April 20, 2016
Facebook A / B Testing
Facebook’s advertising platform can target almost any human characteristic or interest. How can you know what’s important enough to test with so many choices? I’m going to answer that question for you today and stop you from getting lost in the myriad of options. The post below is based on $3 million in Facebook ad spend managed through my company, AdEspresso. With this data, I narrowed down the top 10 factors that result in the biggest changes, and thus, are the most important for testing. The list is ranked in descending order of importance:
- Precise interests
- Age ranges
- Body copy
- Relationship status
- Landing page
- Interested in
Here are some basics of Facebook A/B testing:
- The bigger the audience, the more reliable your A/B test results will be. We suggest at least a 100,000 person audience size for any test you run
- To A/B test properly, change only one variable in your ad set at a time
- The distribution of impressions or budget must be even among all the experiments you’re running to really get an accurate test
With that said, let’s get into it.
Even though I’m Italian, I still make myself exclude my home country from most of my ad campaigns. In fact, when testing, I suggest you only target one country regardless of how alluring it is to see results from across the globe. Testing across countries complicates testing, because there are:
- Language differences
- Cultural differences
- Cost and conversion differences
Thankfully, Facebook’s “Lookalike” audiences are created per individual country.
#2: Precise interests
Precise interest targeting lets you target your audience based on almost everything in their profile:
- Any of their interests
- Apps they use
- Pages they like
For example, if they like a competitor’s page, then they are probably a perfect target for your ad. With AdEspresso, you can test these interests against each other to see which one gets you the best conversions or clicks. The standard approach to precise interests is to target many interests at once with “or” logic. For example, let’s say you do high-end car rentals. You could start by advertising to users who like Audi or Mercedes or BMW – an audience of 315,000,000 people. You could then progress to targeting those who like Audi and Mercedes _and _BMW. With “and” logic, you go from an audience of over 319 million to a much more targeted 1.6 million.
#3: Age ranges
Do you want to appeal to teenagers, young families, or retired people? You have the flexibility to choose any age range above 13 years old within Facebook’s platform.
You can set up your age ranges anyway you’d like. Personally, I test the age ranges as follows:
In our data, 13-17 year-olds have the cheapest cost-per-click – less than a third of the cost per click of those over 55 – but we all know they have way less disposable income, no credit card, and the only way they can buy the product is to ask their mom, dad, or older sibling. Go with your gut when choosing your first age range to test, then experiment from there.
While I’m personally for gender equality, I can’t say the same about my Facebook ad targeting. Split testing gender shows clear differences. While there are 58 different gender options in personal Facebook profiles, Facebook ads currently narrows it down to two: male or female. While certain ads by Seventeen or Cosmopolitan would be targeted only to females, sometimes gender differences are much more ambiguous. Our data shows that females, on average, cost 30% more-per-click than males. However, averages don’t always apply to your industry, and you may find less (but still relevant) discrepancy in B2B marketing.
Your ad image needs to be visually catchy enough to attract a person’s attention amongst all the other News Feed posts. It also needs to convince the someone to take the action you want them to. Testing these is a balancing act, with many marketers creating more “clickable” ads with cheesy copy or images that catch a user’s eyes with contrast: But this doesn’t always lead to sales, even if the clicks are cheaper. This tactic could actually negatively damage the perception of your brand. One way to create visual contrast, without sacrificing your brand’s integrity, is by applying color filters – just like you would do on Instagram. In our experience, we’ve been surprised between what we thought would work and what actually worked with our images for AdEspresso. For one of our ad sets, we only changed the image, and had a cost per sale that became 75% higher than our benchmark!
The headline, or title of your Facebook ad, is the most important element of your ad copy. It needs to grab the user’s attention and define what your product is about. As a basic testing start, try having a more traditional headline tested against a more flamboyant, “clicky,” or playful one. Tread carefully, though, you don’t want to tarnish your brand. For other headline ideas, you can try:
- News headlines that mirror an announcement
- How-to headlines work for blog posts, why not ads?
- **Question headlines **to draw the reader into the rest of the ad copy
For more title ideas, here are 11 formulas to write irresistible headlines.
#7: Body copy
While the headline attracts the user, the body copy gives them that final nudge to click. Some things to test in your body copy:
- Social proof: Testimonials (like in the image above), reviews, the number of users on our email list, or customers of your product
- Urgency: Try testing a discount offer, or offering a free download for a limited time
- Call to action: Experiment with different calls to action like power words or even include Facebook’s own button CTAs (Shop Now, Sign Up, etc.)
#8: Relationship status
For many B2C industries, relationship status can be one of the primary factors in your testing. For example, wedding photographers obviously want to target “Engaged” couples before they change their status to “Married.” The Bouqs (a flower delivery service) recently ran an interesting ad targeted to males who had the often under-utilized “Not Specified” as one of their targets. The main categories you can test are:
- In a relationship
- Not Specified
Whether this is significant for your business greatly depends on your industry, but is a valuable testing variable to keep in mind.
#9: Landing Page
One of the most neglected areas of Facebook ads is where the ad links leads. Keep in mind that the primary element of a good landing page is that it mirrors your Facebook ad. In other words, while you test elements of your landing page, make sure you don’t lose the atmosphere and the promise in the initial ad that led the user there in the first place. Here are the major factors you should test for your Facebook ad’s landing page:
- Call to action
- Wording on a page
- Page layout
- Images on page
- Discounts or offers
With so many options to test, just remember this: only test one element at a time! For more on A/B testing landing pages, we created an entire list of resources.
#10: Interested in
Nope, we aren’t talking about someone’s hobbies and career anymore. We are talking about one of the most primal human desires: sexual preference.
- Male and female
- Not Specified
For example, Insomniac Events has an ad that seems to be a perfect candidate for this. Their testing results could provide a lot of insight into who is most likely to come. Clicks don’t always equal conversions, and I’d be curious about who actually bought tickets after clicking through on this ad. Who do you think? I hope you learned a thing or two about what to test in your Facebook ads from this post. There are many more areas to test (behavior, ad placement, etc), but these are the top ten to start with. And remember, test only one element at a time, and don’t test until you have a big enough audience. If you want to save yourself both time and money on your Facebook ads, try a free trial of AdEspresso as well. Ciao! What kinds of tests have you ran with Facebook ads? What’s led to the biggest wins? Let us know in the comments.