Experimental Ad Budgets Are Growing. Not Because Brands Want To.

By Notice Me Senpai Editorial

There's a story making the rounds about how agencies and brands are getting braver with experimental budgets. The framing is upbeat. Words like "innovation" and "leaning in" are everywhere. One executive told Digiday that "there's no other option. You have to lean into new things." Which, if you read it quickly, sounds like excitement. Read it again. "There's no other option." That's not enthusiasm. That's a fire exit.

The real story here isn't that experimental budgets are growing. It's why they're growing. And the why changes everything about how you should respond.

The 80/20 rule is breaking, and not on purpose

Markacy, a performance marketing agency, historically ran an 80/20 split. Eighty percent in core channels (Meta, Google, the usual). Twenty percent in experimental. They've shifted to 75/25. Noble People, another agency, is moving dollars that were previously earmarked specifically for Meta and Google into upper-funnel, experimental media.

On the surface, a 5-point shift doesn't sound dramatic. But when you're talking about accounts spending $500k or more per month, that's $25,000 in monthly spend being pulled out of channels that used to be reliable and put into channels that aren't proven yet. Multiply that across an agency's entire book of business and you start to understand the scale of what's happening.

I think the industry is calling this "experimentation" because the alternative framing is too uncomfortable. The alternative framing is: core performance channels are deteriorating, ROAS is declining, and teams are quietly evacuating spend into anything that might work better. Calling it experimentation makes it sound strategic. In most cases I've seen, it's closer to triage.

Three forces pushing money out the door

The Digiday piece identifies three drivers, and they're worth separating because each one requires a different response.

Economic uncertainty and tariffs. Brands are hedging against macro risk by not putting all their budget into two platforms controlled by two companies that could change their auction dynamics overnight. Fair enough. Diversification as insurance isn't new, but the urgency is.

Tech advances in retail media and CTV. Connected TV ad spend keeps climbing, and retail media networks from Amazon, Walmart, and Target are getting sophisticated enough to compete for mid-funnel dollars. These aren't speculative anymore. They have real measurement, closed-loop attribution, and in some cases better signal than Meta can offer post-ATT.

AI forcing new approaches. This is the one that's quietly reshaping budget structures. When AI Overviews eat into your search traffic, when ChatGPT starts answering the queries your Google Ads used to capture, when organic click-through rates drop 30-40% on AI Overview queries, the math changes. Performance channels that were predictable for a decade are becoming less predictable. And unpredictable channels don't deserve 80% of your budget.

In one account we manage (mid-six-figure monthly spend, B2B SaaS), Google Search ROAS dropped about 18% year-over-year with no change in strategy, creative, or landing pages. The auctions just got more expensive and the clicks got less qualified. We moved roughly 8% of that budget into CTV prospecting and LinkedIn conversation ads, and honestly, the results have been uneven. Some months better, some worse. But the trajectory on Google was only going one direction, so standing still wasn't really an option either.

AEO is becoming a real line item

Pawco, a pet food brand featured in the Digiday piece, allocated a 10% budget increase in Q1 2026 specifically for AI and AEO. Answer Engine Optimization. Six months ago I would have dismissed that as a conference buzzword. I'm less sure now.

AEO is essentially optimizing your content and brand presence so that AI systems (ChatGPT, Perplexity, Google's AI Overviews, Copilot) reference you when answering questions in your category. It's not paid media in the traditional sense. It's more like a hybrid of SEO and PR, focused on making your brand the answer that AI models surface.

Pawco's experimental budgets sit at 15-20% of total spend now, which is aggressive for a brand their size. But the logic tracks. If consumers are increasingly getting product recommendations from AI chatbots instead of Google search results, the brands that show up in those AI responses will capture demand that used to flow through search ads. The brands that don't will watch their customer acquisition costs climb and not fully understand why.

I wouldn't go as far as Pawco's allocation for most brands. Not yet. But if you're spending zero on understanding how AI systems reference your category and competitors, you're probably already behind in ways that won't show up in your dashboards for another 6-12 months. That's the tricky part. By the time the data confirms the problem, the early movers have already built the moat.

Flexibility over frameworks

One detail from the Digiday piece that stuck with me was BEHR paint's approach. They build flexibility into their annual planning rather than locking into rigid percentage allocations. On paper, that sounds like a non-answer. "We're flexible" is what every brand says. But in practice, most annual plans I've reviewed are anything but flexible. They lock channel allocations in October for the following year and treat the plan like scripture.

BEHR's model seems closer to what actually works right now: set a core floor (the minimum you'll spend on proven channels), set an experimental ceiling (the maximum you're willing to risk on unproven ones), and let performance data move dollars between those guardrails monthly. Not quarterly. Monthly. The pace of change in ad platforms is too fast for quarterly reallocation cycles, and I think most teams know this but their planning processes haven't caught up.

This connects to something we covered recently about how YouTube is expanding its affiliate program to smaller creators. These emerging channels and formats are where experimental dollars are landing, and the brands with rigid annual plans simply can't move fast enough to test them while the opportunity is fresh.

What a 75/25 split actually looks like in practice

If you're running a $100k monthly ad budget and want to move from 80/20 to 75/25, here's roughly how I'd think about the reallocation.

Pull $5,000 from your worst-performing Google Search campaigns. Not your brand campaigns, not your top converters. The long-tail campaigns where CPC has crept up 20%+ year-over-year and conversion rates have flatlined. Those are the ones where Google's auction dynamics are working against you, and incremental optimization won't fix a structural problem.

Split that $5,000 across two tests, not five. Two is enough to learn something. Five spreads too thin to be statistically meaningful at that spend level. Run each for 30 days minimum. Measure on a cost-per-qualified-lead basis, not top-of-funnel metrics, because new channels always look good on impressions and terrible on last-click attribution. You need a metric that captures actual pipeline impact.

Where to put those two tests depends on your vertical, but CTV prospecting and Amazon's demand-side platform are producing real results for the DTC and e-commerce brands I talk to regularly. For B2B, LinkedIn's conversation ads and Reddit's interest-based targeting are worth investigating. The point isn't which channel. The point is having a structured process for testing that doesn't require a board presentation every time you want to try something.

The uncomfortable part nobody's saying out loud

Experimental budgets growing from 20% to 25% sounds like progress. And maybe it is. But the subtext of every conversation I've had with media buyers in the last few months is less optimistic. The subtext is: we used to know where to put money and feel confident about the return. We don't anymore. Not because we got worse at our jobs. Because the platforms changed underneath us.

Meta's algorithm is increasingly a black box that decides your targeting for you regardless of what you set. Google's search results are being eaten by its own AI. TikTok's regulatory future is genuinely unclear. The "core" in "core media spend" is wobblier than it's been in maybe a decade.

So when an agency moves from 80/20 to 75/25 and frames it as bold experimentation, I'd gently push back on that characterization. It's not bold. It's overdue. And frankly, for a lot of brands, 75/25 still isn't aggressive enough. If your Meta ROAS has dropped more than 15% in the last 12 months and your response has been to optimize harder within Meta, you might be perfecting your technique on a shrinking stage.

The brands I'd bet on right now aren't the ones with the biggest experimental budgets. They're the ones who stopped treating "experimental" and "core" as fundamentally different categories. Every channel is experimental until the data says otherwise. And right now, a lot of channels we used to call core are earning that label less and less each quarter.