AdEspresso

How To Slash Your Facebook Ads Costs with Split Testing

Have you been running campaign after campaign without seeing the results you wanted?

You know your products are great because you already sold a bunch of them. Is the cost per lead simply too high?

That amazingly cheap, endless stream of leads you were promised is nowhere to be seen, and the few ones you got costed you an arm and a leg.

By now you must feel like you’ve just wasted your money and hate that Zuckerberg guy…

What you are going through sucks. It feels horrible but it’s completely normal, we’ve all been there.

In this article, I’ll show you how to turn that around, what to do next and how you can go from feeling like your money went down the drain to getting the results you are actually looking for.

If you are interested in learning more about your audience, build a funnel that works all the way from your new customer’s computer (or mobile) to your cash register- and you’re willing to invest some time and money into it, this is for you! 

So go grab a coffee and sit down comfortably, I’ll show you how to get the results you want by focusing on the biggest wins first.

Rule of the rules: you have to be patient

There are so many different things to try on Facebook, that it’s unlikely you’ll find the best stuff early on.

For the sake of this argument, let’s imagine there are 10 different combos and one will give you high quality leads at the price you are dreaming of. What are the chances that you hit that one combo if you only launched one campaign? 10%. So how do we make sure that we will find that one combo out of all those 10? We try them all.

What may be hard to believe is that even if you have to test all 10 different combinations to find just that one home run, it will pay for all those other failed attempts and then some.

We are going to be methodical and focus on getting the biggest wins early on, so we don’t waste our money. If there ever were a magic bullet, this would be it.

Split-testing

Split testing (also called A/B testing) is testing two or more things against each other to find the best performing one. It’s a continuous process and what works best now may change in the future, but you will learn a lot about your audience in the process.

Why would you do this? Because it is the best way to figure out what your audience responds to and then showing that to more people like them.

After a few quick notes and best practices, we’ll dive into testing. First, we’ll talk about the order of testing, what to test first and what to check later (so you’ll get the biggest wins upfront 💰); then I’ll explain each step in detail followed by a couple of notes.

I will not be covering how to set up a campaign step-by-step, how to maintain your campaign nor how to scale your campaign once you are happy with the results you are getting – The guide is pretty long already, so I’ll save that for next time.

Sometimes it’s worth testing the campaign objective earlier, but if you’re a beginner, I don’t want you to be overwhelmed. I recommend that you simply pick an objective at first – usually “clicks to website” or “website conversions” will be fine – and then we will test it later once you feel more comfortable with the whole thing. Sounds good?

Some of you may have heard about multivariable testing. I consider that to be an advanced topic, so I won’t be covering that for now. To explain it really briefly, it means to test several things at the same time to find a great combo faster. However, it does require more work upfront and a higher budget.

Best practices

  1.  When you are testing, it’s important to change only one thing and keep everything else the same. If you don’t, you won’t know what actually changed the result.
  2.  I suggest you check on your ads every 24 hours and about the same time daily so that with each test they will get traffic from the whole day.
  3.  I suggest that for each element you test, you start with wildly different things to get a better overall sense of what’s working for your audience.

That means if you are testing images, test one with a picture of a guy, one of a girl, one of a monkey and one of a garden. Of course, those are random examples, and you probably shouldn’t test the monkey (or should you?). The point is that they will (probably) give you wildly different results, but here and there you’ll find a hidden gem.

Whenever you test something new, I suggest that you duplicate the old campaign (can be done in Power Editor) and set up the new changes in a fresh campaign. So for every new element you test, you want to set up a fresh campaign. The reason behind it is that Facebook doesn’t like paused and resumed ads that much and whenever you change something in an existing campaign it has to be sent for review, which means it will be paused in the meanwhile.

As I bet you are a busy person with lots of things to handle in your business, I’ll limit this guide to only focus on the elements that will give you the biggest wins right now. I don’t want you to feel overwhelmed or waste time on small, useless gains when you could have spent your time in other areas of your business instead.

Now with the boring notes out of the way, let’s get to the meaty stuff!

The order of testing

How many variants of each element you should test will depend heavily on your budget. I’ll explain more about that in the reporting section. I usually follow this order:

Let’s examine them one by one.

#1. Campaign type

Retargeting, Lookalike audience or manual campaign? The first thing you have to figure out is the type of campaign you want to run.

If you want to sell more to your current customers, then you should go with retargeting. If you want to find more people similar to your current customers, you should run a lookalike audience campaign, and if you want to build your audience from scratch, you should manually build the campaign.

Action step: Pick a campaign type.

#2. Audience

Deciding on the campaign type is a strategic decision. The most important non-strategic decision is your audience, which includes two things: who your audience is and how you find them on Facebook.

The latter is what seem to cause the most problems for people and, in my experience, this often happens because they target an audience that they think is their audience when it actually isn’t. I believe Facebook doesn’t deliberately make it as tricky as it is, it rather tries to offer as many targeting options as possible so you can do more cool stuff. They do make it a bit complicated to figure out, but at least that gives us an opportunity to help out 😉

Finding your audience is more important that testing various images, text and so on because we need to find the right people before we figure out what’s best to show them. If we try to A/B test which image works best, but we show it to the wrong people, then what do we learn? – That, that image doesn’t work on those people! Yeah, right, but… they were the wrong audience so, at the end f the day, we don’t learn anything really. So, first find the right people, then the right image and so on.

I usually start out by running five ad sets within a campaign, with a different target audience in each. A target audience could be as simple as a fan-page, interest, behavior or something similar.

The purpose is to figure out which of the audiences is the best to move forward with. The best practice is to test just one interest or behavior per adset (or only one interest per lookalike audience, if you’re doing that) so that you’re able to determine which is better.

If you group several interests together (bulk-interest adset), you won’t know which one performed better and which ones didn’t.

Another thing is that if you group several interests together and you proceed to test images, you may experience that one interest gets you more traffic on the first image and another interest got more traffic on the second one. Hence your test won’t be accurate.

One scenario where I like to use bulk-interest adset is if you are targeting employees at small companies as the traffic is so little that it’s a probably a waste of time to set up a lot of campaigns. I usually set the bar at 1000 people in my audience.

Action step: Pick the 5 interests (or whatever way you choose to target) you think are most relevant to your target audience and set up a campaign with 5 single-interest adsets.

Continue with the best performing audience; we’ll save the rest for later.

If you are following this guide step-by-step, you might benefit from reading the bottom sections about reporting and statistical significance to an idea of how much traffic is enough to decide on your result.

#3. Angle

The angle is something that isn’t mentioned that much in general which is a shame as it’s crucial and will have an enormous impact on your ads. It can be described as the direction or the overarching message you show to the user.

Testing this will set you far apart from your competitors – it is something I’ve seen amazing results with myself, and it’s something that, at least in my experience, many advertising agencies don’t do. They might propose one or two core messages and select which one they are more confident with, or even worse they simply like better.

Testing blog posts to see which angle “clicks!”


As you have a great tool at your disposal and you are testing things anyway, you might as well let your audience TELL you which direction they like more. 
Testing the angle first will make it MUCH easier for you to test everything else later as you know where you are heading. Once you have found the right “topic” for your audience, it will be 10x easier to deal with the overwhelming feeling of “all the things you need to test.” At that point, you’ll know in which direction you need to look for images, what your headline and copy should be about, and what type of landing page you should send them to.

It will give you a great sense of clarity. 

If you just acknowledge that people browsing Facebook might have a different mindset when they see your ad compared to if they read a guest post, saw your ad on Google or something else, you are already miles ahead of your competitors! Why? Because you are now able to connect your angle to how your users think which makes you able to personalize your ad and that has been proven time and time again to be insanely more successful as the user can relate to your ad much better and have a much stronger feeling that it’s made just for them.

Action Step: Come up with five wildly different angles (headlines with different topics) and test them while keeping everything else the same. Pick the winning angle and move on to the next step.

Pro tip: if your audience is based on a certain fan-page or interest, mentioning that indirectly or tying an ad copy with it should improve your results a lot. Keep in mind that it’s not allowed to mention your targeting on the ad, so do it subtly.

#4. Image/Landing Page

Once you’ve found the angle that works best, notice how much easier it becomes to come up with ideas for images and landing pages – you know which direction to go in and, more important, in which not to go.

Whether you should test the image or landing page first depends on your previous results.

If you have plenty of clicks but very few conversions the problem might be that people need to be convinced by seeing something else than your landing page when they arrive from an ad on Facebook. That could be one of your best articles, your about page or even your homepage. I’ve recently seen surprisingly good results with sending people to something else than a landing page. I haven’t been able to figure out for sure why yet, but my best guess is that it appears more trustworthy or that people might want to get to know you better before they sign up.

On the other hand, if your clicks are expensive but people are still converting you’ll probably get a bigger win from getting cheaper clicks by testing more images.


I’ll soon write a blog post about how to read your reports but if you are looking for more details, head down to the reporting section in the bottom.

Images are something I’d recommend testing a lot more than the other things – in fact, the more you test, the better your results are likely to be as the image is what attracts the user to look at your ad and therefore can reduce your price a lot.

For every 10 images I test, I like to test 1 that I think will be horrible as you sometimes get surprised. If you have the budget for it, I’d recommend you to test at least 30-50 images – If you really want cheap, incredible traffic finding the right image for your audience is important. It will lower the cost a lot and, trust me; you will be way ahead of your competitors.

As you get more comfortable with Facebook ads and want to spend more time and money on your campaigns, I’d suggest testing much more than 50 images, but that it’s a great number to start out with, so you’ll get a sense of how each element affects your overall funnel.

Remember that the image has to tie in with your angle, so you are also able to convert the people you sent to your website.

Action step: Set up a campaign with 5 different images. Pick the 4 you think will work and add 1 that could work but probably won’t.

I recommend you test 5 images at a time or your spending could get out of hand if you are not used to running a lot of tests at the same time. When you find the winner of those images, rinse and repeat until you’ve tested 30 images and move on to the next step.

The landing page is what converts the traffic you send from Facebook – it may need to be converted in a different way than other traffic types.

Action step: Test 5 different landing pages and continue with the winner. Remember to have PixelCaffeine set up on all the pages you are testing so you can see conversions in Facebook reporting.

#5. Headline

You should have a fairly good idea which direction to go with your headline as you have found a topic to go with (angle), an image and a landing page that works well.

The headline is less important than the image as that is used to get the attention of the people.

The difference between the angle and the headline is that before we were looking for the right topic and now we are trying to improve the copy of the headline to get more clicks without changing the topic too much. That could be done by adding words like free, secret, hidden and similar words that has been proven to create curiosity.

Here’s our latest post about headlines and a good article from SumoMe.

Action step: Come up with 5 different headlines and test them.

The headline along with the copy (description) is generally what convinces the reader to click and sets their expectations for when they arrive on your site.

#6. Copy (description)

The ad’s description should explain a bit more about what the user can expect from clicking the ad. Keep it simple and remember to look at the preview of both desktop and mobile if you are running these to make sure that your copy is displayed as you’d like it to.

Action step: Test 5 different types of copy.

#7. Campaign objective

Some people will argue that you should test the campaign objective (e.g. website conversions or clicks to website) earlier in the process, that could be a good idea indeed. However, if you are just starting out, I believe it’s better to become comfortable with other things as they will most likely give you a bigger win right away as well as a better understanding of how to work your audience from Facebook.

When you have dialed in your audience and your ad, it’s time to test which campaign objective will work better for you. This is something that is very difficult to predict and something you’ll want to test for any new campaign.

I imagine you would want “clicks to website” versus “website conversion” objectives. Below are screenshots for Power Editor and Ads Manager, depending on what you are using.

(in Power Editor)

(in Ads Manager)

But if you want the easiest of all, here it is AdEspresso:

There are plenty of places where you can place your pixel and let Facebook optimize for that but, again, to keep things simple for now all you have to do is place it on your “thank you” page. Either for sales or subscribing to your emails.

Action step: Set up two campaigns: one with “clicks to website” as objective and one with “website conversions.”

#8. Placement

Placement is where your ad is shown – that could be on mobile newsfeed, desktop newsfeed, right column and so on.

I’d recommend starting out by just running on mobile and desktop newsfeed until you get to this step, and then test them all.

The reason for the newsfeeds are that they are generally performing well, so just like above let’s keep it simple for now.

Action step: If you are starting from scratch on this guide: test only mobile newsfeed and desktop newsfeed. You can see the result in your reports by breaking them down by placement (screenshot below from Ads Manager and AdEspresso).

Ads Manager – Placement

Of course, using AdEspresso makes it easier and quicker: 

AdEspresso Placement Testing

If you have been through the guide and you are ready to test placements: duplicate your campaigns and add all of them except Instagram. I recommend testing that later as it’s a whole different platform with its own quirks.

#9. Misc.

We could move on to test CTA buttons, bid type, ad type and a million other factors – all things AdEspresso makes way easier to do than Power Editor. But you’ll always have to ask yourself the question: are the gains you’ll get from optimizing worth the time you spend on it (instead of doing other things for your business)? Is it worth the effort?

Long story short: if you are not happy with the results you will need to go back and optimize further. If you’d like to test more things, consider trying CTA buttons, the bid type, and the ad type. 

If you are happy with the quality and cost per lead/sale, I will stop with fine tuning, and I’ll move on to scaling that up, which is where Facebook really shines!

Reporting

What to look for when you’re starting out? Just pick one or two metrics to keep things simple – I’d suggest “unique link click-through rate” (CTR) and “cost per click” (CPC) if you are running a “website clicks” objective or “cost per conversion” if you are running the “website conversions” objective.

Make sure you have enough data to test, or your results won’t be trustworthy – they won’t be “statistically significant.” Around 100 clicks For “website clicks,” or about 10 conversions for “website conversions” -for each element, should be good. (That means if you are testing 5 images you’ll need about 500 clicks in total as 5 images*100 clicks = 500 clicks).

Remember, when you look at reporting to look at the breakdowns, they might offer you some interesting insights such as different age brackets that perform better than others. Here’s a screenshot of how to do it in Ads Manager.

Want more? Another advantage of using AdEspresso – is that you can get the results of your testing in PDF format (instead of Excel) as well as having them automatically sent to your email based on a schedule that fits your needs.

Statistical significance

Statistical significance helps you determine how sure it is that one ad performs better than the other – most often people accept a 95% certainty. It doesn’t take your demographics size into account so you will need to know when you have enough data – as mentioned above around 100 clicks or 10 conversions per adset depending on your objective should be good.

Here’s a free calculator that you can use to easily check the difference in conversion rate and certainty.

If you are optimizing by account conversions, just do as it says, and if you are optimizing by clicks just enter the clicks into the second row as shown in the screenshot below:

If you are running more than two variables (e.g. images) at the same time, you should test each one against the best performing one.

Problems with split-testing

You’d better be prepared to lose money, at first, to make this work. If you are unsure or only have a very limited budget, I’d recommend not doing this as you’ll probably be too worried about the results and not ready to test enough to find the combo that gives you what you are looking for. You’ll end up with a bad experience.

A good budget depends on the price, as we are looking to buy enough clicks or conversions to make sure that each test we run is tested properly. The price will vary depending on a lot of metrics, but having set aside a couple of thousands is not a bad idea.

As we are talking about the mindset of testing, Facebook has a tendency to be a bit slow with reporting so don’t be obsessed with refreshing the stats all the time. If you check your numbers every 2 hours, you may think they are performing really bad, panic, and you’ll stop them right away to avoid losing more money. BUT those numbers could change just a few hours later, that’s why I’d recommend checking once or twice daily. Go work on something else in the meantime and let Facebook do its thing.

You should also be aware that if you test on audiences too small, your price will increase, as Facebook will want to spend your full budget for the day. It means that if Facebook can’t generate enough traffic to suit the budget you set, it will increase the price per click, impression or whatever you’ve chosen to pay for). So be aware of small audiences – I’d be cautious about testing much on a group smaller than 5000 users.

I noticed some people were commenting on other posts, worried about showing too many ads to their audience. Unless your audience is colossal, I wouldn’t worry about that as it won’t make much sense to start methodically split-testing a lot of stuff on a small audience you already own – what would the result be?

Testing on new audiences you want to attract makes a lot more sense especially because you can scale your results.

If you already have a large audience I wouldn’t worry much about the most engaged users seeing multiple posts; but if you do, you can always test using unpublished posts and then post the final one on your page.

Key takeaways

And now it’s your turn! Use some or all of our tips and let us now how it goes! I’m cheering for your success 🙂