AI optimises what it is asked to optimise, not what was meant. Six failures keep recurring in automated campaigns: a poorly defined objective, wrong conversions, inflated brand traffic, low-margin products, saturated audiences and platform metrics with no counterpart in the business.
The essentials in a definition
Automated campaigns optimise what they are asked to, not what the business wants. The six most typical failures are: measuring the wrong event, cannibalising brand, ignoring margin, overloading frequency, using weak assets and serving irrelevant terms. Performance Max and other automated campaigns demand reliable measurement and regular review.
Failures 1 and 2: objective and conversion
Asking for “more conversions” without defining quality leads to cheap, not very useful enquiries. Counting as a conversion something that is not a real sale or lead makes the algorithm look for traffic that triggers the event but does not generate revenue.
The fix is to define an objective tied to value (revenue, margin, qualified lead) and validate the event that measures it.
Failures 3 and 4: brand and margin
The algorithm quickly discovers that brand searches convert better and channels budget there. Reports look excellent, but much of the result would have come anyway. In shops, it concentrates on products that sell easily, even if margin is poor.
The fix is to separate brand from non-brand and to tell the system about margin or value per product.
Six failures and what to do
- Define an objective tied to value
- Validate the conversion being measured
- Separate brand from non-brand
- Share margin and refresh audiences
Failures 5 and 6: saturation and empty metrics
Small or repeated audiences saturate: frequency rises, costs rise and results fall. On the other hand, metrics such as clicks, impressions or platform conversions can improve without revenue following.
The fix is to refresh audiences and creative, and always to evaluate with business data (sales, margin, new customers).
When the metric rises and the business does not, the metric is measuring something else.
How to spot them early
Compare platform versus business every week. If they diverge, look for the cause. Analyse the distribution of spend by product, term and audience. Keep a simple alert list: concentrated spend, brand above a given weight, margin drop.
Early detection does not need expensive tools: it needs habit and fixed questions.
Decision table
| Failure | Warning sign | Remedy |
|---|---|---|
| Wrong event | Conversions rise, sales do not | Redefine the primary conversion |
| Cannibalising brand | Large share of spend on brand terms | Brand exclusions and a separate campaign |
| Ignoring margin | High ROAS, low profit | Margin-based value |
| Weak or repeated assets | Low CTR and restricted delivery | Renew assets and vary angles |
Common mistakes to avoid
- Accepting aggregate reports as proof. Always confirm in the sales system.
- Letting irrelevant terms accumulate. Review search categories and exclude.
- Not testing anything. Without tests, you do not know whether automation helps or only seems to.
How to measure the result
For each failure, define an alarm indicator: share of spend on brand, margin per order, average frequency and share of irrelevant terms. Review them weekly.
Limits and exceptions
These six failures do not exhaust possible problems, and not all apply to every account. Some campaigns work well without much supervision because they have excellent data and a simple offer; others require tight control. Adapt the checklist to your case, add the risks specific to your sector and review it when you change strategy, platform or offer, instead of applying it mechanically.
Illustrative example
Illustrative example: a shop sees an automated campaign’s ROAS rise while profit falls. On analysis, it finds the platform favours cheap products that convert easily. It starts segmenting the catalogue by margin and sending margin value; ROAS dips slightly, but profit rises. The review avoided celebrating a misleading number. Respecting ad policies also avoids disapprovals.
Frequently asked questions
Are these failures common?
Yes, especially in accounts that moved to automated campaigns without reviewing objectives and measurement.
How do I separate brand?
With distinct campaigns for brand and non-brand searches, and separate reports.
Do I need margin data?
It helps a lot. Even an average margin per category improves decisions.
Which failure is the most expensive?
Optimising for the wrong event, because everything else, including budget and creatives, will be steered towards it.
Should I avoid automated campaigns?
No. They work well with reliable data and supervision; the problem is using them without measurement or review.
The most common failures of automated campaigns come from measuring the wrong event, ignoring margin and brand and reviewing too little; each has an alarm indicator and a simple remedy.
Sources and recommended reading
- Google Ads Help: Performance Max campaigns
- Google Ads Help: conversion measurement
- Google Ads policies
Last editorial review: October 2026. Informational content; always confirm current conditions in the official sources before deciding.