Demand Gen is a Google Ads campaign format for YouTube, Gmail, and the Discover feed that works on demand a person hasn’t yet turned into a search query. The most common cause of disappointment with this format is judging it by the same metrics and timelines as a regular Search campaign. Below are the principles I follow when working with this format: how to build the structure, when to change the bidding strategy, and which mistakes are best avoided from day one.
What Demand Gen is and who it’s for
Demand Gen is a campaign format with visual ads (images and video) shown across three Google environments: the YouTube and Shorts feed, the “Promotions” tab in Gmail, and the Discover feed. Unlike Search campaigns, there are no keywords and no explicit query here; the platform selects the audience based on interests, behavior, and the advertiser’s own data.
The format grew out of Discovery ads and fully replaced them in 2023. Demand Gen’s purpose isn’t to capture demand that’s already formed (that’s a job for Search or Performance Max), but to shape demand before a person even starts searching for a solution.
This is a fundamental distinction worth understanding before launch: expecting Demand Gen to deliver the same CPA numbers as a Search campaign targeting commercial intent is a mistake from day one.
Expert advice: key principles
Separate remarketing from cold audiences
I always run remarketing and new-audience acquisition as separate campaigns rather than one, because they have different goals, different targeting settings, and often different placements. Combining both tasks into a single campaign makes it harder for the algorithm to understand which action to optimize for.
Give the campaign real time
I plan for a 60-90 day range for results to stabilize, and for a new account with no conversion history, I budget for as much as three months of campaign runtime before making any scaling decisions. The decision to “turn it off after two weeks because it’s expensive” is almost always made on statistically insignificant data.
Don’t judge it by last click
Evaluating Demand Gen by last-click attribution alone systematically undervalues the campaign’s real contribution to sales. The most common mistake I see in accounts during an audit is measuring an upper-funnel channel with lower-funnel metrics, as if it were an ordinary Search campaign targeting a hot, ready-to-buy query.
Start with your own signals, not broad targeting
I never start with cold, broad targeting. I start with the “Google Engaged” audience (people who have already interacted with the brand on Google-owned properties) and Custom Segments built from the niche’s top search queries. First-party signals give the algorithm the strongest possible start, and only then does it make sense to expand the campaign to a colder audience.
Placements can and should be controlled
Even though Demand Gen is marketed as a largely automated format, advertisers can see exactly where ads are showing, broken down by YouTube, Gmail, and Discover, and exclude specific placements if their results are consistently below average.
What the platform itself recommends
Google Ads’ official documentation adds several concrete benchmarks to this. For Maximize Conversions strategies, the recommended minimum budget is substantial (Google suggests a benchmark of $100 a day), while for target CPA/ROAS strategies, the budget should be at least 10 times the target cost per action. Assets need at least 14 days to “ramp up” before you make a decision to pause them. For merging underperforming ad groups, the benchmark is 30 conversions in 30 days, and bids and budget should be adjusted by no more than 5-10% (15% at most) at a time, so as not to disrupt the algorithm’s learning.
According to Google’s own data, advertisers who implement at least three of the platform’s four core recommendations (sufficient data volume, an adequate budget, Optimized Targeting with first-party seed lists on cold campaigns, and creative diversity, meaning at least three vertical, square, and horizontal images each plus one video per format) get, on average, 40% more conversions.
Typical mistakes
The most common mistakes I see in other people’s accounts during an audit are these. The most telling one is too small a budget: a campaign running on a few dollars a day, trying to cover an entire country, physically can’t reach enough impression frequency for the algorithm to have anything to optimize.
The second is launching Demand Gen where there’s no foundation yet. The format works best as a layer on top of already-working remarketing and search, not as a business’s first advertising activity with zero conversion history.
The third is leaving Optimized Targeting turned on where a campaign should be targeting a narrow, specific audience (pure remarketing, for example). This setting lets the system expand the audience beyond what’s defined, which is useful for cold prospecting for new customers (and that’s exactly how Google recommends enabling it), but harmful when the goal is to reach a specific segment precisely.
| Mistake | Consequence | Fix |
|---|---|---|
| Budget too small for the chosen geography | Insufficient impression frequency, algorithm can’t optimize in time | Narrow the geography to fit the budget, or increase the budget |
| Judging performance by last-click after 1-2 weeks | The campaign looks unprofitable even though it’s driving sales indirectly | Evaluate over 60-90 days and look at assisted conversions |
| Optimized Targeting left on for pure remarketing | The audience drifts beyond the defined segment | Turn it off for narrow remarketing campaigns, keep it on for cold prospecting |
| Switching to tCPA/tROAS before 50 conversions | The algorithm learns unstably, results are erratic | Wait for data to accumulate, run on manual or maximize-conversion strategies until then |
| One campaign for both remarketing and cold audiences | The algorithm doesn’t know which goal to optimize for | Split into separate campaigns with different goals |
Editorial analysis
In practice, running campaigns for Ukrainian clients, the biggest challenge with Demand Gen isn’t technical, it’s managerial: explaining to the business why the format can’t be judged by the same rules as a Search campaign chasing a hot query. When a client is used to transparent CPA from Search, any channel where conversion happens with a delay and partly outside the boundary of direct attribution looks, by default, like “wasted spend.” That’s why, before launch, I separately agree with the client on which metrics and what time horizon will be used to evaluate results; this saves the campaign from being switched off prematurely right when it’s starting to gain traction.
Who this is a good fit for
Demand Gen isn’t a replacement for Search or Performance Max, but an additional layer on top of them. It makes the most sense for a business that:
- already has working remarketing and a stable flow of conversions from Search campaigns, since Demand Gen extends reach on top of an existing base rather than replacing it;
- sells a product where the decision isn’t made instantly, requiring several brand touchpoints before conversion;
- has quality visual or video assets, or the resources to produce them, since the format rests almost entirely on creative;
- is willing to wait 2-3 months for results to stabilize rather than making a call after the first week.
For a business with no conversion history at all, a minimal budget, or no visual assets, Demand Gen isn’t a priority channel yet; it’s better to build a foundation with Search first.
What to do right now
- Split existing or planned Demand Gen campaigns into a dedicated remarketing campaign and a separate new-audience campaign.
- Check whether the Optimized Targeting setting matches each campaign’s specific goal: on for cold prospecting, off for narrow remarketing.
- Budget for at least 10x the target CPA if you’re planning a Target strategy.
- Agree with your team or client on an evaluation horizon of at least 60 days, and on metrics that account for assisted conversions, not just last-click.
- Don’t switch to tCPA/tROAS until the campaign has accumulated at least 50 conversions.
Risks and things to watch for
Demand Gen is developing rapidly: over the past year, Google has added shoppable formats for Connected TV, attributed branded search, and dynamic feeds for the travel niche. Some of these new capabilities aren’t yet available in every market or every account; before factoring them into a plan, I always check whether a specific feature is actually available in the client’s account rather than relying on general announcements.
Expert tip. I recommend treating Demand Gen not as a standalone tactical campaign with an end date, but as a permanent layer of awareness and audience warming that runs in parallel with Search, an “always-on” level of the funnel.
FAQ
How does Demand Gen differ from Performance Max? Performance Max works across all of Google’s inventory, including Search, and captures both formed and partially unformed demand. Demand Gen is limited to visual placements (YouTube, Gmail, Discover) and is specifically meant to shape demand before a person starts searching.
How long should I wait before evaluating results? I plan for a 60-90 day range, and for a new account with no conversion history, even up to three months.
Is Demand Gen suitable for a business with a limited budget? If the budget physically can’t sustain enough impression frequency in the chosen geography, the campaign won’t gather enough statistics to optimize; in that case, it’s better to narrow the geography or delay launch until the budget increases.
What is Optimized Targeting in Demand Gen, and when should you turn it off? This setting lets the system expand the defined audience beyond the specified segments in search of new conversions. Google recommends keeping it on for cold campaigns targeting new audiences, alongside first-party seed lists, and turning it off for narrow remarketing campaigns.
How many conversions are needed before switching to automated bidding strategies? At least 50, according to Google Ads’ official documentation. Before that point, the algorithm lacks the data for stable learning.
How often can bids and budget be changed in Demand Gen? Google recommends adjusting by no more than 5-10% (15% at most) at a time; sharper changes disrupt the algorithm’s learning and push the campaign back to an earlier optimization stage.
Do I need dedicated videos for Demand Gen, or can I use existing creatives from other campaigns? You can carry over creatives from other channels (email, social media), but the format benefits from diversity: at least three vertical, square, and horizontal images each, plus one video per format, increase the odds that the algorithm finds a winning combination.
Does Demand Gen affect Search campaign results? There’s no direct technical link, but campaigns that combine multiple formats and screens (Connected TV included) show a lift in incremental conversions according to Google’s data. Demand Gen works here as part of a broader funnel, not an isolated channel.
Conclusion
Demand Gen is a format for businesses that already have a working foundation of search and remarketing, and that are willing to evaluate results over a quarter, not a week. My core principles are simple: separate campaign goals, allow enough time and data, and don’t judge an upper-funnel channel by lower-funnel metrics.
If you want to figure out whether your account is ready to launch Demand Gen, let’s discuss it in an audit. More on working with Google Ads formats is on the platform page, and more on how I lower CPA on Search campaigns is in the article “How to Reduce CPA in Google Ads”. For the attribution model changes that directly affect how channels like Demand Gen are evaluated, see the article on Google Ads’ new 2026 attribution rules.