Facebook’s AI algorithm for ad delivery
Facebook’s use of machine learning became popular through its surprisingly accurate “People you may know” suggestions. Since then they’ve expanded from machine learning to other types of AI, like generative AI, pushing AI use to the forefront of all their advertising products.
So, how do these complicated algorithms work, and what makes them so accurate? It all begins with gathering data. Facebook gets the information needed to feed its AI from two main sources:
- Data users provide about themselves like demographics or pages they’ve liked
- Data advertisers provide, like their email mailing lists
The next step is combining this information with your actual ads. To decide which ads to serve to which specific users, Facebook needs to determine the ad’s total value. It uses the following equation:
Advertisers Bid x Estimated Action Rate + Ad Quality = Total Value
One of the advantages to machine learning being involved is that the more your ads are shown, the more it can make improvements to the delivery of your ads. This increased the overall effectiveness of your campaigns.
Let’s break down the equation to see how you can create your ads to be best suited for this algorithm.
How is the estimated action rate determined?
Facebook determines the estimated action rate by using machine learning. It predicts how likely a user is to complete the goal chosen by the advertiser. For example, if the advertiser wants to gain more traffic to their website, then the ad will be shown to someone who is more likely to click on it.
Factors like the users behavior on Facebook and other websites come into play, in addition to many other things like the time of day and the ad content.
As we mentioned earlier, businesses can provide additional information to Facebook about the actions users take on their website outside of Facebook. This is possible by using Business Tools, which are resources Meta provides for businesses to more easily sync with Meta.
Some of these tools include Meta Pixel which gives you a piece of code to add to your website. This code allows you to track visitor actions on your website and then share that information with Meta. Once the machine learning algorithm has access to this data, it helps you to serve highly personalized ads.
Another Business Tool is Meta App Event tracking which serves a similar purpose to Meta Pixel, but is used for your application. It records events like purchases or app downloads.
Thanks to the complexity of Facebook’s machine learning algorithms, advertisers don’t need to consider each of these factors individually and decide who to show their ads to. Instead, they can simply set their business objectives and let AI do the heavy lifting.
How is the ad quality determined?
The quality of your ads is determined over time based partly on the feedback users give and whether or not there are any “low quality assets” detected.
Some of the user feedback taken into consideration is a user hiding ads and the reason they given, reporting ads, and how long they stay on your landing pages.
Indications of low quality ads
By avoiding these issues, you can show Facebook that your ads are of quality.
- Not providing full context in order to make your products or services sound better than they are. Your ads shouldn’t come across as being “click bait”.
- Using exaggerated language in your ad copy.
- Not using original content.
- Ads that go against the advertising guidelines
What role does the advertiser’s bid play in the ad auction?
The bid amount you pick is one of three choices you have to influence the ad delivery, the other two being your preferred audience and business objectives. The bid you set tells Facebook how much you’re willing to spend in order to meet your business objectives.
Note that the highest bid doesn’t always win the bidding war. Sometimes a lower bid will win the auction if Facebook decides that the user will be more likely to create a conversion.
7 Facebook AI advertising features every marketer should know
Meta has two main areas, AI Sandbox and Meta Advantage, where they implement AI features compatible with Facebook ads. Let's explore 7 features you should know about.
AI Sandbox features
As Meta continues to invest in their AI features, they’ve introduced several innovations that aim to help advertisers create and deliver more effective digital paid ads. The AI Sandbox is an experimental space for testing these new AI-powered ad tools.
One big component is the introduction of generative AI tools. It currently includes three core components.
1. Text Variation
This feature generates multiple versions of ad text to highlight key points, allowing advertisers to experiment with different messaging for specific audiences.
How can you use this with your ad campaigns?
This tool allows you to highlight important points of your copy and try different messages for specific audiences. By using AI to create text variations, you can quickly and easily test out different messages to see what resonates best with your target audience.
2. Background Generation
Advertisers can now create background images from text inputs, enabling quick variations in background visuals and diversification of creative assets.
How can you use this with your ad campaigns?
Use this tool to quickly experiment with different backgrounds and diversify your creative assets. For example, you could input a few keywords related to your brand. The tool will then generate a variety of backgrounds incorporating those keywords. This can help catch the attention of potential customers and make the ad stand out.
3. Image Outcropping
This feature adjusts creative assets to fit various aspect ratios across different ad placements, such as Stories or Reels, reducing the time and resources required to adapt creative content.
How can you use this with your ad campaigns?
You can save time and resources by automatically resizing images to fit different platforms, rather than creating a separate version for each platform. Using this tool can ensure your ads look good across multiple platforms, reach a wider audience, and comply with ad requirements.
Meta Advantage AI features
Meta Advantage is a suite of ad automation products that leverage AI and machine learning to optimize campaign results. These enhancements aim to make advertising campaigns smarter, faster, and more efficient.
4. Easily switch to Advantage+ Shopping
Advertisers can transform existing manual campaigns into Advantage+ shopping campaigns with a single click. This feature makes it simple to create a Facebook ad AI campaign strategy.
As Meta claims these ads will save advertisers time and money, this feature will come in handy to anyone wanting to make the switch.
5. Performance Comparisons
If you do make the switch to Advantage+ shopping campaigns, you’ll want to make sure that the change was for the better. Thanks to Performance Comparisons, you can receive an automatic report that compares the performance of your manual sales campaigns to Advantage+ shopping campaigns.
You’ll then be able to see any benefits of adding automation to your campaigns, or if there’s something you should be doing differently.
6. Video Creative in Catalog Ads
Instead of only being able to use static image creatives in catalog ads, you can use videos as well. Uploading various video creatives allows for dynamic content delivery across various ad placements using AI and machine learning.
Some examples of videos you could upload are:
- Customers using your products. This would be the video equivalent of lifestyle images.
- Videos that showcase your brand
These videos can be shown to users as the watch Reels, Watch, and Stories, or as they scroll through their feed.
7. Advantage+ creative
Advantage+ creative is a product that uses AI to help advertisers make changes to your ad creatives like adding music, 3D animations, and minor improvements to images and videos. These improvements can include adjusting the contrast and brightness and standard enhancements.
Standard enhancements is a feature within Advantage+ creative that will create multiple versions of your images with various enhancements. It then determines which image to show to which users based on their personal preferences.
According to a study run by Meta, these small, personalized changes to your images can result in 14% more incremental purchases. This shows that using the standard enhancements options makes for a powerful Facebook ad AI optimization strategy.
How can you create Facebook ads using AI?
Part of the Advantage Meta suite of products are the Advantage+ campaign ads. In 2022, all Facebook ads were moved under the Meta Advantage umbrella and given new names. We’re using the updated names here, but will specify their previous names as well.
Each of these campaigns use machine learning and AI to reduce the amount of manual labor required for advertisers in the set-up stage.
Advantage+ Catalog ads
Formally known as Facebook dynamic ads, Advantage+ catalog ads provide a highly efficient and effective way to reach the right audience at the right moment with the help of AI automation.
How do Advantage+ catalog ads work?
With Advantage+ catalog ads, you can automatically serve relevant products ads to people based on their interests, intent, and actions. This ad type removes the burden of creating individual ads for each and every item you sell. Instead, you’ll give your product catalog to Facebook and let it do the heavy lifting.
This approach allows you to serve personalized recommendations for all the products and services in your catalog, eliminating the need for manual ad set-up. Whenever a shopper shows interest in an item from your catalog, Meta can dynamically generate an ad specifically tailored to that individual.
It then delivers the ad automatically across mobile, tablet, and desktop devices.
Who are Advantage+ catalog ads for?
If you’re an advertiser whose focus is scaling and automating your ad campaigns then Advantage+ catalog ads could be for you. You may have a large product category or wide variety of products. These ads would be beneficial for you because it eliminates the need to create each ad individually.
Advertisers looking to target audiences based on user interests, intent, and actions can also use Advantage+ catalog ads effectively. These ads use machine learning to deliver personalized recommendations to users, aligning with their preferences and behavior.
Advantage+ catalog ads can run on a variety of advertising objectives, such as:
- App Promotion
Catalog ads offer two targeting options. We’ll take a look at both, outlining what they’re used for and when you should choose them.
The first lets advertisers reach consumers across different stages of the buying journey, including those who have not yet made a purchase but have shown interest in their offerings.
Consider setting up your dynamic ads with a broad targeting strategy. This can ensure your ads have the widest reach possible while also targeting shoppers with a high purchase intent.
The second targets those who have expressed interest in specific products (retargeting). This flexibility in audience targeting makes the ads suitable for businesses aiming to engage both potential and returning customers.
Dynamic ads are highly effective for retargeting users who have previously interacted with your brand through your website. For example, someone might have shown interest in a product by adding it to their cart, but ultimately never moved on to purchasing it.
Facebook’s AI can then identify these users who have shown interest and what specific products they were looking at. It then reminds shoppers about these products with relevant ads, increasing the chances of conversion.
Tips to consider while setting up dynamic ads for retargeting:
- Don’t target shoppers who have recently made a purchase as they are less likely to make another one so soon.
- Use a wide retention window. For example, you can retarget shoppers who have visited your site in the last 90 days on every type of device.
Advantage+ shopping campaigns
Advantage+ shopping campaigns are designed to be the most efficient solution for performance-focused advertisers looking to drive online sales. These campaigns use machine learning to help reach valuable audiences with less set up time and greater efficiency.
Facebook campaign creator
Who kind of advertisers are Advantage+ shopping campaigns for?
There are three main groups who will benefit from these ad campaigns the most:
- Retailers selling directly to consumers
- Retailers selling fast-moving consumer items
- Online retailers who run a single brand
How do Advantage+ shopping campaigns work?
Advantage+ shopping campaigns use machine learning to identify and aim for the highest value customers across all of Meta's family of apps and services. They require a lot less set up than manual ads do.
Advantage+ shopping campaigns target audiences
You can also run up to 8 campaigns at a time in each country. This means that rather than taking the time to set up multiple campaigns each targeted to a different audience, you can instead automatically target larger audiences by showing them ad variations best suited to their tastes.
These campaigns put the customers you already have with potential customers all under one campaign as your target audiences. It then chooses the products from your catalog that are best suited for those audiences.
By running fewer campaigns and automatically testing up to 150 creative combinations, this solution delivers the highest performing ads to the highest value shoppers.
Advantage+ shopping campaigns objectives
There is one objective available for Advantage+ Shopping campaigns – Sales. This objective is focused on driving online sales of your products or services, making it suitable for advertisers who want to maximize their conversion rate and revenue.
Advantage+ app campaigns
These campaigns used to be called automated app ads.
How do Advantage+ app campaigns work?
The aim of these ads is to increase the amount of app installs you receive. Advantage+ app campaigns use machine learning to help you deliver the most creative and personalized ads possible, all while reducing the amount of manual labor required.
You’re able to create up to 50 ads in bulk at one time, removing the need to create duplicate ads.
Advantage+ app campaigns targeting options
There is a more limited amount of targeting options available for Advantage+ app campaigns, which are:
- Device operating system
- Minimum age
- Custom audience exclusion
- Location settings
These simplified options ensure that your ads won’t be competing against each other.
Meta Advantage campaign budget and bidding strategies
Meta Advantage campaign budget is a product that used to be called ‘campaign budget optimization’. Rather than setting individual budgets for your ad sets, this feature allows you to set one budget for your campaign.
Which budget type should you choose?
The budget you set will then be automatically dispersed to your ad sets based on where it will be the most effective.
There are two options you can choose from:
- Daily Budget: Pick this option to set the average amount you want to spend daily. Since it’s an average, you’ll spend a little more some days and a little less on others. It might be the right choice for you if you want to get consistent results each day.
- Lifetime budget: Choose this option if you have a predetermined sum that you don’t want to go over for your budget. It also gives more flexibility to how much is spent each day.
If using more Meta platforms, you can decide what you want your Facebook and Instagram AI ad spend to be, and if you prefer daily or lifetime budgets.
Which bidding type should you choose?
There are three main categories of bidding strategies to choose from, with specific choices for those options.
1. Choose spend-based bidding to make sure you’re using up all of your budget and squeezing as many results as you can out of it.
There are two options within this category:
- Highest volume: Helps you get the highest possible amount of results (don’t choose this option if you want to optimize for Cost Per Action).
- Highest value: This is the option to pick if you want to get the most value out of each conversion, rather than focusing on the quantity of conversions.
2. Goal-based bidding allows you to choose if you want to reach a certain value or cost.
- Cost per result: This goal will prioritize keeping your costs the same as the amount you set, no matter what is happening in the market at the time.
- ROAS: Choose this goal to set the average amount that you would like your ROAS to be throughout your campaign.
3. The third option is Manual bidding with the goal option of ‘bid cap’ but it should only be used by advertisers who have a firm grasp on how to predict conversion rates.
How to set up your Advantage campaign budget
Follow these steps to set up your budget.
- While creating your campaign, switch the ‘Advantage campaign budget’ toggle on.
- Pick which budget type you want to use (Daily or Lifetime).
- Fill in the budget amount you’ve decided on.
- Pick the bidding strategy that best aligns with your campaign needs and goals.
- At this stage of the set up, advertisers who chose to use a Lifetime budget will be able to set a schedule for their ads to run on.
Advantage campaign budget best practices checklist
- Have at least 2 sets of ads to get the best results.
- To determine how well your budget is working for your campaign goals, you should check your results at a campaign level rather than at the ad set level. This is because your budget probably won’t be equally distributed between your ads.
- Keep the amount of ad sets per campaign to 70 or below. There are two reasons for this. The first is that you won’t be able to make as many edits you can make after publishing the ads if you have more than 70. The second is that the more ad sets you have, the longer your campaign will spend in the learning process. This prolongs the time until you’ll be able to experience the benefits of Facebook’s machine learning.
- If you’re going to add more ad sets after publishing the campaign then add them in bulk because it can take up to 2 hours for the campaign budget optimization to calibrate. Additionally, make any setting changes in bulk after your ads are published because each change causes your ads to go back into the learning phase.
- Manually paying and unpausing your ad sets will disrupt the AI’s flow of where it distributes your budget.
AI for Facebook ad targeting
These Meta Advantage features use AI in order to target the right audiences for your campaigns. To be sure you’ve chosen the right targeting option, you can conduct A/B testing.
Using AI to find new Facebook audiences
The following options use AI to find additional audiences to target outside of the ones you’ve already set.
This feature will automatically kick in and show your ads to users outside of your set lookalike audiences if Meta’s AI predicts that it will positively impact your performance. You can enable this setting for the following campaign objectives:
- App promotion
You’re also able to exclude any audiences you think won’t provide you with conversion while this feature is selected.
Advantage detailed targeting
Advantage detailed targeting looks for profitable opportunities outside of your defined detailed audiences. Because of this, it won’t affect age, location, or gender targeting.
Depending on the campaign objective you chose, this option might be automatically enabled. For other objectives, you’ll be able to switch it on yourself in the Detailed targeting section of Ads Manager.
Advantage custom audience
Like the other two options, Advantage custom audience uses AI to find additional users to target that will benefit your campaigns and are outside of your original selected audience.
When setting up this campaign, you can enable or disable this option. Just be aware that if you turn it on for an ad you’ve already created, it could trigger the learning phase again.
This product offers a more flexible approach to audience targeting. Instead of rigid constraints, advertisers can suggest audience inputs, allowing the ad system to find additional potential customers outside those suggestions.
Meta claims that this targeting option is the easiest and most effective one it offers to date, recommending it for all campaign types except remarketing. Its AI system uses data such as:
- Previous conversions
- Data from Pixel
- Interactions with your ads
If you’re not comfortable with Facebook’s AI completely taking over your targeting, you can still have some control over who sees your ads. For example, you can add Audience suggestions and set Audience controls to exclude certain groups from being targeted.
AI use in Facebook ads - Case Study
Meta released a case study they conducted with fashion brand ONE BOY. They added an Advantage+ shopping campaign to their existing advertising strategy to see if it would increase their online sales.
With the Advantage+ shopping campaign they were able to fully utilize Facebook’s machine learning by:
- Removing the need for manual campaign set up
- Avoid audience overlap by putting all of their audiences into one campaign
- Test out different photo and video combinations (up to 10) to know which one performed best
- Making the most of their flexible budget to ensure their ads were seen by new and existing customers
After running an A/B test between their regular campaign and the Advantage+ shopping campaign with their regular campaign, they got the following results:
- 42% more purchase events
- 34% ROAS increase
- 29% decrease in customer acquisition costs
Facebook AI advertising is becoming easy for all marketers to use as Meta increases the capabilities and offerings it provides.
The AI Sandbox, Meta Advantage suite, and ongoing AI infrastructure investments reflect their commitment to improving the experience for both advertisers and users. Further innovations in AI are expected across all Meta apps as they continue to explore the potential of AI-driven campaigns and features.
Check out our article All-in-one Guide to Using AI in Digital Marketing to learn more about taking advantage of AI in your ad campaigns and strategies.