Why your Demand Gen budget is too small to matter

There’s a specific kind of budget line that exists to be technically true rather than actually useful. A brand wants to look like it’s testing upper-funnel formats. So Demand Gen gets a token allocation, a few percent of the Google account. That’s enough to say the channel is “live.” But it’s not enough spend for the algorithm to learn anything, or for anyone to draw a real conclusion from the results. Then, six months later, someone points at Demand Gen’s flat ROAS as proof the channel doesn’t work. Nobody asks whether it was ever funded well enough to succeed.

This isn’t a new problem. Marketing effectiveness research has argued for two decades that brand-building activity needs a meaningful share of budget to compound, not a symbolic one. The IPA’s most cited work on the subject comes from Les Binet and Peter Field’s analysis of effectiveness case studies. It put the ideal long-run split at roughly 60% toward brand-building and 40% toward short-term activation for balanced growth. Most Google accounts run something closer to the inverse. The overwhelming majority sits in Search and PMax, with a rounding error left for anything that looks like awareness.

What a test budget buys you

An underfunded test produces a misleading result. Automated bidding systems need volume to exit their learning phase and find efficient audiences. A campaign running on a few hundred dollars a month rarely gets there. Someone in a QBR calls it inconclusive before that happens, and the budget moves back to Search. The channel doesn’t fail. It never got the chance to.

The part that doesn’t show up in last-click reporting

Even a properly funded Demand Gen or YouTube campaign has an image problem, because its value mostly shows up in someone else’s report. Last-click attribution correctly credits whichever channel closes the sale, which is usually Brand Search or PMax picking up a warm, brand-aware searcher. Attribution models don’t trace that awareness back to the upper-funnel campaign that created it weeks earlier. A marketer looking only at Demand Gen’s own conversion count sees just the smaller half of what it did, as one Fospha Academy lesson on the subject lays out.

What happens when the allocation scales

Fospha’s Full-Funnel Google Report offers a useful data point here. It tracked a cohort of retail advertisers through Q4 2025. Accounts running Demand Gen at a meaningful share of Google wallet saw materially stronger account-wide ROAS than those parked at a token allocation. The downstream channels benefited too, not just Demand Gen’s own numbers. One retailer in the report leaned into that shift heading into peak. As Demand Gen scaled, its Brand Search cost per acquisition fell sharply, alongside broader Google revenue growth for the account overall. It’s one dataset, but it’s consistent with what the Binet and Field research would predict. Fund the awareness layer properly, and the capture layer downstream gets easier.

Where this goes wrong

The mistake isn’t only underfunding Demand Gen. It’s also funding it by taking money away from Search or PMax and calling that a rebalance. Cutting a channel that’s already converting to fund one that’s still finding its footing tends to produce a short-term dip with no offsetting gain. Nothing upstream grew to replace what got cut. The accounts seeing the strongest results funded the new channel as additional spend, at least while it matures.

The decision on the table

Nobody is arguing Demand Gen deserves half the Google budget. The argument is narrower. A line item too small to exit its own learning phase can’t be evaluated on its own reported ROAS. It hasn’t had a fair chance to earn one. Before killing a channel that “isn’t performing,” ask whether anyone ever funded it past the point where a real answer was possible.