Complaints about Meta’s algorithmic targeting are mostly misguided. Meta’s ability to find the people who are most willing to perform your desired action is extremely effective. But there is a fatal flaw that impacts optimization for any event that isn’t a purchase event.
Before you come at me about the issues with algorithmic targeting, I get it. I say that it’s “effective” because it’s efficient at doing exactly what it’s supposed to do. The flaw prevents it from being far more valuable.
Some advertisers will spend without seeing it. They see the results and don’t ask questions. Others will reject algorithmic targeting entirely without understanding why they aren’t getting the results that they desire.
There is a problem that is frustratingly difficult, if not impossible, for advertisers to solve. It’s Meta’s problem to fix, and I’ve been complaining about it for years.
I know, I’m being cryptic. It’s not easy to explain in an opening paragraph.
Let’s back up…
Who Sees Your Ads?
First, it’s important to understand that the definition of “targeting” has changed. I’d say that this evolution is part of what confuses advertisers. We don’t know how to communicate what “this” is now.
Not long ago, I asserted that targeting was the most critical factor to the success of your ads. Good ad copy and creative couldn’t recover from a bad targeting pool.
Of course, our inputs are only kinda sorta considered now when it comes to the audience that sees our ads.
1. Advantage+ Shopping Campaigns allow for virtually no targeting inputs at all. No detailed targeting, lookalike audiences, custom audiences, or much of anything.
2. Advantage+ Audience is the default option for defining your audience now. You can define a few things like location and age minimum, but your inputs are otherwise seen as suggestions (and it’s questionable how much they’re considered at all).
3. Original Audiences tend to be the fall-back for advertisers who want to retain targeting control. But, most don’t realize that their audience is usually expanded, especially when optimizing for conversions.
The primary lever that controls who sees your ads is the performance goal.
If you’re able to strictly define your audience (which is rare), Meta will find the people within that audience who are most likely to perform the action that you want.
If your inputs are seen as suggestions, your audience is expanded, or you don’t provide any inputs at all beyond the basics, Meta will find those people within the largest pools of people.
Is this targeting? Not really. It’s providing some initial suggestions and constraints and defining what you want so that Meta can find the people who will lead to results.
Like I said at the top, Meta is actually very good at this. Fewer constraints will almost always lead to more and cheaper results. But, that’s not without some problems.
When Optimization is Most Effective
Meta is best at generating high-quality results with minimal guidance when you are able to clearly articulate what you want. There are three primary examples of this…
1. Maximize Conversions (Purchase Event).
This requires that you’ve set up conversion event tracking and have defined purchase events. Meta will focus on getting you the most purchases within your budget.
2. Maximize Value (Purchase Event).
This requires that you pass value with your purchase events and have a variety of purchase prices. You may get less volume of purchases in this case, but Meta will focus on generating the highest Return on Ad Spend.
3. Maximize Conversion Leads.
Conversion Leads optimization is possible when using instant forms and requires several months of setup to define your funnel. Meta will then optimize to show your ads to people who will most likely become high-quality leads.
It doesn’t mean that you’re guaranteed to get great results when using any of these three approaches (so many factors contribute to that). But these are the times when you and Meta are on the same page regarding what you want.
Where Optimization Struggles
The reason the above three approaches to optimization work is that there is agreement over what a quality result looks like. You’ve defined that you want more purchases, more value, or more conversion leads, and Meta will focus on getting you those things. If successful, there shouldn’t be a dispute about the quality of those results.
Where this goes wrong is when using virtually any other performance goal. It includes some performance goals that are notorious for quality issues:
- Link Clicks
- Landing Page Views
- ThruPlays
- Post Engagement
But it can also include conversions that don’t result in a purchase. If you choose the performance goal to maximize conversions and select Lead or Website Registration as your conversion event, you likely run into a regular battle.
In all of these cases, you’ve only begun to define what you want. But you and Meta aren’t going to be on the same page.
If you choose to maximize link clicks or landing page views, Meta will focus on getting you as many link clicks or landing page views as possible. But you want quality traffic, not just any traffic.
If you maximize ThruPlays, Meta will show your ads to people most likely to watch at least 15 seconds of your video. But, that’s going to include people who are forced to watch your video. You want quality views of people who choose to watch, not just any views.
If you maximize conversions where the focus is on leads, Meta will try to get you as many leads as possible. But you want quality leads who are likely to buy from you, not just any leads.
In each case, Meta doesn’t care at all about quality. The algorithm’s only focus is on getting you as many of the action that you said you want.
This has always been an issue. But it’s less of an issue when you can tightly define your audience. When you can’t, Meta has fewer constraints to find results — and the likelihood for quality issues increases.
Exploited Weaknesses
This is the perfect storm for quality issues.
- An inability to strictly define your audience.
- An inability to define a quality action.
- Weaknesses that can help Meta generate a high volume of the actions that you want
Understand that Meta’s delivery algorithm knows where to look to find the action that you want. This isn’t always good.
This can be as simple as going after people who are likely to act because they’ve visited your website or engaged with your ads. It can also be going after people who have engaged with similar products or businesses.
But, it can also be due to weaknesses that are exploited to get you more results.
1. Placements.
If you choose a performance goal to maximize link clicks or landing page views, expect that a large percentage of your impressions will be focused on Audience Network. Meta knows that it can get clicks there. It’s not clear whether these are from accidental clicks, bots, or click farms (before they’re detected), but you can bet you’ll get lots of low-quality clicks.
If you choose to maximize ThruPlays, a large percentage of your impressions will go to placements where people are forced to watch at least 15 seconds of your video. Audience Network Rewarded Video, which incentivizes people to watch videos in exchange for virtual currency or something else of value, is notorious for this. I’ve had cases where I’ve had more ThruPlays than people reached for this reason.
2. Countries.
If you target multiple countries at once and there’s an imbalance of cost to reach people in those countries, you may then see an imbalance in distribution. Especially if you choose to maximize top-of-the-funnel actions, Meta will try to get you the most actions possible within your budget. While this doesn’t guarantee lower quality results, it can be a contributing factor — particularly when a country is known for bots and low-quality accounts.
3. Ages.
If you aren’t able to restrict by age, this can be a weakness that will be tapped to generate more results. I can only speak from personal experience on this, but it seems that older people are much more likely to click on and engage with ads. But that doesn’t mean that they are a likely customer. If you are generating a high number of low-quality leads, it’s possible that Meta is focusing impressions on older people because it’s leading to more results.
4. Genders.
Let’s say that your business caters to women. In theory, you may not need to limit your audience when maximizing conversions when the conversion event is a purchase. The algorithm will try to get you more purchases and should adjust when men don’t buy.
But that’s not the case if you optimize for link clicks, landing page views, post engagement, or ThruPlays. Even though they may not be your target customer, men may engage at a high rate. And that will lead to low-quality results.
5. Low-Quality Accounts.
This is a big bucket that includes bots (before they’re detected), spam accounts, and real people who want to click on everything. If they perform the action that you’ve defined in your performance goal, these are going to be some of the primary people who see your ads. They’ll get you a bunch of cheap results, but that doesn’t mean those results are the quality that you desire.
NOTE: These five weaknesses aren’t nearly as big of an issue when optimizing for conversions when your conversion event is a purchase. The reason is that if it doesn’t lead to the action that you want (a purchase), the algorithm adjusts. But this is why these weaknesses are so problematic for any other performance goal.
Age and Gender and Advantage+ Audience
One of the primary complaints about Advantage+ Audience is that age maximum and gender aren’t audience controls. You can provide an age maximum and gender, but they are only audience suggestions.
Once again, this should not be a big deal if you can accurately define the action that you want, like a purchase. But it otherwise has the potential to make Advantage+ Audience unusable when using any other performance goal.
Earlier, I mentioned having this challenge with leads. It’s not always a problem, but I’ve found that when I begin to get “surprisingly good results,” it’s usually because a high percentage of my budget is getting spent on an older audience.
There’s unfortunately no easy way around it. I’ve tried an age maximum suggestion, but Meta immediately ignores it because I can get more of the results I “want” by reaching an older audience. You can switch to original audiences and define the age maximum, but that’s not necessarily a great solution either. I don’t necessarily want to cut off all ad spend to an older audience. I just don’t want it to monopolize my budget.
The Fatal Flaw
The fatal flaw in Meta ads targeting and optimization is that, except in rare cases, Meta doesn’t know what we want. We’ve defined what we want in very general terms (link clicks, landing page views, leads, ThruPlays, etc.).
It’s the combination of this weakness in optimization and the growing reliance on algorithmic targeting that makes the problem worse. Meta’s systems are powerfully good at finding people who are willing to perform the action that you want.
Unfortunately, the action that “you want” isn’t necessarily exactly what you’ve defined with the performance goal. And that’s what leads to low-quality results and wasted ad spend.
The Solution: It’s Complicated
To a point, it’s simple. We don’t necessarily need more targeting control. It shouldn’t be necessary to require the ability to restrict by age or gender. The solution also isn’t to eliminate Advantage+ Audience or audience expansion through the various Advantage Audience tools.
The solution hasn’t changed since I first complained about it years ago: We need to be able to more precisely define what we want.
Instead of any old traffic, we want people who are going to spend time on our website, perform several actions, and make return visits.
Instead of any views of our videos, we want people who signal interest (willingly watch without being forced, search out more videos, and provide other engagement).
Instead of any leads, we want people who perform other actions that prove that they are quality leads — even if it’s not an eventual purchase.
I’m not sure how exactly Meta would implement this. It could be by providing a secondary performance goal. Or maybe it would be providing options of “volume” and “quality” actions where other factors are considered.
But the current flaws in optimization are old and primitive. Not only were they unacceptable years ago, they enhance the problem with the development of algorithmic targeting.
This needs to be fixed.
Your Turn
What are your thoughts?
Let me know in the comments below!
