The Money Was Never Just in the Ad Spend

Half a million dollars a year. That’s what one growth team found by auditing refund requests and failed credit card charges, work that never touched a single ad campaign. When a marketing budget needs to stretch further, the instinct is to go straight to the media plan and push down cost per acquisition. But the real efficiency gains in a growth budget rarely come from a cheaper install so much as from the parts of the business nobody thought to check.

A handful of habits tend to separate a growth team that’s genuinely getting more out of the same spend from one that’s just chasing a lower number: how they read cost per acquisition in the first place, how hard they’ve looked for revenue that was already sitting there, whether they can tell real growth from spend that’s just chasing itself in circles, and whether the rest of the product can even hold onto the users they’re paying to bring in.

A Cheap Install Can Still Be a Bad One

Cost per acquisition looks like a scoreboard, but it behaves more like a proxy for something further downstream. A trial that costs less to acquire isn’t automatically worth more, because what matters is what happens after the install: retention, renewal, and the ratio between lifetime value and what it cost to get there. Paying 50 percent more for a customer worth twice as much over time is the better trade, even when the acquisition report alone makes the cheaper option look smarter.

One subscription business running a 30-day trial learned this by tracking which discount codes moved people from trial to paid and which ones just gave away revenue the business would have collected anyway. Some codes earned their keep by converting hesitant trial users. Others simply lowered the average order value without changing a single mind. 

Once the team separated the two and shifted spend toward the codes doing real work, return on ad spend rose 10 percent, giving them room to let cost per acquisition climb as they scaled, now that a rising number read as growth rather than a warning sign.

That’s the discipline worth building here: stop asking how low this number can go, and start asking what it’s standing in for, then make a habit of checking the answer, before finance asks the question for you.

“CPA is always a proxy for ROAS, not the finish line itself. The real skill is understanding exactly what it’s standing in for.” AGS Berlin 2026

The Revenue That Was Already Yours

Somewhere between failed credit card charges and an overly generous refund policy sits a chunk of revenue most growth teams never go looking for, because it doesn’t feel like marketing’s job, even though it directly changes what the marketing budget can afford to do. Auditing refund approvals and figuring out which payment methods are failing is real grunt work, and it’s how one team found roughly 500,000 dollars a year without spending a cent more on a single campaign.

“Revenue recovery doesn’t sound very sexy, but it is. We’re generating around half a million dollars a year through refund and failed-payment experiments alone, without paying for a single one of those customers twice.” AGS Berlin 2026

Lifecycle messaging has a similar blind spot. Most companies run some version of push notifications and onboarding emails, and most of them play it safer than they need to, especially in the first few days after someone signs up, when a user is most open to hearing from a brand and least likely to have already made up their mind. Testing how far that frequency can go, while watching engagement closely enough to protect long-term deliverability, tends to surface a version of free money: revenue with no corresponding line item in the media plan, and enough of it to justify paying more for a customer on the paid side because more of them will now make it through the funnel.

Some Growth Is Just Redirected Demand

A different kind of leak shows up once a company is running more than one brand in the same category. Every euro spent acquiring a user might just be redirecting someone who would have found a sibling brand anyway, or found the product organically. 

One portfolio dealing with this set a strict pecking order: a lead brand gets priority in a given market, the others get bid down there, so the brands stop competing against each other for the same install. The same logic holds for any team running multiple products, or the same product across overlapping regions: before crediting a channel with new demand, check whether it created that demand or just intercepted it.

The honest way to check is to test it. Marketing mix models, increasingly built with AI, treat a team’s own budget shifts and seasonal swings as natural experiments, separating channels that move revenue from ones just riding along. The common mistake is testing for two days and calling it settled, when the real signal only shows up after running a channel dark for weeks, sometimes a month. One test worth running everywhere: black out branded search, usually the channel most likely to be taking credit for demand that already existed.

AI is a legitimate tool for that kind of modeling, but its judgment has real limits: it doesn’t know a company’s goals, its customers, or a competitor disappearing overnight, the kind of shift that explains a swing in performance better than any dashboard will. One team nearly paused its highest-spending campaign over an AI alert that missed that context, caught by a human just in time. These tools also work best once the repetitive, deterministic work around them is already automated, since the same question can return a different answer each time, fine for a judgment call, risky for anything meant to run the same way daily.

None of This Works If the Product Can’t Hold On

Every one of those gains assumes the product is ready to receive the traffic. A creative can be flawless and a campaign perfectly targeted, and none of it converts if the App Store listing someone lands on doesn’t match what the ad promised, if onboarding buries the product’s value under too many screens, or if the paywall shows up at the wrong moment. Fixing that takes the acquisition, product, and monetization teams sitting in the same room and treating conversion as one shared problem across all three. Part of why this slips through the cracks: the three teams are usually measured on three different things, installs for marketing, engagement for product, revenue per user for monetization, so the leak between them often has no single owner until someone forces the conversation. Skip that step and every other efficiency gain hits a ceiling nobody sees coming until it’s already there.

The teams getting more out of this year’s budget are less often the ones running the cleverest media buy than the ones treating cost per acquisition as a question instead of a target, willing to go looking for the money that was already theirs before asking for more of it.

The biggest efficiency gains in a budget are usually hiding wherever nobody thought to look: refund policies, discount codes, lifecycle timing, the stuff this piece just walked through. The same kind of conversation plays out live too, at invite-only events that rotate to a new city every few weeks, Berlin, São Paulo, Tokyo among them. Check our event calendar and request your invite now, seats move fast.