We’re moving into a new era of Meta advertising. Meta wants us to go broader with our targeting, and in some cases there is evidence that it’s effective. But there are times when broad targeting is bound to fail.
You may even get results that appear to be positive. But that’s part of the issue here. The performance fails once you scratch below the surface.
In this post, I want to highlight some specific situations when this is bound to happen. At the end, I’ll lay out a way that I’m trying to counter this. But ultimately, it would be helpful if Meta would come up with a way that wouldn’t require creative solutions.
What is Broad Targeting?
There’s often confusion regarding what I mean when broad targeting is discussed, so let’s define it.
When “broad targeting” is mentioned, I mean one of three things:
1. All targeting inputs are removed.
This could be because you’re using the original audience options and don’t provide anything for custom audiences or detailed targeting (only using geographic and demographic targeting). It could also be because you’re running an Advantage+ Shopping Campaign.
2. Targeting expansion is on.
You’re using the original audience options and provide custom audience, lookalike audience, or detailed targeting inputs, and you either manually turn on one of the Advantage audience expansion tools (Advantage Detailed Targeting, Advantage Lookalike, Advantage Custom Audience) or they are automatically turned on due to objective. As a result, your audience will be expanded to reach people beyond your inputs if it will lead to more or better results.
3. You’re using Advantage+ Audience.
This is now the default option when setting targeting in manual campaigns, though you can currently switch back to the original audience options. When using Advantage+ Audience, your targeting inputs (which are optional) are used only as targeting suggestions. The algorithm can then go much broader.
Optimization is Literal
This is going to come up repeatedly, so let’s address it now. It’s important that you understand that Meta ads optimization is literal. This is both a benefit and a weakness. Here’s what I mean…
Ad delivery is driven by your performance goal. The algorithm’s focus will be on getting you as many of the actions you want within your budget.
Meta will optimize and make adjustments (including who sees your ad, which will be important) based on satisfying that goal. Whether you want conversions, link clicks, impressions, or something else, Meta’s focus will be on helping you get as many of that thing as possible because that is how you’ve defined success.
That’s a benefit if all you really want is the thing that you’re optimizing for. It’s a weakness if you expect people who perform your optimization event to also perform other actions. You expect, for example, people who click a link to land on your website then behave like a normal human who may do other things.
But the algorithm doesn’t care about those other actions. It only cares about satisfying that initial goal.
When Broad Targeting Works
Broad targeting is most effective when both of the following are true:
1. Your performance goal is Conversions (number or value) and the conversion event is Purchases.
2. Your budget is high enough to generate the volume required for Meta to learn and properly optimize.
The reason this is the ideal situation is that, first, optimization is literal. Meta’s only focus is getting you purchases. When your audience is expanded, the algorithm won’t go after people likely to result in accidental or low-quality purchases. That’s just not a wide-spread and predictable issue.
Going broad is beneficial here because your results are often limited by your audience size. You add those limitations and they can restrict the algorithm from finding you more results. Additionally, that tighter audience can drive up frequency, creative fatigue, and ad costs.
A high enough budget to generate volume is helpful to get great results, but it’s not required. Let’s not get bogged down in what “high enough” means. The bottom line is that the algorithm can better learn when there’s volume of data to learn from.
Optimizing for purchases is when broad targeting is at its best. A notch below would be any other type of conversion. It’s not on par with purchases because the quality of that conversion can be an issue when the algorithm expands your audience and goes for anyone within the targeting pool.
That is also because optimization is literal. You want leads? The algorithm will get you leads. Unless you optimize for conversion leads, the algorithm does not care what those people do after subscribing.
This could potentially be controlled with tighter targeting. But once you allow Meta to target anyone beyond your target group, the characteristics of your ideal audience mean close to nothing.
When Broad Targeting Fails
Broad targeting can be an unmitigated disaster when optimizing for any top-of-the-funnel action. The problem is that, more than likely, you’ll be able to generate what appear to be great results. Meta will think they’re great results, too. But they’re likely low quality.
The reason, again, is that optimization is literal.
This is why most experienced advertisers would tell you it’s almost always a waste of money to optimize for link clicks or landing page views. The algorithm will do everything it can to get you those clicks at the lowest cost. And they could come from people who click everything (for no known reason), accidental clicks, and even bots or click fraud before they are detected.
Low-quality results are already an issue with top-of-the-funnel optimization prior to broad targeting. But if you can at least define your audience succinctly, you may be able to place some guardrails on the algorithm. You’d likely still run into quality concerns, but that will definitely be an issue if Meta can remove those guardrails and target whomever they want.
If you say you want link clicks and can’t place limited restrictions on whom you want to reach, Meta will find link clicks. You may luck out and get a few quality clicks, but most are likely to be a waste of money.
It seems as though Meta knows this because the only time detailed targeting and lookalike audiences are automatically expanded using the original audiences is when optimizing for conversions. You can turn this off when running top-of-the-funnel optimization.
Interesting, isn’t it?
An Approach to Solving Broad Targeting Issues
Let’s try to find a creative solution to this problem of quality when optimizing for top-of-the-funnel actions while going broad. Because if we did, the algorithm could actually benefit us by helping to find lower cost (and hopefully higher quality) actions.
Let’s use the example of link clicks and landing page views to drive traffic. We need an alternative when we have content that we want people to consume — and not another immediate bottom-of-the-funnel action would be expected.
This is my life, actually. My blog is important, and I do still want to drive traffic to it. We’ll want to optimize for some type of conversion, but a purchase or even lead won’t happen at a high enough rate as a direct result of reading my blog post to make that optimization realistic — without needing to spend quite a bit to get it.
You may already know that I have a large slate of custom events that fire on my website that represent some of the quality actions that I want.
Here are examples:
- Timer Events that fire at 15 seconds, 1 minute, 2 minutes, and 3 minutes
- Scroll Depth event that fires at 50% scroll
- Item in View event fires when someone views the comments
- Video Viewed event fires when someone plays an embedded YouTube video
- Click event fires when someone clicks my bot or a share button
- Internal Link Click event fires when someone clicks any link that takes them to another page of my website
These events are helpful for both optimization and reporting. To optimize for one of these events, we’ll need to create an Engagement campaign that uses the Website conversion location and Conversions performance goal.
But, which event should be the optimization event? This may not sound like an important question, but I’ve found some of the same issues with my custom events when it comes to the algorithm being literal. If I optimize for a timer event, I’ll end up with people spending lots of time on the page, but they never do anything else. If I optimize for scroll, they’ll scroll, but immediately abandon.
We also need to consider costs and volume. If it ends up costing $10 for one of these events as the central conversion event, I’ll need to spend $500 per week just to exit the learning phase for what is essentially a traffic campaign.
Here’s an example of what I’m experimenting with now, but I may still make adjustments…
I’m running ads that send people to one of my short-form video custom post types. Because of that, the YouTube video is embedded at the top of the post and there’s a short blog post below it.
For now, I’ve chosen to set the central conversion event as VideoWatched, which is the event that fires when someone starts the embedded YouTube video.
When optimizing for such an event that can happen over and over, a critical element is the Attribution Setting. Make sure it’s 1-day click only, otherwise the results will be inflated.
I’m using Advantage+ Audience, but with targeting suggestions of people who have fired the VideoWatched event during the past 180 days or those who are in the top 25% of time spent on my website. I’m hoping this initial group will provide the actions I want to give the algorithm something to learn from prior to going broader.
I’m also excluding anyone who viewed one of seven different posts for at least 15 seconds I’m going to promote in this campaign. This will also prevent unnecessary frequency.
I’ve created multiple ads for different video posts, hoping to give the algorithm something that will work. Since I already post Reels, people are accustomed to seeing these in the feed. So I use link ads that use the featured image of these videos, making it clear that they will be watching.
So far, the results have been solid. A reasonably high percentage of the people who click are spending at least 15 seconds (or a minute), scrolling, clicking internal links, and watching the embedded video. This isn’t shocking because I’ve experimented with this approach before. The caveat here is that we’re trying to make it work with broad targeting.
It’s early, but the Video Watched event volume is still the lowest of those custom events, so I may make an eventual adjustment and optimize for something else, like internal link clicks.
What Meta Could Do Instead
Coming up with creative solutions is fun, but it’s aggravating that this is necessary. Top-of-the-funnel optimization is already problematic, but if Meta’s going to encourage or virtually force broad targeting via Advantage+ Audience, it’s practically burning money.
Meta could fix this by solving the quality issue. The custom event approach may not be as useful as it could be because the learning is isolated to my website. Why can’t Meta create standard events that represent some of this quality engagement so that there is more data?
Meta could add a Quality element to Traffic campaigns (or other top-of-funnel objectives). Do you want the algorithm to focus on the most link clicks or landing page views? Or do you want the highest quality traffic (people who are more likely to spend more time, click around, and return later), knowing that it will cost more?
This has long been a complaint, but it becomes a bigger problem if you can’t put guardrails on targeting. Whether it’s traffic or some other type of engagement, the algorithm — given a huge pool to go after — will find weaknesses to get you the most results possible. And those results will often be low quality.
Watch Video
I also recorded a video to walk through this. Watch it below…
Your Turn
How do you approach broad targeting?
Let me know in the comments below!