TLDR;
- Two-sided marketplaces like childcare apps often waste money on Google Ads because the platform confuses people looking for a service (parents) with people looking for a job (sitters). The common assumption is that Performance Max campaigns are uncontrollable black boxes that make this problem worse.
- The data from this childcare app account, with over 5,700 signups, shows the opposite. A well-structured Performance Max campaign massively outperformed a standard Search campaign for acquiring parents, achieving a CPA of $2.52.
- The main thing isn't to avoid PMax, but to engineer its inputs. By feeding it precise audience signals built from top-performing search keywords and proven demographic data from Meta campaigns, you can force it to ignore the job-seeker traffic and focus only on high-intent parents.
- This article breaks down the data, shows you how we built the signals, and includes an interactive calculator to model the impact and a searchable keyword database to spot intent pollution in your own campaigns.
I see this all the time. A business is running what looks like a perfectly sensible Google Ads campaign, but the numbers just don't add up. The Cost Per Acquisition (CPA) is through the roof, and nobody can figure out exactly why. This gets a whole lot worse when you're running a two-sided marketplace. For an app connecting parents with babysitters, you're fishing in the same pond for two very different types of fish. A parent searching for a "babysitter needed tonight" looks almost identical to a student looking for a job with that exact title.
The common wisdom floating around on forums is that Performance Max (PMax) just throws fuel on this fire. You hand your assets and budget over to Google's black box, and it sprays your money across the web, supposedly making this intent confusion even worse. The advice is often to stick to manual Search campaigns where you have full control over keywords.
But what if that's completely backwards? What if PMax, when handled correctly, is actually the sharpest tool you have to solve this exact problem? I've been digging into the data from one of our accounts—a childcare services app called SitterTree—to answer a specific question: Can you systematically engineer a Performance Max campaign to surgically target only high-intent parents in a confused marketplace, and in doing so, get the CPA below a target of $2.50?
This isn't just a thought experiment. I’m going to walk you through the data from this single account, which generated 5,779 signups, to show how we diagnosed the source of wasted spend and how the PMax campaign became the most efficient part of the entire strategy, not the most wasteful. We're going to compare it directly against a standard Search campaign to see what really works. Forget the theory; this is a detailed look at the numbers.
So, where's the money actually going?
Before we can fix a leak, we need to find it. The first job in any account I look at is to get a clear picture of the unit economics for each customer segment. It's a straightforward approach, but most people don't do it properly. For SitterTree, the two universes are Parents (the demand) and Sitters (the supply). They behave differently, they're worth different amounts, and they cost different amounts to acquire.
The data from this account, looking at the Google Ads performance between July and November 2024, makes the split obvious. We spent $5,477 to acquire 3,085 sitters, for a CPA of $1.78. Finding people who want flexible work is relatively straightforward. On the other hand, we spent $8,030 to acquire 1,444 parents, for a blended CPA of $5.56. That's more than three times as expensive. And our target was under $2.50.
This initial split is the first clue. A blended CPA of $5.56 for parents isn't just a bit high; it's a sign that something is fundamentally inefficient. Either the audience is incredibly expensive, or a good chunk of that $8,030 spend is hitting the wrong people entirely—it's attracting sitters into the parents' campaigns.
Marketplace Segment Performance Snapshot
Executive audit of the two core customer segments within the account, highlighting the stark difference in acquisition cost and volume.
How do you spot intent pollution in a Search campaign?
This cost disparity led me straight to the QLH - Search (Parents) campaign. This is the kind of manual control campaign that experts often recommend. It's neatly structured into ad groups based on the buyer's journey: information-seeking (Upper Funnel), cost-comparison (Middle Funnel), and ready-to-buy (Lower Funnel).
On the surface, this looks good. But the numbers tell a different story. The all-time CPA for this campaign was a staggering $10.00. Even in its more optimised, recent phase, it was $7.83. This is where the budget was haemorrhaging. The leak wasn't everywhere; it was concentrated in specific areas.
Digging into the ad groups, the Lower Funnel: Hiring group, which should be the most efficient, had spent nearly $2,000 to get 184 conversions at a CPA of $10.53. Why? Because it was full of keywords that are hopelessly ambiguous. Take the top-performing keyword by volume: "babysitter needed". It drove over 1,200 clicks and 102 conversions. But its CPA was $10.41. This single phrase is a perfect example of intent pollution. It's exactly what a desperate parent would type. It's also exactly what a sitter would type when looking for job ads to apply for.
Even with extensive negative keyword lists—we had lists to exclude terms like "jobs," "gigs," "salary," "apply"—it's a losing battle. You're trying to patch a fundamentally leaky bucket. The search terms themselves are the problem. You end up paying for clicks from both sides of the marketplace, and the algorithm gets rewarded with clicks and engagement, even if they're from the wrong audience who will never convert. This drives up your costs and pollutes your conversion data, teaching the algorithm to go and find more of the wrong people.
To really see this in action, you have to go through the keywords one by one. I've pulled the data from the Lower Funnel: Hiring ad group into an interactive table below. I've added a simple Ambiguity Score—a subjective rating from 1 (clearly parent intent) to 3 (highly ambiguous). Have a look for yourself. Notice how the keywords with the highest ambiguity scores often have some of the highest costs and CPAs. This is the wasted spend, right there in plain sight.
Parent Search Keyword Diagnostic
Search the keyword data below. Note how terms with high intent ambiguity (score of 3) often correlate with higher CPAs, indicating wasted spend on job-seeking traffic.
| Keyword | Ambiguity Score | Cost / Conv. | Conversions | Total Cost |
|---|
So how do you force PMax to find the right people?
This is where things get interesting. The conventional Search campaign was struggling, with a CPA over $10. Meanwhile, the LH - Performance Max (Parents) campaign was humming along, delivering 289 signups at a CPA of just $2.52. It was hitting our target. So much for the uncontrollable black box theory.
The main thing isn't in avoiding PMax; it's in how you feed it. PMax is a powerful amplifier. Feed it garbage signals, and it will amplify your ability to waste money. Feed it incredibly precise, high-quality signals, and it will find you customers with an efficiency that a manual Search campaign can only dream of.
Our entire approach was based on this principle. We didn't just give PMax some ad copy and images and hope for the best. We constructed a highly specific set of instructions—the Audience Signal—using data we had already validated elsewhere. It's a layered approach.
Layer 1: Custom Segments from Proven Search Terms. We took the keywords that we knew worked from our Search campaigns. Not the ambiguous ones like "babysitter needed," but the high-clarity, parent-intent phrases. Think terms from the Upper Funnel: Info ad group like "finding a babysitter" and from the Lower Funnel: Competitors ad group which, although paused, showed a decent CPA. We bundled these into a custom segment with the instruction: "show ads to people who have searched for any of these terms on Google." This tells the algorithm to anchor its search around proven, high-intent behaviour.
Layer 2: Audience Insights from Meta Ads. We were also running campaigns on Facebook and Instagram. We knew from that data which demographic and interest profiles converted best. Our top performing ad sets were targeting parents of young children (0-5 years) layered with interests like child care products or behaviours like working parent. We replicated this inside Google Ads, creating audiences based on these same affinities and in-market segments. This gives PMax a clear demographic and psychographic profile of our ideal customer.
Layer 3: Your Own Data. Finally, we added our first-party data: retargeting lists of website visitors, past signups, and newsletter subscribers. This is the most powerful signal of all, as it tells Google to go and find more people who look exactly like these people who have already shown interest or converted.
By combining these three layers, you're not leaving it up to chance. You're giving the PMax algorithm a crystal-clear, multi-faceted blueprint of the person you want to find. You're telling it what they search for, who they are, and what they're interested in. The machine's job is no longer to guess; its job is to execute on a very specific brief.
PMax Audience Signal Engineering Framework
Layer 1: Search Intent (Custom Segments)
Build a custom segment using your top 10-15 unambiguous, high-performing keywords from standard Search campaigns (e.g., "finding a babysitter," "nanny near me"). This anchors PMax to proven buying intent.
Layer 2: User Profile (Audience Insights)
Replicate your best-performing Meta Ads audiences. Add Google affinity/in-market segments for "Parents of Toddlers," interests in child care products, and demographics for working parents. This paints a picture of who the customer is.
Layer 3: First-Party Data (Your Customer Lists)
Upload your customer lists, newsletter subscribers, and create detailed retargeting audiences (e.g., website visitors who did not sign up). This tells Google to find clones of your existing best customers.
What's the real financial impact of signal quality?
It's one thing to talk about strategy, but let's model the numbers. The difference between a poorly targeted campaign and a well-targeted one is not incremental; it's a step change. A campaign bogged down by ambiguous keywords might have a high CPA, say $10, and a low conversion rate. By feeding PMax with cleaner signals, you're not just getting slightly better; you are fundamentally changing the efficiency of every dollar spent.
Let's use a simple model. Imagine your baseline is the standard Search campaign, converting clicks to signups at about 13% (the all-time Google Ads average for this account) but with a high Cost Per Click (CPC) of $1.23 on those ambiguous terms. That gives you a CPA around $9.46. Now, what if you use PMax with precise signals? It can find better-qualified users across Display and YouTube at a much lower CPC, let's say $0.38 (the actual CPC from the PMax Parents campaign). Even if the conversion rate stays the same, your CPA plummets to $2.92. This is before you even factor in that better quality traffic is likely to convert at a higher rate.
The interactive calculator below lets you play with these levers. It's pre-filled with the actual data from the SitterTree Parent campaigns. Adjust the CPC and conversion rate to see how small changes in signal quality (which influences CPC) and landing page performance (CVR) can drastically alter your final CPA. See for yourself how getting from the $7.83 CPA of the Search campaign to the $2.52 CPA of the PMax campaign is just a matter of improving the inputs.
CPA Impact Calculator
Model the impact of signal quality (via CPC) and conversion rate on your final CPA. Baselines are taken from the SitterTree Parent campaign data.
Is PMax really that much better? A head-to-head comparison.
The final proof is in a direct comparison. Let's put the two parent-focused campaigns—QLH - Search (Parents) and LH - Performance Max (Parents)—side-by-side, looking at a recent, comparable period.
The numbers are stark. For a similar amount of spend (around $750 each), the PMax campaign generated 289 conversions, while the Search campaign only managed 102. The PMax CPA was $2.52; the Search CPA was $7.83. That's not a small difference. The PMax campaign was 206% more effective at turning budget into signups.
Why such a massive gap? It comes back to the core mechanics. The Search campaign was restricted to a single channel (Google Search) and was constantly battling intent pollution, paying a premium CPC ($1.23) for clicks that were a mix of parents and sitters. Even with diligent optimisation, it was fundamentally inefficient.
The PMax campaign, guided by our engineered Audience Signal, did three things better:
- It bypassed the keyword ambiguity. Instead of just relying on what people type, it focused on who people are, based on our signals. It could find a parent interested in baby essentials even if they typed an ambiguous search term.
- It hunted for cheaper inventory. PMax isn't limited to the expensive real estate of the search results page. It found our target parents across YouTube, the Display Network, and Discover, where clicks are much cheaper ($0.38 average CPC).
- It harnessed the full power of the algorithm. By giving it a clear blueprint of the target user, we unleashed the machine to find lookalikes and patterns we could never identify manually. It wasn't guessing; it was pattern-matching at a massive scale.
The chart below shows this performance gap visually. It’s not an indictment of Search as a channel. It's an indictment of using a single-channel, keyword-dependent strategy for a complex problem that a multi-channel, audience-first approach is better equipped to solve.
PMax vs. Search: Head-to-Head CPA
A direct comparison of Cost Per Acquisition (CPA) for parent signups between the engineered Performance Max campaign and the traditional Search campaign.
So what does this actually mean for your campaigns?
The biggest takeaway from this deep dive is that you need to change how you think about Performance Max. It's not a lazy, set-and-forget campaign type for people who don't understand Google Ads. It's a high-performance engine that demands high-quality fuel. The control you think you have with manual Search campaigns is often an illusion when you're dealing with ambiguous user intent.
Your job isn't to manually pick every keyword and placement. Your job is to become an expert in your customer and to translate that expertise into the language the algorithm understands: data and signals. The success we saw here wasn't an accident. It was the direct result of a strategy that treated PMax not as a black box, but as a precision tool.
If you're running campaigns for a two-sided marketplace, or any business where user intent can be easily confused, stop fighting a losing battle with negative keyword lists. Instead, focus your energy on building the best possible Audience Signal. Do the research in your other channels. Find out what your best customers search for, what they're interested in, and who they are. Then, package that intelligence and hand it to PMax. You might be surprised when the uncontrollable campaign becomes the most profitable one in your account.
This kind of detailed analysis and strategic rebuilding is what we do day in, day out. Many businesses get stuck with high CPAs because they're working with a flawed model of how these platforms actually operate. If you're seeing similar issues in your own account and feel like you're hitting a wall, it might be time for an outside perspective.
We start every engagement with a free, no-obligation strategy consultation. We'll get into your accounts, crunch your metrics, and show you exactly where the opportunities are. If you're serious about fixing the leaks and scaling efficiently, it's a conversation worth having.
Cite this Research & Data
If referencing this research, data, or strategy in your publication, please use the attribution below.
Lukas Holschuh
Founder, Growth & Advertising Consultant
Great campaigns fail without expertise. Lukas and his team provide the missing strategy, optimizing your entire advertising funnel—from ad creatives and copy to landing page design.
Backed by a proven track record across SaaS, eLearning, and eCommerce, they don't just run ads; they engineer systems that convert. A data-driven partnership focused on tangible revenue growth.