Article

Meta Isn’t JUST About Creative

Advertising Analytics Data Marketing Meta

February 17, 2025

As pitch, conference (and webinar) season have all entered full swing, I’ve heard a steady drumbeat of claims along these lines:

“Meta is just about creative”
“Anyone can run meta ads”
“If you’re not running broad, you’re doing it wrong”
“Targeting doesn’t matter”
“With Meta AI, the ad does the targeting for you”

All uttered by smart, successful, and well-meaning people. And there is a lot of truth underneath the surface of these statements. Meta has removed troves of data over the past 5 years to comply with increasing regulation and to avoid lawsuits. Meta has aggressivelypushed their “power-five”, which includes broad targeting and auto-advanced matching – on advertisers. AI & ML have enabled Meta to understand more of an ad’s context and better predict the right audience for that creative.

To sum it up, I’d go as far as to say that these statements are 80% true. But 80% isn’t 100%.

And while I may be in the minority, I don’t believe media buying on Meta is dead – to the contrary, I believe it’s poised to enter a new golden age – one where channel targets can (finally) be married with dynamic, up-to-the-moment audience research that finally allows brands to access and target hyper-relevant audiences in the ways imagined by the (mostly just garden-variety fraud) Cambridge Analytica scandal.

Before we go on: no, I still think creative quality, diversity, velocity & volume are mission-critical to succeeding in Meta Ads. And I still believe that most brands aren’t making enough ads (or good enough ads) to truly scale their Meta ad accounts.

Both things can be true: (1) most marketers are overlooking the incrementality of targeting and media buying on Meta and (2) despite all of the talks and pixels spent proclaiming the importance of creative, most brands still don’t get it.

Let’s Talk Media Buying:

Way back when in the “golden era” of Facebook Ads there were (quite literally) thousands of stupidly-granular targeting options available in Meta, ranging from individual net worths to home values, salaries, types of investments held, detailed personal preferences, brand loyalties and more. That once-in-a-tech-cycle abundance of data made it not just possible, but easy for marketers to target hyper-specific, overly niche audiences.

But, as with everything, things changed. For many reasons, some of them legal, others economic – Meta gradually removed many of these targeting options and reduced the number of interests shown in the search tool down to 25 (and, typically, the largest 25 interms of audience size) for any particular query/branch.

The end result of this targeting “pruning” was a massive financial boon to Meta: larger, more inclusive audiences mean there is more competition for every audience – which, in turn, results in higher CPMs/CPCs, more data available for machine learning, and larger earnings per share.

It would be easy to say this was the end of advanced targeting + media buying on Meta – but that’s not the entire story. Many – not all – of those targeting signals are still available in Meta, albeit under different names and in slightly-less-than-obvious ways.

I’m an avid golfer. I spend a substantial amount of money on the sport every year, everything from new clubs to accessories, apparel, greens fees, etc. There are few things I’d rather do on a Friday afternoon or a weekend morning than play 18. Put simply: if you’re a brand that markets to golf enthusiasts, I’m your ideal customer.

That leads to a simple question: if you are that brand focused on golf enthusiasts, how would you go about targeting me on Meta?

When I ask this question at conferences or events or in sessions with other media buyers, the answers are (almost unfailingly): target a combination of generic topics (i.e. Golf), along with popular brands that have large golf footprints (like Nike, Puma, the PGA Tour, Adidas – then throw in a golf-specific brand or two (i.e. Titleist, Golf Digest, TopGolf or Callaway) and/or some famous golfers (i.e. Tiger, Phil, Rory, Bryson, etc.). Finally, if necessary, refine this audience by demographics (i.e. men under 40).

There’s nothing particularly wrong with this answer – it’s logical. It’s defensible. It’s prevalent in the marketing industry. And that’s exactly why it’s suboptimal. If everyone is using the same handful of targeting signals, they’re (essentially) worthless.

Let me illustrate.

Depending on how you structure your audience using the above methodology, the US-only audience you’ll get on Meta is ~15,000,000 to 25,000,000 people. If you are unfamiliar with golf, that probably seems like a decent-sized audience; and if you went the extra mile to do some research on the size of the US golfer population, you likely found a nice article from the National Golf Foundation that indicated ~23.6M people participated in a round of golf in the last year. Since 23.6M is right in the middle of the ~15M-25M audience size Meta has estimated, you probably patted yourself on the back and proceeded to make some ads and launch the campaign.

But here’s the problem with that approach: only about 16.1% of golfers spend more than $1,000 on the game each year, all-in (i.e. inclusive of greens fees, equipment, apparel, etc.), and only about 2.7% spend more than $2,000 on the game each year.

If you’re running a niche golf brand that caters to enthusiasts, the people you really want to find are those ~637,000 ($2,000+ crowd) to 3,900,000 people ($1,000+ crowd). Advertising to the larger group above is likely counter-productive — 5 out of every 6 people you’ll reach (in the wildly optimistic scenario that the true golf fanatics actually have those mass-market interests) just aren’t that into golf.

End result: interest stack underperforms your expectations, while some of your broader targeting overperforms. This, in turn, adds fuel to the narrative fire around media buying being “dead” on Meta – after all, you did the right research (allegedly), built out smart audiences (again, allegedly), and it still under-performed basic, broad targeting.

Not so fast.

What really happened is that Meta shoe-horned your targeting into the largest relevant buckets of people and gave you the illusion that the “right people” and the “big buckets” were the same thing. You fell for it, hook-line-sinker. Your interest stack ad set sucked, Meta said “go broad” and you did. Winner, winner Meta stockholder dinner.

But what if there was a way that you could build a hyper-relevant interest stack audience that actually reached a disproportionately large segment (not all, but most) of those 637,000+ big spenders?

Social Media Buying 2.0

There are three secrets to doing exactly that in 2025:

Secret #1: Connect Audience Research To Platform Signals

I’m going to be blunt: this is something virtually no-one does – not because it’s overly complex, but because it’s frustrating, tedious and runs headlong into Meta’s capriciousness (AKA Meta’s tendency to add/remove targeting signals on a Wednesday afternoon whim).

Most brands do basic audience research, then hand that document to their media buyer and say – essentially – “good luck!” In the most optimistic scenario, the media buyer takes it and does (basically) what I outlined above. What you need is research that connects people to platforms.

To do this, I start with SparkToro –

I’ve built a simple search here for a “golf persona” – people who have “golfer” in their bio, visit golf digest or watch PGA tour clips (it’s not perfect):

What you can see – from the checked boxes – are a bunch of brands/interest categories that are out of the mainstream. I then plugged each of those individual sites back into SparkToro to see which one(s) were consistently frequented by a golf-obsessed audience. From this, I started to create a list of the brands that are most central to the golf enthusiast network (nerdy aside: this is a rudimentary version of the same theory that Google uses to rank pages based on outside link profiles):

But just having a list of things golf enthusiasts love isn’t – in and of itself – a Meta Ads targeting strategy. For that, we need another step: the Meta Ads API. If you’re not comfortable working on APIs, there’s a free tool that’s 90% as good: Interest Explorer.

Type in the broad interest you’re looking to explore (i.e. golf), and voila! You’ll end up with 100-200 interests from the Meta API related to “golf” – and all you need to do is find the niche/non-obvious brands/people/places that appear on both your SparkToro list and your Interest Explorer (or Meta Ads API) list.

As a concrete example, here’s some examples of brands I found using the above exercise:

Scotty Cameron
Ping Golf
Titlelist Golf
Bridgestone Golf
Mizuno Corporation
Callaway Golf
Cleveland Golf
Cobra Golf
Dave Pelz (Golf coach)

To an avid golfer, each of these brands/people are eminently familiar. To a non-golfer they might as well be Santa Claus.

When I put those into ChatGPT and asked it to generate an Interest Stack for my hypothetical golf brand, it suggested three sets of “AND” inclusions, and even recommended a few more to add. I dropped these into my demo account, and voila:

A 1.6M-1.8M audience size, based exclusively on these niche targets. That’s far closer to our ideal audience (637,000) than the broad method, but still a ways away. The challenge now: there are plenty of people who are very interested in and/or passionate about golf, but may not have the means to spend $2,000+ per year on it.

Secret #2: Find Acceptable Proxies or Soft Signals

Now, back in the good old days, Meta outright put salaries/net worths/home values into Ads Manager as targeting signals. Those days are, for better or worse, no longer. Instead, we now have to deal with either (a) the amorphous “household income: top XX% of ZIP codes (US)” or (b) infer income/wealth from another interest signal.

Since the “household income” targeting method has nothing to do with the individual income – it’s just people who live in zip codes with the highest average household income, we’re going to go with option b.

There are plenty of ways to accomplish this for wealth/affluence/disposable income, but my favorites:

Identify more generic “silent” signals of wealth/affluence: people who have the income to drop $2,000+ per year on a hobby also likely are fans of nice things – think: luxury cars that are a tier below supercar (BMW, Mercedes, Porsche, Audi, Tesla) and luxury hotels (Ritz-Carlton, Four Seasons, St. Regis, Fairmont, The Luxury Collection, Rosewood Hotels, Mandarin Oriental Hotels, The Waldorf Astoria Hotels & Resorts,the Dorchester Collection, Aman Resorts, etc.).
Target Golf-Specific Affluence Factors: The single-easiest of these is to target people who are members of country clubs or private golf clubs – if someone is willing to pay upwards of $10,000 just to join a club, odds are they have the financial means and the willingness to invest in golf going forward. The other option is to target people who are interested in golf travel & vacations (i.e. Myrtle Beach, Scottsdale, Pebble Beach, Northern Ireland, Kiawah Island, St. Andrews, Bandon Dunes, Pinehurst, Hilton Head Island, Palm Beach, etc.)
Target Tangential Investments in Golf: A serious golfer will often invest in technology and accessories to support their hobby: range finders, simulators, swing aides, etc. Target the manufacturers of these tools (TrackMan, Shot Scope, Garmin, Arccos, etc.).
Other Wealth & Lifestyle Habits: Speaking generally, the types of people who are willing to spend thousands of dollars on a hobby are likely willing to make comparable investments in other aspects of their life – food, travel, etc. Fortunately, there are ways to target these people as well: Fine Dining & Wine Enthusiasts, First/Business Class travel, European Vacations, Investment Interests (Investing, Private Banking, Wealth Management), etc.

Identifying the right soft signals does take effort and research – but you can use the same process as I’ve outlined above. Search a few, find the commonalities, rinse, repeat. The good news for things like wealth is that the signals are consistent across audiences – an affluent yogi likely has many of the same “soft signals” as a wealthy golfer (nice car, luxury vacations/travel, interest in nice food).

This same methodology can be applied to dozens of other categories – from people looking for self-help (i.e. target Brene Brown, Elizabeth Gilbert, Byron Katie, Kwik Method, Simon Sinek, Self-Awareness, Self Love, Louise Hay, etc.) to fitness/health (target CrossFit, Yoga, Weight Training, Bodybuilding, Physical Exercise, Weight Training, Powerlifting, Strength Training, Barbells, Kettlebells, etc.) to early adopters of technology (i.e. smart device, smart technology, emerging technologies, etc.).

If you’re looking for high-earners, pull a list of the highest paid titles/jobs, then find them on Meta. If you want people interested in construction/woodworking, look at pictures of job sites, then target the niche brands you see (or just ask ChatGPT).

With the benefit of hindsight, many of these seem obvious; the problem is that we (as marketers) have not done a good enough job of inferring targeting signals from high probability interests/habits – which has led to us targeting the obvious instead of honing in on the hidden gems.

In fact, when I added those “wealth/affluence” signals to my golf audience, look what happened:

421,300 to 495,600 people – just slightly smaller than the 637,000 target we estimated from the National Golf Foundation’s dataset. That’s ideal. It’s likely hyper-relevant.

Secret #3: Evolve Rapidly

Finally, there’s a nasty habit among marketers that we assume all silent signals/habits/affinities are like wealth: relatively stable. Generally speaking, you could show a marketer from 1980 the brands I listed above (BMW, Mercedes, Audi, Porsche, Maserati, Lexus, Range/Land Rover, etc.) and they’d tell you these are luxury vehicles, likely driven by wealthy and/or financially successful people. They’d be right, in this context.

But if you showed them Scotty Cameron, Bridgestone, Miura or Travis Mathew? They likely would NOT associate those with affluent golfers. Why? Those brands are relatively new and/or niche.

This is where having on-demand, on-command access to user behavior and trends via platforms like SparkToro is wildly helpful: it makes it easy to spot changes, emerging brands, new search terminology, etc. And once you identify those emerging trends/topics/terms, you can get ahead of them by updating your personas/targeting and adding them to your campaigns.

People aren’t static. Your audiences + audience research shouldn’t be, either.

Audience research isn’t one-and-done, so why are your Meta Ads interest stacks set-and-forget for years on end? They should evolve and adapt to the changing preferences of your target audience, else you’ll end up left far, far behind.

The Impact of AI

One of the primary reasons I’m so bullish on the future of media buying is because tools like ChatGPT and Gemini provide the computational horsepower to do this work on an unprecedented scale.

The reason most marketers never did it before wasn’t because they weren’t able, or didn’t think it could be helpful – it’s because they had too many accounts and not enough time. But with AI? Suddenly doing this becomes not just possible, but required. We’re going to see a new golden age of media buying, driven by AI/LLM tools synthesizing vast amounts of user data and audience research, then generating hyper-relevant interest stacks.

And, from a creative perspective, this is going to be absolutely wonderful – because if you don’t have to worry about the creative doing the “targeting”, we can all get back to making creative that appeals to the target audience.

Great creative must appeal to the target audience first, and serve the algorithm second. People make you money; platforms take your money.

Until next week,

Sam

Related Insights