By Anthony Costanzo, Chief Analytics Officer, Mile Marker
Marketing automation platforms were supposed to simplify everything. Feed the machine enough data, trust the algorithm, and let performance take care of itself. For ecommerce brands, that promise has mostly held. For CPG and other “signal-poor” categories, it hasn’t.
Yet Google Performance Max and Meta Advantage aren’t optional. They are now core buying mechanisms—baked into platform roadmaps and sales incentives. Brands that avoid them do so at their own risk. The problem isn’t whether to use these tools; it’s how to use them when your business model doesn’t naturally generate the signals they require.
Automation systems are only as good as the data they learn from. Ecommerce brands hand over transactions, carts, and customer histories. Most CPG brands can’t. Their sales happen through retailers. Their websites are informational. Their conversion events are, from the platform’s perspective, largely invisible. The result is a black box optimized against weak proxies—and often delivering weak results.
Still, there are ways forward. None are perfect. All involve tradeoffs. But together, they form a practical framework for making automation work in categories where signal scarcity is the rule, not the exception.
1. Stop Optimizing to Easy Signals
When brands lack purchase data, the temptation is to substitute engagement metrics: time on site, scroll depth, page views, button clicks. These feel like reasonable proxies for intent. In reality, they’re exactly the kind of signals automation systems exploit.
Machine-learning platforms are extremely good at finding people who will sit on a page for 30 seconds or tap a button—often without any commercial intent. The result is cheap traffic, impressive in-platform metrics, and little real-world impact. In some cases, brands even see spikes in traffic from geographies, devices, or audience segments that don’t resemble their customers at all.
A better approach is to define significant actions—behaviors that are both meaningful to the business and difficult for algorithms to manufacture at scale. These actions usually require multiple steps and genuine effort: using a store locator, interacting with a shoppable retailer module, completing a product finder, downloading detailed content, or participating in a structured lead experience.
The point isn’t friction for its own sake. It’s intent. Signals should represent real consumer consideration, not passive engagement. And just as importantly, brands must validate those signals by analyzing the quality of traffic they generate—not simply celebrating rising conversion counts.
2. Borrow Signal from Retailers—Carefully
Another increasingly common workaround is borrowing signal from retail partners.
Retail media networks now allow CPG brands to run Google and Meta campaigns optimized against retailer-owned conversion events. In effect, the retailer’s purchase data becomes the learning signal for automated campaigns, even though the brand doesn’t own the transaction.
When executed well, this can be powerful. It gives automation engines the kind of outcome data they’re built for. But not all retail media partnerships are created equal. Some require brands to operate through managed services with high fees, limited transparency, delayed reporting, and little control over campaign structure.
The more effective partnerships are those that provide brands with direct access—to pixels, product catalogs, and campaign management—while still delivering closed-loop measurement. Brands give up some control either way, but the difference between visibility and opacity can determine whether borrowed signal actually drives incremental growth.
3. Lease Signal from Third Parties
The most sophisticated—and costly—option is leasing signal from third-party data and measurement providers.
These companies aggregate loyalty card data, receipt scans, and consumer panels to link offline purchases to digital identities. On platforms that allow third-party measurement inputs, such as Meta and TikTok, this data can be used not just for targeting, but for optimization and learning.
There are limitations. Coverage is incomplete. Attribution is probabilistic. Costs add up quickly. And Google remains largely closed to third-party conversion signals. But for brands with sufficient scale, leasing signal can provide a bridge between offline sales and digital optimization that simply doesn’t exist otherwise.
The Bigger Picture
None of these strategies solve the transparency problem. Performance Max and Advantage campaigns still reveal little about where ads run or why decisions are made. Brands shouldn’t expect that to change. Automation aligns with platform incentives, not advertiser curiosity.
What can change is how brands feed the machine. Weak signals produce weak outcomes. Strong, intentional signals—even imperfect ones—give automation something real to learn from. For CPG marketers, success with automation won’t come from blind adoption or outright rejection. It will come from creativity, discipline, and a willingness to assemble a patchwork solution that reflects how their business actually works.
The black box isn’t going away. But with the right inputs, it doesn’t have to be a liability.
