Every marketer knows it’s vital to run several versions of an ad (Facebook ad included) – also called split testing or A/B testing – if you want to find out what works for your clients. What you might not know is where to start or even what to test.
We’re going to share how and what we decided to test for our clients in a series of experiments we call the Quadrant Method for the majority of our Facebook ads this quarter.
In education, an effective way to present new material is to engage students in their own learning process. A KWL Chart is a type of graphic organizer designed to help students take charge of their learning. Instead of passively receiving instruction from the teacher, students are able to discover:
- What do you KNOW already about this subject?
- What do you WANT to know?
- What did you ultimately LEARN about your inquiry journey?
In marketing, we can apply the same technique to crafting Facebook ads that succeed in the framework of our particular campaign’s objective.
We know data beats intuition. We may think using a bright, cheery orange will attract more people to click on our ads than using a blue hue, but does our data back that up? We may think people in highly targeted audiences are more likely to click on our new car ads, but does it yield better results than targeting everyone in a particular geographical area surrounding the dealership?
These are the kinds of questions our new method is designed to answer. Let’s start testing!
How the Quadrant Method Creates Data that Beats Intuition
Each quarter, we decide what we want to test, considering themes we hope to learn about resulting in client success across multiple verticals. In the past, we’ve tested campaign types, ad scheduling, and so forth.
We try to avoid testing too many factors at once. That’s why we focus on areas where we can develop effective best practices and strategies for our clients, identifying which themes we can change over time.
This past quarter, we started our strategy meeting by creating a list of everything we already knew based on previous campaign data. Our list items in the “Want to know” column naturally fell into four categories. This is where the Quadrant Method gets its name:
- landing pages
The goal of our testing is to produce clear results that will lead to improvement. Breaking down these tests into four categories allows us to test one variable at a time over multiple campaigns for statistical relevance per quadrant.
Because it can be easy to fall into the trap of testing too many variables at once, it’s important to note which quadrant we were testing when setting up campaigns with our constants (placement, budgets within ad sets, etc.).
Facebook allows many combinations of audience targeting. We work with a variety of verticals, so each audience we build considers the business goals combined with other data points like age, gender, interests, and more.
Our goal in this quadrant is to understand how valuable each of the audiences is in achieving the best results for the clients’ campaigns. We’ve narrowed our test audiences down to three categories.
These audiences contain consumer behavioral information available from Facebook’s third party partners and additional Oracle data.
We created several varieties of these highly targeted audiences, such as car shoppers looking to buy a new vehicle in the next 90 days, or homeowners with a certain income level interested in custom built homes.
Our research has led us to theorize these audiences will yield the highest ROI because customers are already prepared to participate in the behavior. But will it actually?
These audiences were the simplest to set up. As an example, we’re looking at what happens if we target everyone 18-65+ in geographical areas either directly around the client or in areas based on their sales data. Will casting a wide net be a statistically relevant boon or burn when we run the same ads using additional data partners?
Customer Databases and Lookalike Audiences
We obtained customer databases with names, phone numbers, and emails to see if past and current customers responded as well as the other audience types. We ensured our databases were robust enough to still leave a substantial audience size after Facebook matched database information with users.
How will current and past customers respond compared to audiences containing people who are unfamiliar with the business?
Trying out a variety of statements will teach you how to craft effective, relevant messages. Do short, pithy statements resonate better than entire paragraphs? For an auto ad, what about if we mention the vehicle’s price in the carousel descriptions rather than just the make and model? How about if we focus on what the company can do for the customer using “you” statements v. what the company offers using language focusing on “our” or “us?”
For example, here are messaging options for a custom home builder, a longer and shorter one, both incorporating value statements and addressing “you” as the customer:
- Whether you have an older home, a storm damaged one, an empty lot, or nothing at all, we can help you with every aspect of the home building process for an affordable price. Sign up to learn more today!
- Now is the perfect time to build your dream home, and we can get you there.
And for a clicks-to-site campaign with a used car carousel focusing on length:
- Quippy: Get great deals on used vehicles at ____!
- Regular: _____ offers a variety of used vehicles at competitive prices! Check out our handpicked specials just for you!
- Longer: At ___, we take pride in our ever-changing new used car inventory. With tons of great models at competitive prices, we have the perfect vehicle for every budget and lifestyle. Explore our hand-picked used car selection today!
We want to know how our choice of where we send customers who click on our ad affects a variety of factors, such as site goals, conversions, bounce rate, campaign KPIs, and more.
Landing pages we’re testing are:
- Search results v. actual product pages
- Existing pages v. pages created specifically around the ad
- Landing pages with multiple options versus pages with one singular service focus
Regardless of which landing page option we send people to, we always want to be aware of page speed and user experience to the sites we use in our ads. Is there an opportunity for conversion? Is the page user-friendly? What about mobile-friendly?
One tip we can already give you is to make sure your UTM codes are as descriptive as possible. Doing so will help you sort page traffic efficiently in Google Analytics.
We’ve teamed up with our Design department to help craft a variety of creative graphic options. Best practices said friendly, smiling female faces and bright colors performed well, but is this always the case?
In addition to using stock photos as a constant (which, based on our data, we already know almost never outperform custom graphics and original photos), we are testing:
- Compilations with the car model and humans interacting in a natural manner
- A variety of colors
- Cartoon or vector images
- Carousel images v. single images
To Be Continued
Currently, we’re still compiling what we’ve learned just two months into our Quadrant Method experiments. We’re taking more time to gather data before we share the rest of our results (and we can already say there are some you might not expect), so keep an eye on this space for a follow-up article.
In the meantime, we hope sharing our methodology will get you started on your own experiments.
Every piece of data we obtain is valuable to us and to the industry at large. By crunching the numbers, we gain new insights and are able to make continuous improvements. And when we share those insights, we advance the conversation surrounding these best practices.
Reach out to us if you would like to hear more about our results, share tests that work for your clients, or get a free digital analysis to see how we can help your business achieve unprecedented ROI. We’d love to chat strategy and help you form a plan for success.