Imagine: you walk into a local bakery, and instead of being met with a glass case full of delicious looking pastries, you see a chest-high countertop with someone standing behind it, ready to take your order. You want to order “The Best Cookie Ever” from the menu, but cookies can be very, very different based on what kind of ingredients are put in, and you have no idea what this cookie is made of. Chocolate chip? Oatmeal? Peanut butter? The person behind the counter looks like a qualified baker, but something seems odd.
You ask about these supposed “Best Cookies Ever,” and they claim their cookies are baked so well, on average a customer experiences 18% higher satisfaction after eating them. You raise an eyebrow, because that’s a weird thing to say about cookies.
Would you trust this bakery?
Would You Trust Performance Max?
Performance Max (PMax) campaigns, pioneered by Google, but now growing more common across advertising platforms, are exactly as the bakery I just described. Instead of ingredients for cookies, we’re talking about ad placements on search engines, display banners, videos, and more, and Google is spending your money based on how its AI believes it will be best spent to achieve low cost conversions.
Microsoft has since rolled out its own version, and Meta has Advantage+, which optimizes across Facebook, Instagram, Messenger, and Audience Network. They all aim to take the guesswork out of ad campaigns by automating the entire process: targeting, placement, creative, and spend distribution. This sounds great in theory, but so does “The Best Cookie Ever.”
A really important piece of context these ad platforms all glaze over is that they’re designed for eCommerce advertisers. To their credit, eCommerce by nature should not be limited by interests, since selling as much product as possible is always the goal. Whether it’s a knife set, clothing, pillows, or any other consumer good, anyone in any demographic that has money can buy the product. Performance Max shines here, since it optimizes creatives, platforms, placements, and dollars based on getting sales. However, even with eCommerce, Google doesn’t share placement or targeting data on PMax campaigns with advertisers, so if they wanted to expand or replicate results with their own campaigns, there’s no way to break the black box.
How’d We Get Here?
Around 2018, automated platform optimization was innovative and fantastic on Meta (then called Facebook), as it was novel and honest. Dollars were distributed based on opportunity, and campaigns worked great. Agencies had access to actual salaried account managers that knew the platform technology and helped get results for advertisers. But after ad buying outpaced hiring capacity during the pandemic in 2020, and data privacy regulations decimated tracking and measurable results in 2021, Google and Meta began to shift priorities in an effort to maintain their dominance.
Both companies fired or transitioned account managers to sales roles (incredibly pushy and misinformed sales roles, at that), stacked phony data points and dark patterns on their self-service ad buying platforms to trick businesses into spending more, and now they’re leveraging their incredibly expensive custom AI platforms to pull even more dollars away from businesses in the name of optimization. To go back to our earlier analogy, once-delicious cookies now taste like crap.
These platforms used to put together great, well-researched, scientific studies on consumer behavior on their platforms to help advertisers (like this great piece on effective video creative from Facebook in 2016). Now, they generate bogus citations claiming their own month-long studies from years ago that have no verifiable data to back up their claims of efficacy or effectiveness. In the blink of an eye, these ad companies went from transparent to “Trust me, bro,” and it should make everyone think twice about their motives to push Performance Max as a viable option.
To PMax or Not to PMax?
If a business needs results from specific or niche audiences, or has concerns around brand safety, Performance Max should not even be considered, plain and simple.
Audience networks (websites that partner with Google, Meta, Microsoft, and many other walled garden platforms) are much lower quality these days than they used to be, and given the amount of click bots that crawl some of those network websites to increase ad revenue, Performance Max campaigns only see those clicks as positive signals and can’t reduce or eliminate that traffic on its own.
Additionally, campaigns built on Meta’s Advantage+ setup suffer greatly from convergence, which happens when a machine learning model feels additional training will not improve the model. This means that no matter how long you run your campaign, performance ends up delivering all the ad dollars to one specific audience, like women over 65 on Facebook, because the system feels that audience has the highest likelihood of success. Since it’s a machine trying to find the lowest cost conversions possible, it’s not caring who or what is converting, which also means some of those traffic bots get counted in your conversion results.
The safest bet is to manually set campaigns up, splitting the budget based on how a business wants to spend that budget, setting up audiences manually, and optimizing and controlling everything manually. This takes some technical knowledge and can be time consuming, but knowing where ads are showing up, how those dollars are spent, and who the ads are reaching are all paramount signals to help make a campaign go from good to great. Not that it needs to be said, but having humans convert, and not bots, is also critical to success in any digital marketing effort.
Working with an agency partner like ChatterBlast can help with that, especially since we’ve been building campaigns out manually for close to ten years, and know all the ins and outs of that process. (Hi, I’m Joe, and I’m very much a human!)
At the end of the day, these ad platforms are not inherently bad. They’ve gotten painful to work with and cumbersome to navigate, but they do hold incredible value in the hands of a savvy advertiser. Their claims of helping businesses iterate faster, however, are merely sales tactics to take more money, and their claims that AI can run ads better are completely misguided. ChatGPT, CoPilot, and Claude (Anthropic) are the best AI models currently available, yet none of them are accurate enough to answer simple questions, let alone solve complex business problems like optimizing bids, targeting, or placements based on public response to tailored digital advertising.
Down the road, Performance Max or Advantage+ may come up in conversation with an agency or partner, but unless you don’t care about who you’re targeting, where your ads show up, or the quality of your results, you should always reject Performance Max to save your hard earned ad dollars for a better solution.
