Google Demand Gen Campaign Best Practices
A team launches a Demand Gen campaign with the same creative and targeting logic they used for standard Display ads, then wonders why performance underwhelms. Demand Gen looks like just another Google ad product on the surface — but running it like one is the fastest way to waste the budget behind it.
The Problem: Old Playbooks Don’t Transfer
Demand Gen replaced and expanded on what used to be Discovery ads, and it runs across some of the highest-attention surfaces Google owns — YouTube, Discover, and Gmail. But many teams still approach it with Search-style targeting logic or Display-style generic creative, missing what actually makes the format effective: it rewards native-feeling, visually rich content shown to warm, intent-adjacent audiences, not broad reach with repurposed banner ads.
Why Misusing the Format Gets Expensive
Demand Gen campaigns optimized poorly tend to fail quietly rather than obviously — they spend budget, generate impressions, and produce mediocre conversion rates without a clear signal of what went wrong. Because the format blends programmatic delivery with social-style creative consumption, a campaign that ignores creative quality or audience signal quality will underperform without necessarily showing up as an obvious red flag in surface-level metrics like impressions or reach.
Best Practices That Actually Move Performance
Lead with video, not static images. YouTube is one of Demand Gen’s primary placements, and video consistently outperforms static creative across this inventory. Even simple, well-produced short-form video tends to beat a static image carousel.
Build audience signals deliberately, not generically. Demand Gen performs best when fed specific custom audience signals — relevant search terms, competitor URLs, app categories, and YouTube channels your ideal customer actually engages with. Leaving targeting broad and letting the algorithm “figure it out” usually produces weaker results than feeding it strong starting signals.
Match creative to placement context. A Gmail-native ad format and a YouTube in-feed placement reward different visual styles. Designing one generic asset for all placements, rather than tailoring creative to where it will actually appear, leaves performance on the table.
Test multiple asset combinations, not single creative sets. Demand Gen’s asset-based structure allows multiple headlines, descriptions, and images to be tested in combination. Supplying only the minimum number of assets limits the algorithm’s ability to find winning combinations.
Give the campaign a real learning period before judging it. Demand Gen relies on machine learning to optimize delivery, and like most algorithmic campaign types, performance typically improves over the first one to two weeks as the system gathers conversion data. Pulling budget or making major changes too early disrupts that learning process.
A Practical Example
A travel brand running a Demand Gen campaign for last-minute getaway deals might feed the system custom audience signals built from recent travel-related search terms and competitor booking site visits, pair that with vertical video creative optimized for YouTube Shorts placement, and let the campaign run uninterrupted for two weeks before evaluating performance — rather than judging it against day-three numbers.
Key Takeaways
Demand Gen rewards specificity: specific audience signals, placement-appropriate creative, and patience through the learning phase. Treating it as a generic awareness format, or reusing Display-era assets, consistently underperforms what the format is actually capable of.
If your Demand Gen campaigns are underperforming expectations, it’s worth auditing creative variety and audience signal quality before assuming the format itself isn’t working. We’re happy to help with that review.
