What Meta’s broad targeting teaches us about optimization signals

Over the last several years, marketers have largely accepted that broad targeting works on Meta’s platforms. They’ve also accepted that interest targeting and lookalikes matter less than they used to. Which means the audience-building tactics that performance marketers spent years perfecting matter less as Meta’s algorithms improve. Where does that leave us?

It leaves us with a problem: Many marketers took the wrong lesson from this evolution. Broad targeting doesn’t work because targeting no longer matters. Broad targeting works because Meta’s algorithm is so good at identifying future customers that it often knows more about who is likely to buy than the advertiser does.

It evaluates far more signals and behavioral patterns than advertisers have access to. We simply need to tell Meta what we care about. For most advertisers, that’s sales.

Applying that same thinking to other campaign types within Meta — and increasingly across other advertising platforms — can waste millions of dollars. In conversion campaigns, the optimization signal does the heavy lifting. Let’s take a deeper look at what I mean.

Signal optimization is the new targeting

For years, we’ve said, “Creative is the new targeting.” I still believe that’s true. Great creative naturally attracts the right audience and puts off the wrong audience. But a second shift gets far less attention than it should.

The conversion event you choose increasingly determines who Meta finds, how it spends your money, and ultimately, the business outcomes you generate. In many cases, the signal you’re optimizing toward matters more than the audience settings themselves. Nowhere is this more obvious than in traffic campaigns.

The broad-targeting lessons that work so well in conversion campaigns often fail spectacularly in traffic campaigns because the signal Meta receives is fundamentally different. When you optimize toward purchases in a conversion campaign, you’re effectively telling Meta, “Find me more people who look like the people who buy my products.” 

Every purchase improves the model. Every conversion teaches the algorithm something about the characteristics of a future customer. In that environment, broad targeting makes perfect sense because the optimization signal does most of the work.

Why traffic campaigns are different than conversion campaigns

Traffic campaigns differ from conversion campaigns on a fundamental level. When you optimize toward traffic, you’re no longer telling Meta to find future customers. You’re telling Meta to find people who click on ads. Those two groups aren’t the same.

Clicks are abundant, and Meta can generate them all day long. The challenge isn’t finding clicks. The challenge is finding clicks that ultimately lead to business outcomes. When you optimize toward traffic with broad targeting, you’re essentially removing both mechanisms that typically improve quality: audience constraints and strong optimization signals.

If you’ve ever wondered whether this is happening in your own account, there’s a simple way to check. Pull up the audience and placement breakdowns for your traffic campaigns and compare them to your conversion campaigns. Ask yourself:

• How much delivery goes to older demographics?
• How much delivery flows into Audience Network?
• How many “Amen” or “Praise Jesus” comments show up on your ads?
• How different is the distribution relative to campaigns optimized for purchases?

I’ve seen countless examples where traffic campaigns skew dramatically older than purchase campaigns. Did the ideal customer suddenly get 30 years older? Probably not. More likely, Meta found a segment of users who click on ads frequently: our parents and grandparents.

The same thing often happens with Audience Network, where traffic campaigns push a disproportionate amount of spend into placements that generate inexpensive clicks but very little downstream value. If Audience Network is still enabled in your traffic campaigns, I’d start by fixing that.

You get what you optimize for

Anthony Bourdain famously talked about ordering a steak well done. His argument: When you order a steak well done, you’re signaling to the restaurant that you don’t particularly care about the quality of the cut itself. At that point, you’re mostly evaluating whether you received a steak. The restaurant knows that and optimizes accordingly.

Traffic campaigns often work the same way. When you tell Meta that all you care about is traffic, you’re signaling that you don’t particularly care where that traffic comes from. You aren’t evaluating customer quality, purchases, or incrementality. You’re evaluating whether clicks show up in the reporting. Meta knows that, too. 

Just like a steakhouse happy to monetize the worst cuts of steak, Meta will happily monetize clicks that will never convert. It’s not doing anything wrong. It simply optimizes toward the objective you’ve given it.

The incrementality trap

Traffic campaigns get particularly risky when brands start talking about incrementality. Most growth conversations eventually arrive at the same conclusion: We need to reach more people who aren’t already buying from us. That’s a reasonable objective. The mistake is assuming the path to incremental customers is simply generating more traffic.

More traffic doesn’t automatically create more incremental sales. More traffic doesn’t automatically create more incremental customers. Remember, your team’s objective isn’t to maximize visits. The objective is to maximize the number of people who purchase who otherwise wouldn’t.

While that often means going beyond the audience Meta thinks will convert immediately, too many brands miss the right next step. The solution usually isn’t a traffic campaign, especially not a broad-targeted traffic campaign. A better approach often combines a conversion campaign with a different signal:

• Instead of optimizing toward traffic, can you optimize toward a PDP view?
• Instead of optimizing toward a click, can you optimize toward a collection-page engagement event?

This same strategy plays out even within traditional conversion campaigns because optimizing toward purchases doesn’t mean Meta will find incremental purchases. That’s what makes incremental attribution so interesting.

If you’re getting lower-than-ideal iROAS reads in incrementality testing on your conversion campaigns, consider optimizing toward new-customer purchases, high-LTV customers, or a specific category that matters most to the business instead of generic all purchases.

Each step gives Meta more information about what success actually looks like. Each step provides a stronger signal. As the signal improves, the need for prescriptive targeting declines.

The richer the signal, the less you need targeting

That’s the heuristic I increasingly use when evaluating paid digital strategies: The richer the signal, the less you need targeting. The weaker the signal, the more you need targeting.

Most marketers understand the first half of that equation because they’ve seen broad targeting outperform in conversion campaigns. Far fewer understand the second half, which is why so many traffic campaigns continue generating millions of clicks while creating surprisingly little business impact.

But the broader lesson extends well beyond traffic campaigns.

The conversion event you choose increasingly determines who the algorithm finds, how it allocates spend, and ultimately, the business outcomes it creates. Whether that’s a purchase, a new customer, a high-LTV customer, a PDP view, a specific product category, or another meaningful business outcome, the signal itself is one of the most important strategic decisions marketers make.

The best advertisers no longer just think about audiences and creative. They think deeply about the signals they feed the algorithm and whether those signals align with the outcomes they’re trying to create.