Facebook, Instagram and Google are the biggest players in the ad system space. With huge audiences and opportunities to fine-tune campaigns, this is where the majority of marketing dollars are going in 2020. However, even some of the most experienced marketers can get confused about Facebook and Google’s ad systems.
Facebook decided to clear up misconceptions and release information for how their ad system actually works. In this guide, you’ll learn how Facebook chooses ads to display, leverages machine learning and allows users to navigate controls.
Before heading into some facts, we first want to dispel some common myths surrounding the Facebook advertising system. Firstly, no data is sold to third parties. Unless users give express permission, Facebook doesn’t share any information that will identify individuals with advertisers or any other party.
Secondly, Facebook does not use the microphones on devices in order to provide more accurate ads. Similarly, it doesn’t read text messages to provide a more tailored News Feed.
In the overview released recently, Facebook said that its overall goal is to provide ad experiences based on personalization and data. When the right people see the right ads, everybody gets value:
But how does Facebook deliver the right ads at the right time?
According to Facebook, there are two main factors that determine which ads to show users. Advertisers on Facebook are aware of the variety of advanced targeting tools. When creating a campaign, we choose a target audience as the starting point. Rather than having every single Facebook ad competing for a placement, the ad platform is segmented into niches.
While some audiences are based on demographics such as gender and age, others focus on clicks, app downloads or other types of conversions. Alternatively, the advertiser might have a list of email addresses from people who have visited their website. With Facebook, we also need to consider a lookalike and custom audience. No matter how the audience was formed, Facebook always considers this first.
From here, everything proceeds on a case-by-case basis. When an ad opportunity arises, Facebook gathers all the ads that have the individual as part of their target audience. As it so happens, that person will be in one company’s audience because of their age, another because of their previous actions, another because of their gender and so on. Which ad wins? This is decided at an auction.
In the past, people simplified the auction process. The advertiser that bids the most will win, right? Not necessarily. In reality, Facebook calculates a total value score for every advertiser. The following formula is used:
Bid x Action Rate (Estimated) + Quality
The estimated action rate is the likelihood that an individual will take the action the advertiser wants; this could be an app installation or a website visit. As you can see above, the estimated action rate is multiplied by the bid.
After this Facebook adds the quality score, which is an assessment of the overall ad quality. You might be wondering how Facebook calculates the estimated action rate and the ad quality score. The answer? Machine learning.
Most tech solutions require some form of programming to know what actions to take in a given situation. The difference with machine learning is that it gathers data and makes decisions without explicit programming. Over time, the system will gather more data than any human could process. That is how we arrive at an accurate estimated action rate and ad quality score.
As we’ve just learned, the estimated action rate assesses how likely somebody is to take the action that the advertiser desires (the action selected by the advertiser when setting up the campaign). The machine learning models will consider behavior on Facebook, behavior away from Facebook, time of the day, ad content, and general interaction people have with ads.
Behavior on Facebook includes actions taken on the Facebook app, history of ad clicks, likes and engagement with other posts. Actions taken apart from Facebook are reported from businesses that share data through Business Tools. With this compatibility, Facebook can see when people buy a product, visit a web page or install an app.
Using similar systems, Facebook will assess the quality of an ad based on previous performance. If a lot of people have hidden an ad, this tells Facebook that the ad doesn’t generate results with the intended audience. Meanwhile, an ad with a lot of clicks will have a high ad quality score.
Facebook will gauge the low-quality attributes that exist in an ad. This could be engagement bait, too much text, or sensationalized or over-dramatic language. If your ad displays one of these poor qualities, it will have a lower quality score compared to ads that avoid these attributes.
You can now see that bids play a small role in determining whether or not an ad will win at the auction stage. Because this information isn’t widely known, many marketers assume that outspending other advertisers will gain the best results. Instead of spending more than everyone else, decrease the required bid amount and focus on improving your estimated action rate and ad quality score. Improve these variables by creating an ad that resonates. For this, you’ll need to know your audience and craft outstanding creatives and messages.
Why is this new clarification from Facebook important? For one thing, the machine learning models used at Facebook are only getting stronger and more accurate. As more ads display, Facebook generates feedback and the models continually improve in terms of gauging ad quality and estimated action rate. Considering the enormous size of the Facebook advertising platform, it is making significant strides every day.
As marketers, we need to realize that bidding isn’t everything. When businesses attempt to advertise on Facebook, we see two big problems:
While the first leads to a campaign that never wins auctions, the second leads to overspending and relatively mediocre results. Today, the message is that the highest bid doesn’t necessarily win the auction. Facebook openly admitted that many ad placements are actually won by the lowest bid. The platform doesn’t just consider money, it wants high-quality ads to which consumers will respond. If it thinks that your ad is poor and that people won’t respond, it would take a huge bid amount to even be considered (and still you might struggle).
For small businesses, the lesson is that you can win ad placements against even the largest brands in the world. It doesn’t matter if you can’t compete in the bid amount; you will still win with a better ad quality score and estimated action rate. Facebook has made this system more accessible for businesses of all sizes, which means good ad placements are just as possible for startups as established names.
In explaining the ad system on Facebook, we shouldn’t ignore the controls that consumers have in the ad experience. For example, all Facebook users can change their ad preferences. After consumers make adjustments, Facebook is sometimes limited in what information it can use to decide ad placement. As well as hiding ads and reporting ads, users can opt-out of ads that are based on partner data.
If you use Facebook yourself, you’ll see ‘Why Am I Seeing This?’ buttons next to ads. Here, users can get context for why they are seeing certain ads and change their preferences if they’d like. Finally, users can get a summary of all off-Facebook activity and clear this information completely. When any of these three things happen, it makes ad placement and advertising in general more difficult.
Thank you for reading our guide to the Facebook ad system. Please feel free to pass it on to other businesses and marketers who could benefit from learning more about this subject.