A Demand Gen test that burns its budget with zero conversions has not proven the channel is dead. It has proven a cold, broad audience cannot learn on a B2B budget, and the fix is a smaller audience, not a bigger one.
Google Demand Gen for B2B usually fails the same way the first time. The test runs for a month, serves tens of thousands of impressions across YouTube, Gmail, Discover and the Display Network, and reports zero conversions.
The reflex is to make the audience bigger or the budget larger so the system has more room to find buyers. I think that is the wrong next step for almost every B2B advertiser.
The short answer: after a zero-conversion Demand Gen test, run the next flight against people who already know you (website visitors, YouTube engagers, a CRM list), plus a small set of placements your buyers actually read. Turn optimized targeting off, judge the flight on what known visitors do next, and write the kill rule before a dollar goes out.
This is for demand-gen and paid media leaders at mid-market and enterprise B2B companies who tried Demand Gen once, got nothing, and have to decide whether to try again.
Why this decision is coming for every B2B account
Even teams that never chose Demand Gen are about to be running it.
Google is folding its Display campaigns into Demand Gen. A migration tool started rolling out in June 2026, and Google's own timeline says that later, "New Google Display Ads campaigns can only be created within Demand Gen," with remaining campaigns moved over automatically (Google Display Ads campaigns have a new home in Demand Gen).
So the old B2B display remarketing campaign, the one quietly following pricing-page visitors around the web, ends up inside the same campaign type as the broad prospecting test that just failed. The settings that made that broad test fail come with it.
Ready for paid ads that pay off?
Book your free auditWhy a B2B Demand Gen test reports zero conversions
The budget was never big enough for the bidding to learn. That is usually the whole story, and Google says so in its own guide.
Google's Demand Gen performance guide says "Your budget should be at least 10 times your target CPA" and that the campaign "needs to generate at least 50 conversions" before the system understands what a good impression looks like (Demand Gen campaign performance guide).
Run that arithmetic on a B2B number. Say a qualified demo request costs you $200 from search, which is modest for a mid-market sale. Ten times that is a $2,000 daily budget, roughly $60,000 a month. Fifty conversions at $200 is $10,000 of spend just to finish learning, and that assumes cold feed traffic converts as well as a search click, which it does not.
Most first tests I see run on $3,000 to $5,000 for the month. That is $100 to $170 a day, which the 10x rule only supports at a $10 to $17 cost per conversion. Those are consumer numbers.
Then look at what the money bought. Take an illustrative month of 50,000 impressions, a few hundred clicks at a feed-style click rate, and a cold-traffic conversion rate around 1%. The expected result is two or three leads. Zero is well inside the noise. The test did not fail the channel. It never had enough data to say anything.
Your audience was a suggestion, not a fence
The second reason is that the audience you built was not the audience you bought.
Optimized targeting is on by default. Google's documentation says "Optimized targeting is automatically turned on for all campaigns," and for Demand Gen specifically, "Your ads may show to users beyond your selected signals." Demographic expansion switches on with it (About optimized targeting).
Lookalikes are going the same way. Through 2026, Google is moving Demand Gen Lookalike segments to a suggestion mode. "Your seed list will act as a signal for Google’s models," and the narrow, balanced and broad reach slider "will no longer function as a targeting constraint" (Use Lookalike segments to grow your audience).
For a consumer brand selling to millions, that expansion is the point. For a B2B company whose buyer pool in a country might be 20,000 people with a specific job, it means a carefully built audience of IT directors quietly turns into whoever the system finds cheapest to convert. On a small budget, cheapest usually means the least qualified.
Why "go broader" is the wrong next move
Google's own guide pushes the other way. For an ad group with no conversions for many days, it says "consolidating or broadening your targeting may improve performance," and it recommends running "across all channels unless there's a specific reason not to" (Demand Gen campaign performance guide).
That advice is built for advertisers who have the conversion volume to feed it. A B2B account with a $200 lead and a $150 daily budget does not. Broadening a campaign that cannot learn just gives it more places to spend without learning.
There is a specific reason not to, and B2B has it. The people who can buy from you are a small, identifiable group, and most of them have already shown up somewhere in your data.
What to run in the next flight
The next flight is a different test, not a bigger version of the last one. It asks a narrower question. Does showing up on YouTube and the Display Network move people who already know us closer to a sales conversation?
1.) Build the audience from your own data. Demand Gen supports website visitors, YouTube users and customer lists as "Your data" segments (Demand Gen audiences overview). Start with visitors to product, pricing and demo pages in the last 90 days, plus people who watched your videos. Exclude current customers using a CRM list, so the budget is not spent reselling to people who already pay.
2.) Turn optimized targeting off in every ad group. This is the single setting that turns a fenced remarketing flight back into broad prospecting. Check it again after launch, because Google switches it on for every new campaign unless someone deselects it.
3.) Add one small placement ad group. Demand Gen supports placement and topic targeting on the Display Network and YouTube channels. Pick the 20 or 30 trade publications, newsletters and YouTube channels your buyers actually consume. Your sales team can usually name them in five minutes. Google's placement documentation also warns that popular placements "may need higher bids" (About placement targeting), so avoid the big general-news sites.
4.) Choose channels on purpose. Channel controls live at the ad group level and let you pick YouTube in-stream, in-feed and Shorts, Discover, Gmail, Maps and the Display Network individually (Channel controls in Demand Gen campaigns). With a fenced audience of known visitors, leaving channels open is fine. On the placement ad group, keep it to the Display Network and YouTube, since that is where placements apply.
5.) Stop asking the bidding to optimize for a lead it will never see 50 of. A remarketing audience of a few thousand people is a frequency problem, not a conversion-efficiency problem. I would rather run a click-based or conversion-volume strategy with a sensible cap and judge the result downstream than set a Target CPA the campaign has no data to hit.
6.) Send the creative to a page that matches the question. People who already visited your pricing page do not need the homepage again. Send them to the comparison page, a customer story or the demo booking page, whichever answers the objection most likely stopping them.
How to judge the flight without fooling yourself
Do not let the Platform Comparable column be your verdict. Google built it so Demand Gen can be compared with social platforms. It counts view-through conversions and, in Google's words, "The last Demand Gen campaign in the conversion path gets the full credit" (About conversions (Platform Comparable) columns). That is a fair way to compare against LinkedIn or Meta reporting. It is not a pipeline number, and on a remarketing audience it will flatter the result, because those people were already on their way.
Judge a remarketing flight on what the people in the audience do next:
→ Return visits from the audience to pricing, demo and comparison pages, compared with the same period before the flight.
→ Branded search volume and branded search clicks over the flight window. Our post on B2B brand campaign measurement covers why branded search is a decent stand-in for a lift study.
→ Sales-accepted leads and opportunities in the CRM where the contact was in the remarketing audience before converting.
→ Frequency per person. Under two a week and nobody remembers you. Above eight a week and you are just annoying the same 3,000 people.
If the budget allows a cleaner answer, hold out part of the audience and compare. The method in our guide to incrementality testing for B2B paid media works here with a list split instead of a geography split.
Write the kill rule before you launch
The first test failed partly because nobody wrote down what success looked like, so "zero conversions" arrived with no agreed meaning. Do not repeat that.
Before the next flight goes live, put three numbers in writing:
→ The spend and time limit. For example, six weeks or $6,000, whichever comes first.
→ The pass line. For example, a 20% rise in return visits to pricing and demo pages from the audience, or at least two sales-accepted opportunities that touched the flight.
→ The action on a miss. The money goes back to non-brand search, where B2B intent is easiest to buy, and Demand Gen stays in remarketing-only mode. No third broad test until non-brand search stops losing impression share to budget.
Sign it off with whoever approves the budget. The point is that the next result, good or bad, produces a decision rather than another debate.
When broad Demand Gen does make sense for B2B
There is one B2B profile where I would run broad Demand Gen prospecting early, and that is low-price, self-serve software where the conversion is a free signup that costs $20 to $40. At that cost the 10x budget rule is affordable, the campaign can reach 50 conversions in a few weeks, and optimized targeting has enough signal to find people.
Everyone selling a five or six figure contract through a sales team should treat Demand Gen as a remarketing and placement channel first. Prospecting can come later, once search is fully funded and there is a qualified conversion volume to train on. Our B2B paid channel selection framework covers how to size that order, and if video is the creative you are bringing, the cut list for B2B video ads covers the formats Demand Gen asks for.
Before you approve the next Demand Gen budget
Open the failed campaign and check four things: whether optimized targeting was on, what the daily budget was divided by your real cost per qualified lead, which channels took the impressions, and whether any of the spend reached people who had already visited your site.
If optimized targeting was on, the budget covered fewer than ten qualified leads a day, Discover and Display took most impressions and almost none of it reached past visitors, then the test measured a setup that could not work at your budget. Rebuild it as a fenced remarketing flight with a placement ad group and a written kill rule.
If you want a second set of eyes on a Demand Gen or display setup before the migration forces your hand, that is part of how we run B2B Google Ads for our clients.
