In short: automated campaigns promise results without work, and sometimes deliver — but they have their own expensive ways of failing: site farms that manufacture clicks, gaming apps where taps are accidental, PMax eating your own brand and view-through conversions that make everything look profitable. Automated does not mean unattended.
Automated campaigns (Display with broad targeting, Performance Max, Demand Gen) promise results without effort. Sometimes they deliver. But they share one feature that makes them dangerous for beginners: they spend in places you cannot see. These are the mistakes we have seen lose the most money, and they are among the hardest to discover because aggregate reports hide them.
Mistake 1 · What are site farms on Display?
Case: a services company's Display campaign with 8,000 clicks a month at €0.04 each. Conversion rate: 0%. The placement report showed hundreds of domains with generic names, copied content and dozens of ads per page; several shared the same IP and the same publisher ID. They were farms: websites built to charge for impressions and clicks, often with non-human traffic. The same network of farms turned up in three other clients' accounts.
How to spot it: the "Where ads showed" report sorted by clicks: a laughable CPC, an anomalous CTR, zero conversions, domains you do not recognise. Visit them: if the site makes no sense, it is a farm.
How to fix it: exclude placements (at account level, which also applies to PMax) and, better still, do not target Display "at everybody": specific audiences, excluded content categories, and a weekly placement review. It is never-ending work: farms create new domains every week.
Mistake 2 · Why exclude mobile apps?
Case: a B2B lead generation campaign with 60% of its Display spend going to mobile apps — free children's games, torch apps, wallpaper apps. Accidental clicks on an ad for business management software. Zero leads.
How to spot it: a placement report with "mobileapp::" rows or app names; segment by device (disproportionately mobile) and by placement type.
How to fix it: exclude the app categories or, in many businesses, all apps (an app placement exclusion at campaign or account level). App inventory is legitimate for promoting apps and not much else.
Mistake 3 · Is Performance Max eating my own brand?
Case: an ecommerce launched Performance Max and its ROAS "rose" to 9. At the same time the brand Search campaign lost 60% of its conversions. PMax had claimed the brand searches, which used to convert just as well and cost less. The account's total ROAS had not changed; only how it was split in the report.
How to spot it: the brand campaign dropping at the same time PMax launched; PMax search terms containing your name.
How to fix it: brand exclusions in PMax (a campaign setting), and judge PMax by the total incremental gain for the account, not by its own ROAS.
Mistake 4 · What are view-through conversions and why are they misleading?
Case: a Display report boasting "300 conversions a month". They were view-through conversions: people who saw (did not click) a banner and bought later through some other channel. Measuring only click conversions: 6.
How to fix it: judge Display and video with click conversions and, if you do use view-through ones, with very short windows and knowing they are only indicative. And never in the headline column of a report for the board.
Mistake 5 · Do automated campaigns need watching?
The common thread through the previous four. "Automatic" does not mean "unsupervised": it means Google decides within the limits you set, and that if you do not set limits, Google's own interest sets them (more reach, more spend). Placement and brand exclusions, app exclusion lists, clean conversions and reviewing placement reports are the minimum limits.
How do you audit an automated campaign in five minutes?
- Display/PMax placement report by clicks: do you recognise the domains? → Mistake 1.
- Are there apps in the placements? → Mistake 2.
- Did brand drop when PMax launched? Are brand exclusions in place? → Mistake 3.
- Which conversion column does the Display report use? → Mistake 4.
💡 Ninja trick: these mistakes are the reason Ninja Shield exists. Every night it analyses the sites and apps where your ads show (content, ad density, trust signals such as a CMP, a trust seal or a company tax ID, domain age, shared IP and publisher), detects the farms as networks — when one domain falls, its siblings fall with it, even if they have not yet spent a single euro in your account — and excludes them at account level, where PMax respects it. What a manager cannot review by hand every week, the script reviews every day.
What you should remember
- Display and PMax spend where you are not looking: the placement report is compulsory.
- Farms and apps: laughable CPC + zero conversions = exclude.
- PMax steals the brand unless you forbid it: brand exclusions, and judge by the total.
- Click conversions to judge; view-through, with tongs.
- Automatic ≠ unsupervised: set the limits or Google sets them.