Demand Gen Campaigns: Google's Newest Ad Format Explained

Google Demand Gen campaigns launched as the successor to Discovery ads in 2024, and the paid media world largely responded with a collective shrug. That was a mistake. At Workflow AI Advisors, we've been running Demand Gen for clients across the US, UK, Australia, and the UAE since the format became widely available — and in the right setup, it consistently delivers ROAS figures that surprise even sceptical performance marketers.

This guide covers everything: what Demand Gen actually is, how it differs from Performance Max and standard Display, how to configure a campaign that doesn't bleed budget, and the audience and creative strategies that actually move the needle. No fluff, no vendor spin — just what works in practice.

What Are Google Demand Gen Campaigns?

Demand Gen is Google's visual, upper-to-mid funnel ad format that serves across YouTube (in-feed and in-stream), Discover, and Gmail. Think of it as the paid equivalent of social media advertising — immersive, image-led, designed to interrupt passively scrolling users rather than capture active searchers.

The format replaced Discovery ads in mid-2024. It brought with it expanded creative options (video assets are now supported alongside static images and carousels), lookalike audience capabilities that were previously absent from Google's stack, and tighter creative controls than Discovery ever offered. Critically, it also brought a bidding structure that rewards advertisers who understand funnel mechanics rather than those who just throw budget at automated systems and hope.

Where Search campaigns capture intent, Demand Gen creates it. That distinction matters enormously when you're building a paid media strategy that goes beyond bottom-of-funnel harvesting.

Demand Gen vs Performance Max: What's the Actual Difference?

This question comes up in almost every client onboarding conversation we have. The confusion is understandable — both formats use Google's automation, both can appear across multiple placements, and both sit inside the same Google Ads interface. But they serve fundamentally different strategic purposes.

Performance Max is a full-funnel, fully automated campaign that Google controls almost entirely. You supply assets, audience signals, and a conversion goal, and Google decides where, when, and to whom your ads appear. You get broad reach across all Google properties, but limited visibility into what's actually working. It prioritises conversion events Google can attribute.

Demand Gen gives you meaningful control back. You choose specific placements (YouTube, Discover, Gmail or combinations thereof). You define your own audience segments. You A/B test creatives at the ad level. You control bid strategies more granularly. It's designed explicitly for advertisers who want to reach people before they're actively searching — the consideration and awareness phase — rather than simply capturing existing demand.

In practice, we often run both simultaneously for clients. Performance Max handles the bottom-of-funnel harvest while Demand Gen builds the pipeline feeding into it. Treating them as competitors rather than complements is a strategic error.

Placements: Where Demand Gen Ads Actually Appear

Understanding placement behaviour is critical to setting realistic expectations and correct budgets.

• YouTube In-Feed: Ads appear in the YouTube homepage feed and alongside search results. Users see a thumbnail, headline, and description before choosing to click through to the video or your landing page. High intent relative to other display formats.
• YouTube In-Stream (Non-Skippable and Skippable): Standard video pre-roll and mid-roll formats. Reach is enormous but requires strong video creative in the first 3–5 seconds or you're paying for nothing.
• Google Discover: The algorithmically curated news and content feed on Android devices and the Google app. Users here are in a passive consumption mindset — scrolling, not searching. Visual-first, native-looking ads perform best.
• Gmail: Promoted placements at the top of the Promotions and Social tabs. Often dismissed, but Gmail placements convert surprisingly well for e-commerce and lead gen when paired with strong offer-led copy.

You can opt individual placements in or out at campaign level. We typically recommend starting with all placements enabled and then trimming based on actual performance data after 4–6 weeks — not before.

Campaign Setup: The Configuration Decisions That Actually Matter

Bidding Strategy

Demand Gen supports Maximise Clicks, Maximise Conversions, and Target CPA bidding. For most advertisers, the correct answer depends on where you are in the campaign lifecycle:

• Weeks 1–3: Use Maximise Clicks if you have limited conversion history, or Maximise Conversions if you have at least 30–50 conversion events per month on the account. Do not set a Target CPA constraint during the learning phase — it artificially restricts the algorithm's ability to gather the data it needs.
• After the learning phase: Introduce a Target CPA once you have statistically meaningful data. Set it at 20–30% above your actual CPA during the open phase before tightening.

One of the most common mistakes we see from advertisers coming to us after self-managing their accounts: they set a Target CPA on day one that's too aggressive, the campaign never exits the learning phase, and they declare Demand Gen "doesn't work." It does work — you're just strangling it before it can prove itself.

Audience Targeting

This is where Demand Gen genuinely differentiates itself from Performance Max. You have real audience controls:

• Custom Segments: Built from URLs your ideal customer visits, apps they use, or search terms they enter on Google. This is the closest thing Google offers to interest-based targeting outside of Meta.
• Customer Match Lists: Upload your CRM data. Use this to exclude existing customers from prospecting campaigns (crucial for e-commerce) or to create lookalike audiences from high-LTV customer cohorts.
• Lookalike Audiences: New to Demand Gen (absent in Discovery). Google builds an audience of users statistically similar to your seed list. Quality depends entirely on the quality of your seed audience — a lookalike of your top 500 customers outperforms a lookalike of your entire subscriber base every time.
• Life Events and In-Market Segments: Standard Google audience layers. Useful as secondary targeting signals but rarely sufficient on their own.

Our recommendation: stack Custom Segments with Customer Match lookalikes as your primary targeting combination. It's the closest Demand Gen gets to the precision targeting that made Facebook Ads compelling before iOS 14.

Creative Requirements and Best Practices

Demand Gen is a creative-led format. Technical setup is table stakes — creative quality is the actual performance variable. The format accepts:

• Single image ads (1.91:1, 1:1, and 9:16 aspect ratios)
• Carousel ads (2–10 cards, each with its own image and headline)
• Video ads (horizontal, vertical, and square formats)

Upload all variants. Google's system will test them automatically, but more importantly, it gives you actual comparative data on what's working. Don't upload one image and wonder why performance is flat.

On creative strategy: Demand Gen rewards thumb-stopping visuals over text-heavy layouts. Your first frame — whether image or video — needs to communicate the value proposition without the user reading a single word of copy. Headline and description copy matters, but it's the image or video frame that determines whether they stop scrolling.

Practically speaking, we see consistently stronger performance from ads that show a real outcome or result (before/after, product in use, specific number/stat) versus generic brand imagery. This mirrors what performs on Meta and TikTok — because the user psychology is identical. These are passive scroll environments. You're interrupting, not being found.

Budget Allocation and Scaling Logic

Demand Gen campaigns need sufficient budget to exit Google's learning phase and gather meaningful data. The general rule: set a daily budget of at least 10–15x your Target CPA. If you're targeting a £50 CPA, budget a minimum of £500–£750 per day before restricting the algorithm.

This is a hard conversation with clients who want to "test with a small budget." A small budget doesn't give you test data — it gives you noise. If full-scale budget isn't viable, it's often better to delay the campaign launch until it is, rather than burn three weeks getting statistically useless impressions.

Once a campaign exits the learning phase and is hitting CPA targets, scale budget in 20–25% increments every 7–10 days. Larger jumps reset the learning phase. Patience here pays off compoundly.

Measurement: What to Track and What to Ignore

Demand Gen operates in the consideration phase, which means last-click attribution will chronically undervalue it. This is not a flaw in the format — it's a flaw in how most advertisers measure upper-funnel activity.

The metrics that actually tell you whether Demand Gen is working:

• View-through conversions: Users who saw your ad but converted later via another channel. Set a reasonable view-through window (7 days is defensible for most categories).
• Assisted conversions in GA4: How often Demand Gen appears in the conversion path even when it doesn't get last-click credit.
• Brand search volume uplift: Run Search Impression Share reports for branded terms before and after Demand Gen campaigns go live. Genuine brand awareness drives branded search. If branded searches aren't growing, the campaign isn't building awareness effectively.
• New visitor rate: Demand Gen should be bringing new audiences to your site. If your new visitor rate isn't moving, you're over-indexing on retargeting audiences rather than prospecting.

This multi-touch measurement approach integrates naturally with the SEO and GEO work we do for clients — organic visibility and paid demand generation compound when they're targeting the same funnel stages and audience segments coherently.

Common Mistakes to Stop Making

Based on the accounts we audit regularly, these are the errors that kill Demand Gen performance most frequently:

1. Setting Target CPA too early and too tight. Already covered, but worth repeating because it's the most common.
2. Uploading only one creative variant. No creative diversity means no learning, no optimisation signal, and eventual creative fatigue.
3. Running Demand Gen with retargeting audiences only. It's a prospecting format. Running it exclusively against past visitors defeats its purpose and usually results in frequency capping issues.
4. Ignoring placement-level data. Some placements will outperform others on your account. Check placement reports after 4–6 weeks and make informed decisions.
5. Evaluating performance after one week. The learning phase alone takes 2–3 weeks. Campaigns need 6–8 weeks of data before drawing meaningful conclusions.

Who Should Be Using Demand Gen Right Now

Demand Gen campaigns are well-suited to:

• E-commerce brands with strong visual product catalogues and established conversion tracking
• SaaS companies looking to build pipeline at the top of funnel before capturing intent with Search
• Lead gen businesses in competitive verticals where Search CPCs have made bottom-of-funnel acquisition prohibitively expensive
• Brands entering new markets who need to build awareness before demand exists to capture

If you're a small advertiser spending under £3,000/month on Google Ads total, Demand Gen probably isn't where you should start — Search and Shopping campaigns will generate more attributable revenue at that scale. But if you're spending north of that and your Search campaigns are already optimised, Demand Gen is the logical next layer.

The paid media strategies we build at Workflow AI Advisors almost always include Demand Gen as a pipeline stage for clients once their account has the conversion volume to support it. The average ROAS lift we see when Demand Gen is added to an optimised Search and Shopping setup is meaningful — because you're no longer just harvesting existing demand, you're manufacturing new demand that flows into your lower-funnel campaigns.