Sometimes you hear people complaining that Facebook ads don’t work. While it is true that Facebook ads don’t work for every product out there, the people who say this usually aren’t doing Facebook ads testing and scaling right.
There are many factors that come into play to create a successful ad campaign, but the most important one is testing. By only setting up one ad campaign… with one image… and one headline… and one copy variation – it is very unlikely that you’ll succeed.
In baseball, if you went up to bat and were only allowed one strike, you’d be more likely to strike out. But when you have 3 strikes, you give yourself more opportunities to hit a home run.
To be successful on Facebook, you must test. A LOT.
For testing new campaigns quickly, I use AdEspresso to quickly generate campaigns that can test multiple variants at once. The variants I usually test that I find make the most impact are:
Setting up these campaigns are simple enough. But where many people struggle is figuring out how to declare winners and losers, or when to increase or decrease the budget.
The first thing you must do to determine a successful campaign is to define what success looks like.
Most of the campaigns I run are to generate new leads, so my #1 decision factor for these type of campaigns is the cost per lead.
Based on previous campaign data, I know I can generate leads for a certain product at $5 each. If I’m setting up a new campaign for that product, my goal will be to try to find leads for under $5.
If you’re starting from scratch, you might not have any ad data to help you set goals. Maybe you have other data to use that will help you calculate how much you can pay for a lead depending on the profit of the product and the conversion rates.
Once you start a test campaign, don’t tinker! Once I verify the campaigns are running and nothing looks out of the ordinary, I don’t even look at the campaign at all for the first 5 days.
It takes a while for algorithms to figure out what’s going on and for the data to settle. If you look at results in AdEspresso, you’ll probably notice the first few days of data jump all over the place until they even out.
It’s also important to make sure each variant you’ve tested has been served enough times to start making decisions. This is going to depend on what your daily budget is and how many different variants you’re testing.
If you’re doing Facebook ads testing a lot of different audiences, it will take longer for each of those to spend enough to produce useful data.
To make sure you get this data faster, make sure to distribute the ad budget evenly across all ad sets, rather than proportionally. That way, all audiences will spend the same amount so you can make a better decision, faster.
When you use proportionally based on audience size, often all the budget will go toward the largest audiences leaving the smaller audiences with mere crumbs of a budget. One audience may be spending $30 a day while another is only spending $2 a day.
In this case, each audience has spent almost $50 which is plenty for accurate data. The minimum I usually go off of is at least $15 spent per ad set. For the campaigns, I run and the cost per leads I get, that is usually enough to make accurate decisions. Depending on what you’re advertising, you might find you need to spend more before you get enough data.
Using the Metrics
Although my main goal is generating leads, the first thing I look at is CPC. This is the best indicator to tell me if the people who see my ad are interested. For me, clicks under a dollar are good, and clicks under $.50 are great.
Right off the bat, I’ll shut off any of the variants that are really far off on the high side of the range. If I have CPCs in the $.30’s, $.40’s, $.50’s, $.60’s, $.80’s, $.90’s, and $1’s, I’d cut out the $.80’s and above.
As more time passes, I’d keep an eye on the ones in the $.50’s and $.60’s and either lower the budgets or cut them if they get more expensive. I would allocate more budget to the cheaper ones.
However, if you have 5 different variants, and they are all super cheap, leave them all running! If they are all close like $.15, $.19, $.25, $.26, $.31 – you don’t have to shut off the $.31 just because it’s performing the worst. If you already know anything under $.50 is great for you, these are all great!
CPC is good to use for a first round of cuts, but remember in this scenario we care about leads. Now we’ll check the actual leads generated from each variant. If we find that one of the variants with a high CPC was producing the cheapest leads, we’ll want to turn that back on. This would be an outlier, as most often you’ll find the cheaper clicks correlate to the cheaper leads.
In the AdEspresso dashboard, usually CPA (based on the Conversions column) is the number we’d look at for the cost per lead. These numbers are good estimates, however, they are based off the Facebook Pixel which isn’t always 100% accurate.
Whatever system you use to track your leads and where they come from, these are the numbers you should be using. We use UTMs to track which ad each lead came from.
Using this information, we can decide to shut off any other variants that hasn’t been producing leads, even if the clicks are cheap. If you find yourself in the situation of really cheap clicks but low conversions, you may want to wait a few more days to see if any come in. You may also consider other benefits such as more website visitors = a larger audience you can retarget to later.
When in doubt about whether or not you should turn a variant off, leave it running for a few more days to see if it gives you any clearer data. You are looking for a big enough difference between the variants to choose clear winners or losers.
For example, these images are too similar to stop any of them. Plus, they all have very low CPC’s with $.28, $.37 and $.41.
The blue image is far enough apart from the yellow and red to declare it a loser. The farther apart they are, the more certain you can be to stop the loser. The yellow and red are close enough together to let them both continue to run, especially since they’re both at or below our threshold of a $1 CPC.
Scale Your Facebook Ads
Once you’ve cut the losers, you’re ready to scale the winners. Just like you would cut out the variants that are way behind the pack, you want to scale up the variants that are way out in front.
Assuming that the audience size is large enough, you should increase the budget for any variant that is performing very well. It is best to increase this budget slowly. When the Facebook algorithm is used to finding you people at a certain budget, it gets thrown out of whack when that budget is drastically changed.
Try increasing budget by 10-25% every few days so it’s a more gradual increase.
Another way to scale is to take the winning variants and create new campaigns to find new audiences. You can also test different images or copy to try and beat the existing winners.
Facebook Ads Automation
Once you come up with a standard process that works for your campaigns, you can automate it. Facebook allows you to set up rules to help manage your campaigns.
If your CPC goal is $1, you can tell Facebook to automatically stop any ads or ad sets that go over $1. Or instead of automatically stopping it, you can opt to get a notification so you can review and manually decide what you want to do with it.
Once you create a process for creation, testing, and optimizing it’s easier to teach others how to replicate the system and grow your advertising efforts.
Share Your Facebook Ads Testing Stories
Do you already have a Facebook ads testing system in place that works for you? Share with us in the comments below! Get more digital marketing knowledge delivered right to your inbox by subscribing to our newsletter.