Find out why your Demand Gen campaign is failing before you cut it
Share this Article
Key takeaways
Demand Gen rarely fails as a channel; structure, signal, and scorecard fail.
One ad group cannot teach Google anything; build prospecting, lookalikes, remarketing pillars.
Budget at least twenty times target CPA, or pick a shallower conversion.
Image ads win CTR and lose CPA; video drove ninety percent of conversions.
Last-click hides Demand Gen's impact; track branded search impressions and impression share.
Google Demand Gen campaigns usually fail for one of three reasons: the account structure is too thin, the bidding signal is too weak, or the team judges the campaign on last-click CPA inside six weeks. The channel itself is rarely the problem.
We know this because we watched it fail twice for the same client. Fyxer's first two Demand Gen tests spent about £23k and produced 7 conversions. The rebuild used the same product, the same channel, and the same budget owner. It produced 2,226 conversions and got monthly CPA down to £207. The full story is in our Fyxer Demand Gen case study.
If you are a founder or a paid media lead asking "should I kill Demand Gen?", find your failure below first.
Why do Demand Gen campaigns fail? Quick diagnosis
Match your symptom to the likely cause before you change anything.
[table]
Symptom in the account | Likely cause | First thing to check
CPA 5x to 10x target from day one | No audience architecture | Number of ad groups and audience signals
Campaign barely spends | Budget too small for the bid goal | Daily budget vs target CPA
High CTR, almost no signups | Click bidding or image-heavy creative | Bid strategy and format split
Lots of signups, sales calls them junk | Missing exclusions | Customer, age and income exclusions
Lookalikes underperform prospecting | Weak seed list | What the seed list is built from
CPA dropping, then the campaign got paused | Killed during learning | Change history and spend before the cut
Relaunch never recovers | Too many changes at once | Change history in the relaunch week
CPA looks bad but branded search is rising | Last-click reporting | Brand impressions and impression share
[/table]
1. You launched one ad group and called it a test
A single ad group with mixed signals tells you nothing. Fyxer's first attempt ran one ad group with no audience architecture and posted a £1,804 CPA. The second was a video-only test, also one ad group, at £3,865.
The rebuild used three pillars per market, each in its own ad group:
- Prospecting: affinity, competitor custom intent, competitor landing pages, and non-branded search themes.
- Lookalikes: seeded from first-party customer lists.
- Remarketing: 7-, 30-, and 90-day site visitors, 180-day YouTube viewers, and branded search themes.
Separate pillars let you read, budget, and cut each one on its own. A blended ad group hides which signal is working.
2. Your budget can't feed the conversion you picked
Demand Gen needs volume to learn. Google recommends a daily budget of at least 20x your target CPA for conversion bidding and says ad groups with 50+ conversions perform better.
Run that math on a demo booking. A $400 demo CPA needs $8,000 a day. Most Series A budgets can't support that, so the campaign starves, and the founder concludes "Demand Gen doesn't work."
Optimise to the deepest event that still gives you enough volume. We optimised Fyxer’s Work Email Signup at roughly 400 a month. For B2B, a work-email signup or a qualified signup (filtered by company size) is usually the right trade.
3. You bid for clicks or judged the campaign on CTR
Clicks on Demand Gen are the cheapest and least useful thing Google can sell you. Fyxer's desktop-only Max Clicks test spent about £7k for 4 conversions.
CTR misleads too. Fyxer's image ads had the highest CTR in the account and the worst CPA. Two GB image sets drew 178k clicks for 54 conversions. Discover, Gmail, and Shorts placements produce plenty of accidental taps. Use conversion-based bidding from day one and treat CTR as a creative diagnostic, not a success metric.
4. You ran image ads like it was Display
Video does the selling on Demand Gen. At Fyxer, video took 69% of spend and delivered 90% of conversions at a £280 CPA. Image ads ran at £1,070.
What worked:
- A mid-funnel product explainer, the single highest-volume creative.
- A paid UGC workplace skit at £250 to £340 CPA.
- Short cuts. The 30-second versions were 82% cheaper per signup than the full-length originals.
If you only have static assets, Demand Gen is not ready to launch. Build three to five short, problem-first videos before you spend.
5. You didn't exclude the people who shouldn't see it
Without exclusions, Smart Bidding finds the easiest conversions in your account, which are often existing users, students, and personal Gmail signups. For a freemium AI product, this can make CPA look great while pipeline stays flat.
For Fyxer, we excluded existing paid users and predicted zero-LTV signups everywhere, limited age to 25 to 44, and removed the bottom 50% of household income. Pull your Demand Gen signups from the CRM and check the email domain and company size. If sales would reject most of them, fix exclusions before you touch bids.
6. Your lookalike seeds are weak
A lookalike is only as good as its seed. GB Lookalikes was Fyxer's best ad group at £126 CPA. US Lookalikes ran on the same logic at £496, roughly 4x worse. The structure was the same, so seed quality by region is the open question.
Seed from your best customers, such as paying users or a predicted-LTV list, not from every signup. Seeding from all signups teaches Google to find more free users.
7. You killed it during learning
Demand Gen's first month is expensive, and most programmes die there. Fyxer's GB campaign posted a £1,829 CPA in month one, and US posted £730. Roughly £165k went in before efficiency arrived. Then CPA fell every month with no goal changes.
Decide three things before launch: the learning budget you can afford to lose, the date you will judge it, and the metric you will judge it on. Google's Demand Gen performance guide advises against changes during learning and asks for 30 to 50 conversions before you evaluate a change.
8. You changed four things at once
The fastest way to break a working Demand Gen campaign is a "refresh" that changes everything. Fyxer's spring relaunch added a new ad group, new creative, a Max Clicks test, and a conversion-goal migration together. CPMs doubled from £10 to £19, and CPA hit £1,776. Nobody could isolate the cause.
Google calls a conversion-goal change on a live campaign the biggest reset to learnings. Change one variable, wait for 30 to 50 conversions, then decide. When creative fatigues, refresh it inside proven audiences. Fyxer's rebrand videos in GB Lookalikes hit £119 CPA.
9. You judged it on last-click
Demand Gen creates demand that Search closes. Engaged-view and view-through conversions carry no GCLID, so your offline conversion upload credits branded and non-branded Search.
At Fyxer, branded search impressions rose 2.7x during the flight. Brand impression share fell from 41% to 62% because demand outgrew the brand budget. When Demand Gen was switched off, branded impressions dropped to 51k, below the pre-flight baseline.
At the time, we also missed two things, and we would do both next time:
- Design measurement before launch. A geo holdout in one market would have made the brand halo causal, not correlation.
- Close the loop to Search. Add Demand Gen's YouTube viewer lists to Search as observation audiences.
Also add the "Conversions (Platform Comparable)" column when you compare Demand Gen to Meta or LinkedIn. It counts engaged views the way those platforms do.
When Demand Gen is the wrong channel
Sometimes the campaign is fine, and the fit is wrong. Demand Gen is a poor first choice when:
- Search isn't saturated yet. Fyxer only moved to Demand Gen after Search had covered every worthwhile keyword.
- Your buyer is a narrow job title. Selling SOC 2 compliance to startup founders and security leads is a LinkedIn targeting problem. Our Oneleet LinkedIn Ads case study shows that motion driving $1.63M in influenced closed-won revenue.
- You can't fund learning. Google suggests at least $500 a day per ad group for efficient performance. Two pillars at that level are $30k a month before you know anything.
For placements, formats, and how Demand Gen replaced Video Action campaigns, read our Demand Gen and YouTube ads playbook.
Get a second opinion before you cut the budget
Most Demand Gen campaigns we audit aren't broken channels. They're missing pillars, running on weak seeds, or being judged on a report that can't see what they do.
We'll review your structure, conversion goal, exclusions, creative split, and measurement, and tell you whether to fix, pause, or move the budget. If you're still comparing partners, see our list of top Google Ads agencies in the US and what clients say in our ScalixAI review.
