Who's Actually Driving Your Media Budget?
A Calgary media strategist breaks down what Google and Meta's ad recommendations actually optimize for, and when it's smarter to say no to the algorithm.
Calgary has about 1.6 million people in it. Meta's Advantage+ tool told Gillian Lemishka she could reach 10 million of them with a single food-and-drink interest audience.
Lemishka is director of digital strategy at Mediology, an independent media agency, and she pulled that number live during a SocialNext Marketing Alliance webinar on a question most marketing teams don't stop to ask: when a platform pushes you toward more spend or less control over targeting, who does that recommendation actually serve? Her answer, built around the old fable of the frog and the scorpion, was that platforms aren't lying to you exactly. They're just built to maximize spend, the same way the scorpion is built to sting, and it's on the advertiser to ask a few more questions before the ride starts.
What “AI” actually means here
Lemishka didn't argue for turning everything off. For sales and transactional campaigns, especially ecommerce, tools like Performance Max and Advantage+ can genuinely lift results, provided the advertiser accepts less visibility into targeting and placement in exchange. For awareness or behaviour-change campaigns, the tradeoffs get riskier: less control over where ads run, weaker signal on who's actually seeing them, and real exposure to brand safety issues if ads land next to unsafe content.
The dividing line, in her framing, is what the campaign is optimizing for. Search deserves click-based optimization. Nearly everything else doesn't. She cited Nielsen research that found no reliable correlation between click-through rate and ad recall, brand awareness, or purchase intent across sectors.
Google Ads: the optimization score and the budget nudge
Nowhere was this clearer than Google's campaign optimization score. Log in and see 70%, and the instinct is to chase 100%. Lemishka called this a psychological play by design, because the score doesn't actually correlate with campaign performance. It's a mechanism to get advertisers clicking “apply" on suggestions.
Some of those suggestions:
“Your campaign is limited by budget." Lemishka showed a real campaign with a 17.55% click-through rate and top-of-page rate in the nineties that Google was still flagging as budget-limited, nudging toward more spend.
Opting into Google Search Partner Networks. This one promises more conversions. Lemishka cited research from marketing data scientist Dr. Augustine Fu showing that turning on search partners significantly increases non-human traffic compared to standard Google search. More clicks, not necessarily more customers.
Ad strength labels. Lemishka ran a live test with the group, showing two real search ads and asking the audience to guess which one Google rated higher. The ad with better click-through rate and lower cost-per-click was labelled “poor." The weaker performer was rated “average." Her takeaway: ad strength scores don't affect delivery, they're just another nudge toward "improving" copy that may already be working.
Irrelevant recommendations. Google flagged app conversion tracking and in-app engagement suggestions on a campaign that had no app at all, pointing to how generic some of this “optimization" really is.
Her advice: check whether the campaign is actually underperforming before acting on the warning. Platforms have an obvious incentive to nudge spend upward regardless of whether it improves outcomes.
The fix Lemishka recommends is simple: dismiss what doesn't apply, and don't be surprised when the same recommendation resurfaces about four weeks later. She also flagged a real incident where Google re-enabled search partner networks on a live campaign after her team had turned it off during setup, catching it only because a client noticed the traffic quality drop within two days.
Meta: opportunity scores and audience math that doesn't add up
Meta runs a parallel playbook with its own opportunity score and an auto-apply setting that lets recommendations get implemented without review. Mediology avoids auto-apply entirely, on the logic that a human still reads the room better than an algorithm does.
The budget recommendation carousel got specific pushback. Meta shows several spending tiers with a “recommended" option pre-selected in the middle, alongside a chart where more spend always produces a smoothly rising results curve. Lemishka called this a familiar psychological play: present a middle option as reasonable so it reads as restrained rather than as a sell.
The sharpest example was the Advantage+ audience number from the top of this piece. Manually defined, Lemishka's Calgary food-and-drink audience came to about 4.7 million people, itself already well above the city's population. Advantage+ more than doubled that estimate to 10 million, and scored the inflated version as “100% optimized."
She also flagged Audience Network, a placement expansion Meta pushed hard until recently. The same Dr. Fu research that examined Google's search partners found Audience Network placements carry a similarly inflated share of non-human traffic.
Both platforms drew the same complaint from attendees during Q&A: settings can revert or features can re-enable themselves after a campaign is live, without a notification. Lemishka's rule is to check configuration again after publishing, on any platform, every time.
Where the AI tools actually help
Lemishka didn't argue for turning everything off. For sales and transactional campaigns, especially ecommerce, tools like Performance Max and Advantage+ can genuinely lift results, provided the advertiser accepts less visibility into targeting and placement in exchange. For awareness or behaviour-change campaigns, the tradeoffs get riskier: less control over where ads run, weaker signal on who's actually seeing them, and real exposure to brand safety issues if ads land next to unsafe content.
The dividing line, in her framing, is what the campaign is optimizing for. Search deserves click-based optimization. Nearly everything else doesn't. She cited Nielsen research that found no reliable correlation between click-through rate and either ad recall or purchase intent across sectors.
From the Q&A
Asked how to pitch a new advertising platform to leadership, Lemishka's advice was to start with who's actually there. TikTok's average user is now in their mid-twenties to thirties, not the teen audience most marketing teams still picture from the platform's early years. Pairing platform demographics with a case study or consumption research makes a stronger case to management than intuition alone.
On employment ads, which lose access to age and income targeting by policy, Lemishka described a workaround her team has used: building a custom intent audience based on people who've visited specific industry job sites or searched relevant terms, rather than relying on demographic filters that are off the table entirely.
