. A word of advice from one ad scientist to another, only test one variable at a time. If you change two variables (say the headline AND the ad placement) you won’t know if any changes that result occurred because of the headline or because of the ad placement. You may have achieved a higher CTR with your ad, but you haven’t learned anything definitive and can’t reproduce your results throughout the rest of your campaign. Make sense? 

Common factors to test include:

-Still image vs video
-Video lengths or image variations
-Different headlines
-Different primary texts
-Copy lengths
-CTA buttons

Once you’ve highlighted what variable you’ll be testing and formulated your hypothesis, move forward with one of the following methods:

Ad Testing Method # 1
Several Ad Variations In One Ad Set

You can use this method to manually test multiple variables at the same time by creating ad variations of each possible combination. As you can imagine, this would get really complicated if there were too many variables being tested at once!

With this method, result response is largely in your hands and it’s important for your CPR that you turn off ad variations with a lower ROI. Remember, Facebook prioritizes ads with the highest click-through rate. If you realize that your ad has a high click-through rate, but a lower conversion or download rate once users click through, turn it off. If pure traffic isn’t your goal, a better metric to look at instead of CTR is the ratio of click-through to campaign objective. 

For example, if you were running ads with a campaign objective of getting users to download a new app – an ad with a high CTR but low download rate would actually eat up most of your ad spend and not help you achieve the objective of your campaign. An ad with a slightly lower CTR but a high download rate would be a stronger ad.

Ad Testing Method #2
Facebook Split Testing

For this method, Facebook runs your test for you by dividing your audience into random groups that don’t overlap. It duplicates the multiple ad sets you’ve created for the test and tests the ad sets against each other. Each set’s performance measure on your campaign objective (click-throughs, lead form fill outs, app installs, etc).

The variables you can split test include:
– Creative Variants
– Delivery optimization
– Audience
– Placements

Keep in mind, if you’re split-testing two bad audiences, two bad placements, or two weak headlines, you won’t learn much. It’s most effective to split test against an ad you already know performs well with your audience.

There are a few drawbacks to this method. A/B testing with Facebook uses some of your ad spend and makes your ad set more expensive than if you were running just one set with several ads. Depending on your audience size it can take up to 14 days to get statistically significant results so don’t go with this method if your budget can’t handle paying for that higher rate during the testing period.

For this method, we suggest only testing two variations (instead of the possible 5) and testing significant change – different concepts as opposed to a single word shift. It will make it far simpler to analyze results and reach a large enough audience to have results that matter.

The bottom line – Split tests are ONLY helpful when you want to test an isolated variable for a limited period of time. They’re a powerful tool, but shouldn’t be the primary way you optimize your ads.

Ad Testing Method #3
Dynamic Creative Optimization (DCO)

Similar to Google’s responsive ads, Dynamic Creative is a tool that allows the Facebook algorithm to choose which combination of your creative assets would come together to create an ideal ad to put in front of each individual user within your target audience.

To create a DCO ad set, you can upload up to 30 assets:
– 5 headlines
– 10 images or videos
– 5 bodies of the primary text
– 5 descriptions
– 5 CTAs

Make sure that all of your assets make sense together. No two headlines will run at the same time, but each headline needs to make sense with each video, and so on. If you have primary text that only makes sense with one CTA, this isn’t an appropriate asset to load into a DCO set.

When you use this testing method,  use broad targeting so that Facebook has a large enough audience to work with. Watch the individual ad performance and realize the algorithm’s limitations. The ads will optimize to have the highest CTR and as we noted earlier they could end up having great click-through but a terrible ultimate conversion rate.

DCO is powerful, but it’s important to remember that it’s only a test. If you rely on it for all of your ads, you’ll end up wasting money.

__

As you dive into the world of optimizing your ads it can be easy to feel like you’re in over your head. Get in touch with us if you have any questions about testing Facebook ads. For the sake of science, we’re always happy to help.