RevelOne’s Spotlight Series regularly features insights from top experts in our Interim Expert Network. We cover a broad range of topics at the intersection of marketing, growth, and talent. If you’re interested in exploring these topics further and engaging with one of our 200+ executive or mid-level experts, please contact our team at experts@revel-one.com.

Written by Amanda Berg
This article was originally published in Amanda’s email newsletter, Growth Therapy.  

If you asked any paid media buyer how to run Facebook ads in 2017, they’d give you a completely different answer than if you asked that same question today (and if they don’t, run!).

Facebook used to be THE channel for ultra-precise, narrowed-in targeting, which, of course, made it an incredibly valuable tool for advertisers looking to reach highly relevant audiences.

But a lot has changed in the last ~7 years. Aside from the fact that we call Facebook “Meta” now, precise, advertiser-controlled audience targeting is no longer a key feature of the platform. Like most of the big paid digital channels, Meta has removed more and more advertiser control over time, relying more on its modeling and algorithms to serve ads to the right people at the right time.

AI and ML advancements aside, there were a few key turning points that led to sweeping changes in how advertising on Facebook works.

Let’s dive into how it used to work, what these turning points were, and how it works now, so you can leverage Meta properly — and not like it’s 2017.

How Facebook Targeting Used to Work

Back in the day, you could target your Facebook ads to virtually any granular subset of people you could dream up, based on demographic info, interests, and behaviors. This included:

• Precise demographic info, including race, religion, and age as low as 13
• Household income bands
• Relationship statuses
• Subsets of parents based on the age of their kids
• Folks experiencing particular life events like anniversaries or birthdays
• Political leanings
• Education level
• Employers and job titles
• Virtually any interest or behavior you could think of, like: affinity for specific brands, activities, sports, public figures; or commuters, soccer fans, travelers, purchase behaviors, device users

See this (massive) graphic from 2018 that details all the possible targeting options on Facebook. Keep in mind that each of these interest categories contains thousands of individual interests.

This level of targeting was incredibly useful to all advertisers and made it easier than ever to reach new and high-intent users at a massive scale.

For example, think about how a nascent clean beauty brand, political campaign team, or consumer fintech company could all use the same tools on Facebook to achieve massive reach to their precise audience.

But if you’re thinking to yourself, “Wow, it must have been easy for Facebook and its advertisers to abuse all this information,” you’re already ahead of where Facebook was (publicly, at least) at the time.

The Audience Insights Tool

If you remember the old audience insights tool, you might not even know there’s still a version of it currently available because it’s so watered down.

The old version of Facebook’s audience insights tool was an absolute goldmine. Facebook used to let you pop in ANY targeting inputs you’d like, and would spit out the demographics and interests of those folks so that you could:

• Learn all about the people in your audience
• Target your audience’s interests — critically, interests you may not even have thought of (This strategy was pivotal in scaling Facebook while I was leading growth at SmartAsset, and got us spending ~$1M a month on Facebook ads during my time there).

Luckily, I was able to find a screenshot of what this tool used to look like (apologies for the blurry image). You entered your inputs into the gray menu on the left, and could then learn all about this audience on the right:

The real golden nuggets came from the “Page Likes” tab, where you could see a list of pages liked by your audience. You could then target these as interests.

There’s still a version of this tool that exists, but it’s so watered down that I don’t see how it’s useful. The inputs are much more limited, to age, gender, location, and a limited set of interests.

When I tested it, the outputs were virtually the same across the various demographics I tried, and were incredibly generic, as you can see in the list below.

This is the “top pages” list for my demographic, Women 26 – 34 in Colorado. Aside from the Denver Broncos (and the total wildcard that is Eminem 🍝), the same “interests” showed up in virtually every demographic breakout I tried.

Custom Audiences and Lookalikes

Custom and lookalike audience lists have also gone through a lot of change, though not nearly as visibly.

They certainly still exist and are widely used, but they’re not as powerful as they used to be, particularly for smaller brands with less reach. To understand why that is, it’s important to note the two most common ways of building a custom audience:

• Uploading a customer list of emails, phone numbers, or some other identifier that you already have from users who have made a purchase or used your service
• Using your Facebook pixel to define a set of users based on an action they took on your site

Once you build a custom audience this way, you can then tell Facebook to make you a net new audience of people that “look like” the folks in your seed list. Facebook’s ability to do this accurately relies on being able to take your seed list and match it up with folks on Facebook.

Match rates used to sit in the 60% – 80% range. The missing 20% – 40% wasn’t a deficiency in Facebook’s ability to understand your data — more often than not, you just had folks in your list that either didn’t use Facebook or used Facebook with different contact info.

You can see why having a large audience is necessary to inform Facebook’s ability to model, even when match rates used to be this high. Today, match rates are much lower. I’ll explain why.

The Two Major Turning Points

Two key events over the last 10 years meaningfully changed the way Facebook functions as an ad platform. The first takes us back to 2014…

1. The Cambridge Analytica Scandal

This story made global news in 2018, so you may remember pieces of it. Here’s the TL;DR:

Back in 2014, a developer and academic named Aleksandr Kogan released a Facebook app called “thisisyourdigitalife” and billed it as a personality test. Around 270,000 Facebook users downloaded the app and consented to share their data with the app itself.  Behind the scenes, the developer was sharing all this data with political consulting firm Cambridge Analytica — without users’ consent — a violation of Facebook’s policy.  Further, Facebook’s API at the time allowed developers to access information about people’s Facebook friends, so the data of around 50 million people, unbeknownst to them, was ultimately shared with Cambridge Analytica.

Cambridge Analytica then used this data to build psychographic profiles of users that could be used to target political ads with extreme precision. A whistleblower finally leaked the story, and Facebook had to pay the price by significantly tightening up its rules around:

• What customer data could be shared, and with whom
• The types of targeting they allowed advertisers to access

We started seeing many granular interest and behavior-targeting segments disappear from Facebook as a direct result. To this day Facebook still occasionally removes targeting options, often blaming the fact that they’re “duplicative” or “rarely used.” You certainly can’t find anything related to race, religion, or political affiliation anymore, and tons of interest segments have been removed over time.

2. Apple’s iOS 14.5 Rollout

The second key turning point made big headlines in the digital marketing and tech circles but may have flown under the radar for Gen Pop. This was Apple’s rollout of iOS 14.5 in the spring of 2021.

The key change in this OS update was that all apps would be required to show a one-time popup to all users, asking them to opt into being tracked by that app. If a user opted out, they could not be asked again to opt in.

This was annoying for app developers, but not the end of the world for most apps. And it was such a non-event for typical consumers that you probably don’t even remember this happening. Meta, however, was uniquely positioned to lose out big time — Facebook and Instagram are both mobile apps AND massive advertising platforms that rely heavily on tracking user behavior to properly inform their ad-serving algorithms.

It’s estimated that as many as 90% of people opted out of being tracked — a massive loss for Meta.

In other words, up to 90% of the portion of your audience that uses an Apple device is no longer trackable or able to be used in custom or lookalike audiences. If your audience happens to skew heavily Android or desktop, this has less of an impact on you.

Let’s bring this back to custom audience match rates. As a result of users being given the power to opt out of being tracked by Facebook and Instagram, match rates have sharply declined, and typically come in around 25% – 40% these days. This means Facebook has far fewer data points to power its models, and means it’s especially challenging for smaller or newer orgs with less data to leverage these tools.

This opt-out behavior also impacts the interest and behavior targeting options that remain, accurate conversion tracking, and anything else on Meta that relies on engagement and conversion signal — think budget optimization and creative optimization.

The Key Changes

Of course, it’s good for consumers not to have their data shared or tracked without their consent. Ultimately, the two major impacts on advertisers from these events were:

• Reduced availability of granular demographic, interest, and behavior-targeting options
• Weakened effectiveness from using your first-party data

How Meta Targeting Works Now

For better or worse, Meta is heavily invested in AI and machine learning, which are more crucial than ever in both measuring performance and powering its ad-serving algorithm. Since many granular targeting options have been done away with, Meta now encourages advertisers to use what they call “Advantage+ targeting”, a silly name that means you’re letting Meta decide for you who to show ads to, and when and where to place them.

One huge misconception here is that by going “broad” this way, your targeting is fully open. Unless you’re optimizing purely for reach, it’s virtually impossible to have “open targeting” on Meta, because their algorithms have gotten so good at honing in on who to serve ads to. Advertisers just don’t have as much visibility into who their ads are being served to as they used to, back when they could make more of these decisions themselves.

It’s also more difficult now to achieve ultra-precise geographical targeting. According to Meta’s documentation:

“Meta technologies use a variety of signals to show ads to the people who are a part of your location targeting selections. Because these signals vary, complete accuracy cannot be guaranteed. You may occasionally see a small number of ad impressions, or even receive a message or lead, from outside of your location settings.”

Finally, when using custom or lookalike audiences, know that bigger is better. Consensus on exact numbers varies, but I like to use 10,000 matched profiles as an absolute minimum — meaning you probably need somewhere between 20,000 and 30,000 profiles in your list, depending on what your match rate looks like.

When you bear in mind that a 1% lookalike audience in the U.S. contains somewhere between 2 and 3 million people, basing that list off of just 10,000 profiles means Meta still has to fill in a lot of blanks.

These numbers probably sound like small potatoes for medium and large brands, but this can be difficult to achieve for newer, smaller, or budget-constrained brands.

Does This Mean Meta Ads Don’t Work As Well Anymore?

I would say different, not worse. Meta is still a critical channel for B2C brands that are focused on acquisition, and it’s an incredibly powerful platform, in part because of how it’s needed to adapt over time and leverage machine learning.

It’s certainly gotten harder for very small or very local brands, because of all of the changes we discussed. CPMs have also gotten a lot more expensive over time as competition has increased and advertiser controls have lessened.

How to Use the “New” Meta Ads to Your Advantage

For starters, you don’t have to worry about as many campaign settings anymore. Unless your targeting needs to be restrictive, you should minimize your targeting parameters and let Meta figure most of it out for you. Its machine-learning algorithm is just going to be better than a human’s assumptions at determining what audience you should reach, especially with so many interests and other targeting options removed.

This isn’t to say you don’t need to understand who your audience is because Meta will understand it for you — quite the opposite.

Secondly, and I’ve talked about this before, your creative is now your targeting. It’s now the job of your creative assets to grab the attention of the people you want to reach with the right visuals and messaging to do so. A deep understanding of your customers and what makes them tick is essential to doing this well.

Invest in getting to know your customers — get on the phone with them, read the reviews they’re leaving, put thought into your post-purchase surveys, etc.

About the Author
Amanda Berg is a fractional head of growth and writer of the Growth Therapy Newsletter. She works with early-stage B2C startups to strategically launch, hone, and scale their marketing functions to effectively drive key outcomes. Previously, she led growth teams in-house at companies like SmartAsset and Zola. She’s based in Denver, CO.

About RevelOne
RevelOne is a leading go-to-market advisory and recruiting firm. We help hundreds of VC/PE-backed companies each year leverage the right resources to achieve more profitable growth. We do 250+ retained searches a year in Marketing and Sales roles from C-level on down for some of the most recognized names in tech. In addition to our Search Practice, our Interim Expert Network includes 200+ vetted expert contractors – executive-level leaders and head-of/director-level functional experts – available for interim or fractional engagements. For help in any of these areas, contact us.