It’s a question we get all the time – how do I scale my Facebook ads once I find what works? We get many clients who find some success on their own using Facebook ads but then have trouble scaling up spend while maintaining good results. Often their campaigns will not spend or the CPA/ROAS tanks as soon as they try to increase budget.
So how do we do it? It’s part secret sauce, part tried-and-true best practices and we’ll break them down for you here.
This guide is adapted from Facebook’s own published best practices (along with our expert additions) and is broken down into 3 sections based on how much traffic your website gets on a monthly basis. Make sure you have the Facebook pixel installed on your website before beginning this guide and verify that it is set up properly. The sections are broken down by the number of monthly pixel events on your website – pixel events are any time someone completes an action on your website such as viewing a page, downloading a guide, or making a purchase.
About Facebook Ads Optimization
Optimizing Facebook ads is a bit different from optimizing in AdWords, Amazon, LinkedIn, or other ads platforms. Whereas in those platforms much of the optimization and scale comes from identifying successful segments and growing them, in Facebook it’s often about steering Facebook in the right direction, and letting them auto-optimize based on your pixel data. You tell Facebook what your goals are (leads, sales, downloads, etc) and they measure who is completing those goals and tailor your ad delivery accordingly. Thus, the more successful conversions you send to Facebook via your pixel, the more they know about what types of people convert and the better they’ll deliver your ads. This leads to CPAs dropping, ROAS increasing, and the ability to scale. How you get to that point varies based on how much data Facebook has from your pixel.
This guide is for websites that have had less than 500 pixel events in the last month.
Before optimizing for bottom-of-the-funnel conversions, Facebook needs to learn more about your website visitors and the types of people who complete your desired actions. With less than 500 pixel events, they don’t have enough data yet to be able to effectively optimize your ad delivery toward your goals so it’s important to keep your optimization goal high up the funnel at this point.
Objective & Optimization Guidelines
With less than 500 pixel events per month, we recommend using either the “traffic” or “conversions” objective and optimizing toward “clicks” or “landing page views.” This will get traffic to your website for the lowest possible cost and Facebook will be able to start learning what types of people interact with your content and complete events down the funnel.
At this point in your journey we recommend using a lowest cost bid strategy with no bid caps. This keeps things as open as possible and let’s Facebook bid toward your goals with as few restrictions as possible.
You want to give Facebook as much information as you can in order to start them in the right direction with ad delivery. Here are the best targeting types to use here:
- Email List Custom Audiences – If you have a list of past purchasers or even newsletter subscribers, upload that list and target them directly and/or create lookalike audiences to target people similar to your existing list.
- Demographic / Interest Audiences – If you don’t have an existing list to upload, target by demographics and interests in order to find the people you think would be interested in your product or service. Here’s a full list of Facebook and Instagram ads targeting options and below are some guidelines to steer you in the right direction:
- Geographic – where are your customers located geographically?
- Age – what age ranges are your typical customers?
- Gender – are your customers primarily male/female or a mix of both?
- Interests – what pages and brands would my customers likely follow on Facebook?
- Exclusions – don’t forget that you can exclude people: past customers, certain pages, etc.
Remember with your targeting that you should only include demographic or interest targets if you think you know your audience better than Facebook does. Here’s Facebook’s official word on the topic:
“Do I have better insight into a given type of targeting than Facebook would? This can be difficult to know in every situation, but any time you think you have information that we don’t have, you should provide that information through targeting to help us find the best people to show your ads to. If you don’t think you’d know better than Facebook, then you shouldn’t add that to your targeting criteria. Keep in mind that Facebook’s delivery system is trying to show your ads to the people most likely to get you the result you’re optimizing for no matter how broad your targeting is. Because of this, ill-informed targeting choices that restrict our delivery system could end up hurting performance rather than helping it.”
This guide is for websites that have had 500 – 10,000 pixel events in the last month.
Once you’ve exceeded 500 pixel events per month, you can move on to the Intermediate guide and start experimenting with various conversion-optimization strategies. Optimizing toward website conversions means telling Facebook what action you want people to take on your site and Facebook will attempt to deliver your ads to people most likely to complete that action.
Your action can be either a standard event (lead, add to cart, purchase, etc) or a custom conversion that you define on your own.
- Conversion volume – When optimizing for conversions, Facebook recommends that you have at least 50 conversion events per week per ad set. When setting up your ad sets and selecting the conversion events, they recommend selecting one with at least 100 conversions per month without running any ads to it. This is so they’ll have enough data to work with in order to optimize the traffic toward your goals.
- Move up the funnel – Rather than starting a new ad set and optimizing for purchases right away, try moving up the funnel and optimizing for product page views or adds to cart.
- Use automatic placements – This gives Facebook the ability to find conversions across Instagram, Messenger, Audience Network etc. The bigger the pool they have to choose from the better.
- Website Custom Audiences – Now that your pixel is recording a decent amount of traffic, use that data to create audiences. You can retarget people who have visited and not purchased, or you can use Lookalike Audiences to find more people similar to those who did complete a conversion on your website.
- Audience Expansion – Use the same demographic and interest guidelines as we recommend in the beginner guide, but also test out the audience expansion functionality. This allows Facebook some leeway to expand your detailed targets to a wider pool of similar people.
- We recommend starting with a lowest cost bid here and then moving to a target cost bid once you see what your cost per conversion is at the start. The risk in using a target cost bid from the start is that you set your bid too low and Facebook won’t deliver your ads because they don’t think they can meet your goal cost per conversion. If you do want to start with a target cost bid, just set it high at ~2x your actual target cost per conversion.
Here’s Facebook’s official word on bidding:
“Keep in mind how each bid strategy and ad auctions work: A cost target tells our system to bid for results with an average cost as close to your target as possible. If you use a cost target, set the average amount you actually want to (or can afford to) pay. A bid cap sets a maximum amount we’ll bid for you, and bids are not the same as what a result costs. In fact, you may end up paying less than your bid in many cases. Given this, consider setting your bid cap at the maximum amount you’d be willing to pay for any given conversion.”
This guide is for websites that have had over 10,000 pixel events in the last month
If you have over 10,000 pixel events per month, you’ve seen some success driving sales or leads through your website and you’re looking to scale up your campaigns, this is the right section for you.
“Old school” paid advertising methodology was to hyper-segment your audiences into small buckets in order to maximize your budget and control costs. This doesn’t work in Facebook and in general they recommend combining audiences into a single ad set with a larger budget. Here’s their explanation for why this is better than keeping things separate:
Each ad set accumulates data. An ad set (not a campaign) needs about 50 of the conversion it’s optimized for for Facebook to be able to deliver it with stability and efficiency. If you have 10 ad sets that get 5 conversions each in a week, you have 50 conversions, but also 10 ad sets without enough data, and which may stop being delivered. If, instead, you have 1 ad set with 50 conversions, your campaign has the same amount of conversions as the other set up, but also an ad set with enough data to start delivering more stably and efficiently.
Say it takes $100 to get enough conversions to start delivering an ad set stably and efficiently. If you have 10 conversion-optimized ad sets with $100 budgets, then, in your best-case scenario, you’ll have spent $1000 on data gathering and $0 on optimized delivery. If you have 1 ad set with a $1000 budget, $100 is spent on gathering data, and $900 is spent on optimized delivery.
So while it goes against much of what we consider paid advertising best practices, you may find better performance with this method on Facebook.
As you scale up spend, you may find that your cost per conversion starts to rise. This is partly due to exhausting the low-cost conversions in your existing audience(s). In order to combat that, it may be necessary to expand your audiences and Facebook recommends utilizing lookalike audiences heavily here. Here are a few specific tips for your lookalikes:
- Avoid broad or low quality lookalikes – Instead of creating a lookalike of everyone who visits your website, build your lookalike based on only people who add a product to their cart. The problem with the former is that it doesn’t differentiate between low-quality, average, and great customers.
- Focus on source quality – The ideal size for your lookalike audience source list is 1,000 – 50,000 people but it’s more important to focus on the quality of that list rather than the size. So if your ideal customer list is only 500 people, try that out instead of expanding to a lesser quality list just to hit the 1,000 person minimum.
- Expand high quality lists rather than mimicking them – If you have a high quality list and a 1% lookalike audience that is performing well, test a 2%, 3%, 5%, etc version of the lookalike audience. This should provide better results than trying to create a similar high quality source list with a 1% lookalike.
- Avoid targeting on top of lookalikes – Lookalikes already factor in age, gender, interests, etc so trying to layer on additional constraints can be detrimental to their performance. Facebook recommends that if you still want to layer on additional targeting options, just use targeting expansion to give their delivery system as much flexibility as possible.
Optimize for an event as close to your true goal as possible. For example, if you’re optimizing for AddToCarts, but you have at least 50 people initiating checkout per week, try moving the optimization goal to the InitiateCheckout event to steer Facebook closer to your higher quality users.
Test, test, test
As with everything we do, best practices are great but it’s important to always be testing. Here are some guidelines from Facebook around how to test efficiently:
- Use target cost bids when testing
- Give your tests at least 1 week to run, preferably 2
- Shoot for at least 10 conversions per ad set when testing and make sure you have the budget to afford that. If your cost per conversion is $30, you’ll need $300 per ad set to test.
Here are a few tests Facebook recommends running
- Test static images vs videos in otherwise identical ads to see which performs better
- Test lookalike audiences based on your pixel data vs your CRM (email list) data
- Test different conversion windows, such as a “7 day click” window vs “7 day click, 1 day view” window
And a final bit of info from Facebook on testing:
“Remember that you can’t test and optimize at the same time. Putting restrictions on our delivery system in order to do a well-run test prevents the system from choosing the best results from the full-range of options. We fully endorse testing to help you find the best strategies. (We even recommend consistently dedicating a portion of your budget to testing so you can stay aware of what’s happening in our dynamic ads marketplace and experiment with new features.) However, if you need to do a major test, do it 5right and then act on the results. Let our delivery system optimize based on what you’ve learned.”
Now you have all the fundamentals you need to start having some success scaling up your Facebook Ads campaigns. Start with the tips here as the foundation and then let the data guide you to what works best in your unique case. Every business is different and your optimal campaign isn’t going to look like anyone else’s. But if you use the methods explained in this guide to build and test your campaigns, you’re going to be much more likely to find just the right combinations that work well for you.
Need some help creating Facebook Ads campaigns that work? Contact Us today to find out how we can help you reach your marketing goals!