Article

7 Keys To Making Meta Ads Work In 2024

Advertising Analytics Data Meta

October 21, 2024

I hope you’re all doing well, enjoying the fall weather and getting ready to close the year out strong. This past week kicked off the conference season, with events in Las Vegas (SMASH) and Chicago (ShopTalk), then off to London tomorrow for the just-launched SMX London event.

Across all of those travels, I found myself talking a lot – far more than I’m accustomed – about Meta Ads: what’s working, what’s not, and how to structure accounts for successful outcomes.

With that, let’s get into it:

#1: The Basics

Meta Ads, like all marketing, is a tool – not a magic wand. The best analogy I have is that they are a catalyst – if you have a wonderful product/service, offered at a fair price for the value provided, with a compelling story (preferably one that creates an emotional connection), Meta Ads can work fantastically well.

But if you have a non-differentiated, non-distinct, low-value, overpriced product – the best Meta Ads creatives and the most sophisticated media buyers will not transform your account into a money-printing machine. It is simply not going to happen. For a catalyst to be effective, the reagents need to be present.

#2: Your Data

While there are arguments if data is sand, oil, water or gold – the one thing every marketer has agreed on is that it’s valuable. If you want Meta Ads to be successful, you need three types of superb, reliable data:

Conversion Data: This comes in a few forms – purchase data (if you’re an eCommerce brand, retailer, SaaS platform or other B2C brand) or qualified lead data (for both B2B and some B2C brands). That data needs to be connected to your Meta Ads account (Conversions API) and it needs to be used as the primary optimization event.
Audience Data: Audience data is insanely important – not only does this reduce incidental delivery to existing customers (who likely aren’t incremental), but it also unlocks a host of targeting tactics (more on this below) that can dramatically improve the performance of your account. The ideal setup is to directly integrate your CRM or email platform with Meta Ads, such that as users move through the customer journey, convert, churn, etc., they are automatically added to the correct Meta Ads list (and removed from the incorrect lists/audiences).
Product/Brand/Business Data: The majority of brands get (at least) portions of #1 and #2 correct – but often neglect #3. This encompasses everything from optimizing your product feed to ensuring that the correct COGS + contribution margin targets are associated with each product/service. While Meta (and Google) both have truly remarkable amounts of data on your industry, vertical, and customers, they don’t neither platform knows your business the way you do. The more smart, relevant data you can share with Meta, the more Meta can help you acquire the right kinds of customers.

PS. If you’re one of those people who is terrified of telling Meta how much your margin is on a given product, or doesn’t want to give away your customer data, or doesn’t want to leverage non-obvious insights (i.e. different geos perform better), just remember that Meta is a trillion-dollar company with effectively-endless resources. If they wanted to enter your market, they’d just do it.

Put simply: if your data sucks, your account is going to suck, too.

#3: Optimization Events

After looking at (quite literally) hundreds of Meta Ads accounts, there’s one commonality among the under-performers: a failure to use the proper optimization event. For all of its flaws (and there are plenty), Meta personifies the third of the (alleged) Chinese curses: “May the gods grant everything you ask.”

If you ask Meta for link clicks, you’ll get link clicks – but the quality of those clicks might not be what you’d hoped or expected. Likewise, if you just ask for total sales, you’ll get sales – it just might be sales on your loss-leader SKU.

Equally important to the success of your account (though often overlooked) is the attribution window. By default, Meta operates on a 7 Day Click / 1 Day View (7DC / 1DV) attribution setting, which allows it to take credit for anywhere from 30% to 95% of the conversions you were going to get anyway – all by showing your ad to people who had added products to their cart already, started (but didn’t complete) a form, etc.

The easiest solution to this: remove the 1DV. This forces Meta to drive users to your site / landing page in order to take credit for the sale. Your in-platform ROAS will look materially worse, but your P&L and bank statement will look materially better.

#4: Structure

There are no hard-and-fast rules to Meta Account structures, but there are three (general) types I like to start with in an account:

Audience/Offer/Angle Specific Campaigns – for most brands, your go-to-market approach will vary based on your audience. My standard approach- budget withstanding – is to separate out each core audience/offer into its own campaign, with both prospecting + remarketing running side-by-side.

In a simple example, a financial advisor is very likely to have a radically different message to seniors (likely living on a fixed income w/ substantial assets, focused on preservation + legacy) vs. younger homeowners (likely earning more, but with limited assets, focused on building wealth and saving for other goals). Those audiences, in turn, are very likely to have different approaches to pursuing their objectives, different online behaviors, different interests, different news sources, etc. Segregating these two audiences into different campaigns provides the ability to target each one independently, while preventing signal pollution (i.e. Meta getting “confused” because two radically different types of users are both clicking on the same ad for very different reasons).

The added benefit of this setup is downside protection – anyone who has run Meta Ads long enough has come across the black swan situation where a previously high-performing campaign simply stops spending, often without any trigger, rhyme or reason. If your entire Meta strategy is contingent on that single campaign, you’re in a world of hurt. Conversely, if you have 3, 4, 5 campaigns running (each to a different audience or with a different offer/angle), and one falls off, you have the ability to continue running the other 2-4 campaigns while you attempt to address the collapsed campaign.
Cost Controls – It’s no secret that I’m a fan of cost controls – I view them as free insurance that every brand should take. Especially if you’re a smaller brand, or launching a new product, cost controls can be a lifesaver because they can tell you a lot without spending a lot: in the event a cost-controlled campaign doesn’t spend, this can (and should) be interpreted as Meta saying, “Given the constraints provided, we can’t find a sufficient number of auctions where your bid, the quality of your creative and the ability of your product + site to convert, is sufficiently high.”

That may not be a welcome message, but (news flash) it was a message that was going to be sent in one of two ways: (1) the above version, where your ads just don’t serve or (2) the “highest volume” version, where Meta chews through $5,000, $10,000 or more over the period of 1-2 months with little-to-no conversions to show for it. I know which option I prefer.

Testing Ad Set / Campaign – I’m a big fan of having a structured way to introduce new creatives into an ad account, whether that’s a testing ad set, a testing campaign (usually something like a Advantage Campaign Budget (formerly known as a “CBO”) campaign with 2-3 ad sets – one broad, one LAL, one Interest Stack). Throw a cost cap on it, duplicate the ads across each ad set using Post IDs, give it a healthy budget (5x Cost Cap/Day), then get ready. One of three things will happen:

Option #1: A winner emerges. One ad will get the vast majority of spend, and will serve the overwhelming majority of the time. Congratulations. Copy that ad (use Post ID) over to your always-on campaign, then go make a bunch more creatives. Rinse & repeat.

Option #2: Nothing. The second-most-common outcome is that most of your ads simply don’t serve. This tells you (like above) that Meta doesn’t think your bid + creative + product/offer is good enough – so improve one (or all).

Option #3: Multiple pseudo-winners. This is the most vexing of all possible options. It occurs when Meta spends (relatively) equally across 2-4 creatives, with none emerging as a true winner. In this case, you have several paths open to you: (a) review each creative in detail, find the commonalities, and attempt to create a “fusion” ad that incorporates those commonalities; (b) promote all of them or (c) wait until a true winner emerges. There’s no right answer, but my preference tends to be (a) or (c).

I’m not saying that the above is the ideal way to run a Meta Ads account (heaven knows that I’ve seen accounts spending $5,000 a day profitably using a single Advantage+ campaign on Highest Volume with three creatives), but I think it gives brands the best chance at success while providing a substantial amount of downside protection.

In general, the more complex, niche or unknown your product/service/solution, the more you’ll benefit from the above structure; the more mass-market, impulsive, easy-to-buy your product is, the more likely it is that you’ll be able to scale regardless of the structure. While the latter might seem like an enticing situation, it’s a double-edged sword: if you can scale easily, so can your competition.

#5: Audiences

For the last three-plus years, the “bro-dience” strategy has been the single-most-prevalent targeting setup in Meta Ads. But in both the accounts I see and in my conversations with CMOs, Media Directors and Agency Execs, one thing is apparent: that pendulum is starting to swing the other way.

Mathematically, this makes absolute sense: there’s a segment of the overall Meta audience that’s (essentially) an online QVC: they’re tuned in, they have some money, and they’re keen to buy (or submit lead forms) for products/services that are cool/fun/interesting/potentially useful. Meta knows exactly who these users are, and a significant percentage of them likely fall outside of your traditional interest stacks or lookalikes (this is both a case why you should do more audience research and why you need to improve your LALs – but more on that later). Meta’s solution to this “problem” was broad targeting (and now, Advantage+ audiences) – two settings that allow it target whomever it desires (while respecting the geographic + exclusion constraints). At the outset, this worked great: Meta was able to serve ads to high-proclivity buyers cheaply, resulting in some eye-popping case studies. These wonderful case studies helped increase adoption, which kicked off a vicious cycle of more and more bidders for the same high-proclivity audience.

Long story short: when every advertiser uses broad, the available advantage declines. The purchase-happy segment of the Meta user base is only so large. The end result? Meta now has to work harder and longer to find the commonalities among each advertiser’s converting users. The machine, smart as it is, has picked all the low-hanging fruit (and most of the mid-level fruit, too). The QVC audience is overly diluted and tapped out.

The alternative to broad-only is to use a strategic combination of broad audiences, lookalikes (“LALs”) and interest stacks (“IS”) to help Meta identify the users with the highest probability of purchasing/converting.

Instead, what high-performing accounts do is:

Conduct diligent, ongoing customer insights research
Use those insights to craft both relevant targeting (audiences) and creative
Leverage existing customer data to create both exclusion and targeting lists
Segment existing customer data into relevant, discrete segments – buyers of X product/brand; prospects who requested a demo, but didn’t show up; individuals who exhibit Y behaviors. Then use those segments as seeds for LALs.

Of these, the 4th is the most important: Lookalikes are incredibly sensitive to initial conditions:

The above visualization shows the relative overlap between a total population and two segments of the same population. In this case, it’s an entire CRM (“All Leads” – in blue), just the “ideal” segment (in red), and just the “nurture” segment (in yellow). If this brand had followed the standard playbook and simply uploaded all their leads and created a LAL from those leads, they’d only target the ~12M people in the US encapsulated in that large, blue circle on the right.

They’d miss the ~8M people who are most similar to their ideal leads, but NOT included in the All Leads LAL, and they’d miss the ~11M people included in the “nurture” LAL who are not included in the All Leads LAL. In this example, simply segmenting your seed audiences can unlock ~19M more people who are actually behaviorally more similar to your ideal or nurture leads.

The same is true for Interest Stacks. If you’ve done your audience research (and seriously, do more of it), odds are you’ll be able to find high-proclivity combinations of interests (+ behaviors and demographics) that allow Meta to convert better.

Mathematically, think of using Interest Stacks and Lookalike Audiences as holding a variable constant. When you target broad audiences, Meta has to try to find the Audience + Creative combination that results in a positive outcome, then repeat that process tens of thousands of times until it’s able to identify the properties of each (creative + audience) that correlate to positive or negative outcomes. That can be quite expensive and time-consuming. With the above, you’re holding the “Audience” section of the equation constant – effectively telling Meta, “These are the people most likely to convert; tell me which creative gets them to do so.” That’s a significantly less complex problem – and it’s one Meta is uniquely good at solving.

#6: Creative

This all brings us to the “Creative” section – which is (arguably) the second-most-important (data is the most important, and I’ll die on that hill – if you can’t consistently pass correct data to Meta, you’re effectively flying a plane without a blacked-out windshield and fault instruments – it’s only going to end one way).

Let’s start with the obvious: most brands don’t have a single great creative in their account. Hell, most brands don’t even have passable creatives.

The foundation of all great creative is audience understanding – why are they interested in what you’re offering? Why are they buying what you’re selling? In the overwhelming majority of cases, people are interested for one of the following seven reasons:

It solves a problem or addresses a pain point
It prevents something from happening (think: insurance, routine checkups)
It is something that they must have (think: soap, makeup, shoes, medicine)
It provides status or signaling (think: luxury items, car brands, shoe brands)
It addresses a third-party need (think: gifting, wedding venue shopping, etc.)
It speaks to a passion (i.e. new gym clothes or hiking boots or rock climbing gear)
It makes them feel good

Depending on your brand/product/offer, more than one of these may be applicable (for some brands, all seven are relevant) – though each audience/angle is likely to over index into 1-2 of these. The next step is to layer the reasons why people buy (above) with the reasons why they should buy from you. Those might include:

Unique / First-To-Market product (true innovation)
Superior product (in one or more ways)
Superior functionality / better addresses their needs
Brand reputation (this is what Hermes, Chanel, etc. rely on)
More favorable Price:Value Ratio
Better/More Favorable Availability
Other Intangibles (values, sustainability, etc.)

Ultimately, your creative needs to connect who the audience is (hence, why research is so damn critical) to why they buy to why they should buy from you, and present that in a format that will appeal to the audience.

To do that, you need three things:

Way More Creative – I’ve written about this extensively, but for Meta Ads creative, quantity drives quality. You must have a critical mass of assets flowing into your account in order to scale. If you’re not adding new creative into your account regularly, you’re going to have a difficult time.
Way More Diversity – If every ad in your account looks like it was created by the same designer, or shot by the same videographer – you have a problem. The same is true if your entire account is chock-full of statics, or carousels, or videos. The best performing accounts have a stunning diversity of creatives, from multiple creators, to multiple editors, to a balance of ad formats. If you’re wondering how to get creative, we use Soona and Trend for eCommerce product/photo shoots, plus have a full creative team and 3 photographers on staff.
Way More Creative Formats – your account should have a mix of different ad units, from Product Intros + Unboxings, to Us vs. Them style ads, to UGC/Faux UGC, Reviews, Founder Stories, Testimonials, Stats/Studies, Real/Fake, etc.

Put simply: you need to make more high-quality, tailored-to-each-audience ads if you want to succeed long-term on Meta. Running the same 3 ads in every ad set (yes, I’ve seen that) simply isn’t going to get it done.

#7: Management Philosophy

Finally, there’s the management philosophy – how you run the account. I’m a vocal proponent of an “eyes-on, hands-off” approach that focuses on batching changes, introducing new creatives at predictable intervals, and avoiding “day trading” style management that ultimately just costs far more money than it makes.

That’s not to say I think you should let an account that’s clearly ignoring your guardrails continue to do so (please don’t), or that you should never change anything (please do). Modern machine learning is like a container ship – it’s powerful and massive and it can pretty much get anything, anywhere given enough time. But it is not a speedboat. It isn’t nimble. In order to maximize the impact of Meta’s machine learning, you need to surprise let the machine learn. High volatility and variability of inputs leads to signal loss, which leads to suboptimal performance.

Eyes on, hands off. Give the machine the information it needs, give it the time to process it, and continually refuel it with new creatives and up-to-date data.

That’s it for this week! I’m off to London now, but I hope you all have a wonderful week!

Cheers,

Sam

Related Insights