Est. reading time: 7 minutes
Both campaigns land in the same table, sorted by the same ROAS column, and Demand Gen loses that comparison in nearly every account we open. The budget decision follows within a week. Spend shifts into Performance Max, account-level ROAS ticks up, and total revenue sits where it already was.
The Demand Gen campaign is not underperforming in most of those accounts. It is being graded on a scorecard built for a campaign type that has access to the search query.
These two campaign types buy different inventory, count conversions off different events, and support different measurement tools. Below is the diagnosis we run before either one gets a budget change, and the reporting each has to be judged against.
Demand Gen and Performance Max are not buying the same inventory
Performance Max is a goal-based campaign type that reaches all Google Ads inventory, including YouTube, Display, Search, Discover, Gmail, and Maps, from one campaign. Bidding is limited to conversions or conversion value. There is no hard audience targeting. Audience signals guide the learning, and the campaign-level controls available to you are keywordless targeting, final URL expansion, search themes, negative keywords, and brand exclusions.
Demand Gen serves on YouTube including Shorts, Discover, Gmail, Maps, and the Google Display Network. It supports clicks as a bidding objective alongside conversions and conversion value, it lets you target specific audience segments including Lookalike segments, and its channel controls let you choose which surfaces run. Google Display Ads campaigns are also moving into Demand Gen as the Google Display Network, with a voluntary migration tool available to eligible advertisers since June 2026 and remaining eligible campaigns set to migrate automatically later.
One difference carries most of the argument. Search sits inside Performance Max and sits outside Demand Gen. A campaign with access to the query stream can convert people who have already decided to buy. A campaign restricted to feeds and video has to produce that decision first, then wait for it to surface somewhere else in the account.
Performance Max ROAS reads high because Search is inside the campaign
Google’s own prioritization rules describe the overlap. Search campaigns containing an exact match keyword are prioritized to serve ahead of Performance Max when a query hits that keyword, and search themes carry the same priority as phrase and broad match keywords. If your Search campaign is limited by budget, Performance Max may occasionally serve on exact match terms.
Brand queries are where this gets expensive. Brand exclusions and negative keywords exist as campaign-level controls for exactly this reason, to prevent overlaps and block unwanted queries. An account that never applied either one is reporting a Performance Max ROAS built partly on people typing the brand name into Google.
The channel performance report settles the question. Find it under Campaigns, then the Insights and reports menu, then channel performance. Segmenting by channel splits the campaign across Google Search, YouTube, Discover, Gmail, Maps, the Google Display Network, and Search partners, with conversions, conversion value, and cost reported for each. Data is available for date ranges after June 6, 2025.
Read that split before you compare the campaign’s ROAS to anything. If Search carries most of the conversions, apply brand exclusions and negative keywords, then read the report again.
The overlap runs past Search. Both campaign types serve on YouTube, Discover, Gmail, and the Google Display Network, and Performance Max gives you no direct control over how its budget is allocated across them. The inputs you can move there are asset quality, audience signals, conversion tracking accuracy, search themes, and exclusions. Demand Gen does have channel controls, so once the channel report shows where Performance Max has concentrated, point Demand Gen at the surfaces it is underserving instead of funding two campaigns into the same in-feed inventory.
The conversions column counts different events in each campaign
Demand Gen has a view-through conversion optimization setting currently in open beta, supported on YouTube and the Discover feed (video formats only on Discover). It is off by default on new and existing campaigns. Turning it on moves view-through conversions into the conversions column as a biddable conversion.
Two consequences follow. Your ecommerce backend will not confirm the increase, because impression-based metrics like view-through and engaged-view conversions are not exported. And a Performance Max campaign sitting in the next row is counting a different mix of events, so the two ROAS figures stop being comparable.
Google records each conversion once, against the highest-ranking ad event type available.
- Clicks produce click-through conversions, the same event most reporting outside Google Ads counts.
- Engaged views produce engaged-view conversions, triggered when someone watches for more than ten seconds and then converts without clicking.
- Impressions produce view-through conversions, triggered when a view precedes the conversion and no click does, inside a window Google recommends setting to one day for primary view-through conversions.
Segment the report by Conversions, then by Ad event type, before any cross-campaign comparison. Impression rows are view-through, engaged-view rows are engaged-view, and click rows are click-through. Compare click-through against click-through first, then decide separately what a view is worth to your business.
The thresholds underneath those metrics also move by surface. In Demand Gen reporting, an engagement on YouTube in-stream requires ten seconds of watch time, while YouTube feeds, Shorts, and Discover video count an engagement at five seconds.
Grade Demand Gen with the tests it is eligible for
Demand Gen supports A/B experiments in two forms. Custom experiments compare audiences, bidding strategies, formats, or creative across as many as ten arms. Asset A/B tests isolate creative as a single variable. You choose one success metric from clickthrough rate, conversion rate, cost per conversion, or cost per click, and a confidence level of 70%, 80%, or 95%. Google recommends a 50% traffic split.
The volume requirement decides whether the test is worth starting. Using conversion-based bidding, Demand Gen experiments need at least fifty conversions per arm to surface results, and conversion-related metrics need at least a hundred data points before any result appears. Google’s own guidance is to optimize the experiment toward a shallow event such as add to cart or page view for that reason.
For the incrementality question, the lift options differ by campaign type. Conversion Lift based on users is available in Google Ads for Video, Discovery, and Demand Gen campaigns, while running it on Performance Max means going through a Google account representative. Conversion Lift based on geography covers both campaign types, supports online and offline conversions, and trades a noisier group split and no demographic breakouts for that coverage. Brand Lift is available for Demand Gen campaigns running on YouTube. Search Lift covers Demand Gen across YouTube Shorts, in-feed, home feed, watch next, YouTube Search, Discover, and Gmail.
That last one deserves attention in accounts running both. Search Lift measures how much more likely your audience is to search for the brand after seeing the ads, which is the exact mechanism feeding the branded queries Performance Max keeps converting.
Setting budget when both campaigns run in the same account
Budget floors come first. For Demand Gen campaigns on target CPA bidding, Google recommends a daily budget of at least ten times your target CPA. Both campaign types also have a budget panel on the campaign review page that returns up to three daily budget levels with performance forecasts attached, which is a reasonable check on a number that was set by habit.
Then there is the floor nobody sets. Fund Demand Gen at a level where its own experiment framework can return an answer inside a quarter. Below fifty conversions per arm, the campaign cannot be tested, and an untestable campaign eventually loses a ROAS-column argument to Performance Max. That is the sequence most of these accounts repeat.
After significant edits to budget, bidding, targeting, or assets, give the account time. Google’s guidance for Performance Max is one to two weeks of stabilization, since frequent adjustments can push a campaign back into learning or stop it serving temporarily.
Both campaign types can earn their spend in the same account. Neither number means much until the channel split is clean, the ad event types are separated, and the incremental question has been asked with a study rather than a column sort. That audit is an afternoon of work, and it changes what every number in the account means. If you want it run alongside you, get in touch.










