How to A/B Test Your Facebook Ads

Fighting the Algorithm: How to A/B Test Facebook Ads

Last month I got a great question from Jen, one of the advertisers in the Absolute FB Ads Support Group. She was putting together a campaign for the launch of a program her client runs a few times a year. Based on previous results, she knew which image performed best, but she was frustrated because none of the other images had really been given a chance to shine because of the whole “survival of the fittest Facebook ad” thing.

Jen is smart, and knows that other images might do better, but she didn’t know the proper way to A/B test given this whole “survival of the fittest ad” thing.

Here’s my answer on how to A/B test Facebook ads the smart way:

1. Try out two ads at a time.

Every ad set should start with just two ads in it (but you can run as many ad sets as you want to different audiences). Take your winning image from the last campaign and test it against a new one. Make sure everything else about the ad is exactly the same.

2. Let the winning ad win.

Facebook does you a huge favor by optimizing your campaign, so let it! If you have one ad that gets all the results, is that really a big deal? Good results are good results. Don’t let a desire to test a bunch of different things get in the way of that. (Imagine how your client will feel if your other images perform a lot worse! Nobody wants that.)

3. It will be hard to test something new against a winning ad.

If one of your ads has brought in a bunch of clicks and conversions, then Facebook will always declare it the “fittest,” even if you add new ads into the ad set. Chances are good your new ads won’t get more than a handful of impressions! A winning ad will always beat out an ad with no clicks, even if we want the new ad to have its chance.

4. Don’t touch an ad that’s getting killer results.

If you really want to test out winning creative against new creative, you have to recreate all the ads in a new ad set. Recreating an ad essentially resets it in the “survival of the fittest” game.

But Jen saw that when she did that, the ad that was doing so well in one ad set got some pretty expensive cost per conversions this time around. Why? Because Facebook takes time to optimize your ads; part of the reason that the ad was winning before was because their system helped it find it’s groove.

Starting over means no groove. So if you have one ad that’s doing really well, let it run its course. I often end up with one ad in an ad set that uses up 90% of my budget because I don’t want to fuck with its groove.

5. Wait for the winners to stop winning.

Eventually all ads deteriorate over time, even the winners. Let your winning ad run all by itself in its ad set until the cost per conversion starts to rise significantly.

I don’t watch an ad’s  overall cost per conversion, but rather how much conversions cost me on individual days. If the cost per conversion rises for two consecutive days, I’ll pause it and try something different.

6. Avoid audience fatigue.

Last week I saw 3 different ads from the same person advertising her webinar. I wondered if she was running multiple ads at the same time, more than one ad set at the same time, or changing up her ads a little too quickly.

Using more than one ad set is a good way to test out different creative. Just be careful how you do it. If you only use different interests to differentiate your ad sets, you will absolutely end up with crossover. Here’s what I mean:

Say you run 3 ad sets, each with one interest. Your interests are Gary Vaynerchuk, Marie Forleo, and Smart Passive Income with Pat Flynn. I like all three of these pages on Facebook, so it’s very possible that I’ll be targeted with all three of your ad sets.

Instead of segmenting your audience by interest, use hard demographics to make sure you reach different people. In a recent campaign I ran, my client wanted to see if moms or single women would convert the best. So I ran two ad sets for each call-to-action: one that targeted moms and one that targeted single women. (Everyone converted at the same rate on her landing page, but moms ended up being more expensive to target. Just in case you were wondering!)

Here are some other demographics you could use to segment your audience:

  • Location: test different creative for different cities, states or provinces
  • Age: you could segment your audience using the same age groups Facebook uses in their reports: 18-24, 25-34, 35-44, 45-54, 55-64, 65+

Never Stop Testing, Tweaking, and Trying

I tell everyone who I train one-on-one the same thing: just because something works this week does NOT mean you’ll get the same results the next time you try it. Your CPM (cost to serve your ads 1,000 times) depends heavily on who else is running ads to the same audience as you, and those numbers fluctuate like the US dollar on the Argentine black market: a shit ton.

So Jen knows that the image that worked for her last time might not be the “fittest” ad this time around. And when I restart a list building campaign that killed it last time around, I have to watch all the metrics to see what’ll happen this time.

I say this not to scare you, but so that you’re prepared for anything:

Never assume that you’ve FINALLY figured out the key to Facebook ads; you never know what’ll happen a month from now.

Need some hand-holding or clarity on all these crazy details? Click here to check out my one-on-one Facebook ad training service. Fill out the application and we’ll jump on the phone together to see if you and your business are a good fit for that investment!