Almost all the damage done to a Google Ads account is done by somebody with good intentions inside the interface: by touching things. This lesson closes the module with the opposite of a tutorial: the list of things not to do, why, and the rule that prevents nearly all of them.
The ten mistakes
1. Changing the target (CPA/ROAS) every few days
Every change restarts the learning. A campaign touched twice a week lives in «learning» and never performs. Rule: one change of target, of 10-20% at most, and 2-3 weeks of waiting.
2. Cutting the budget of the campaign «that spends a lot»
It spends a lot because it works (Google gives money to what converts). Cutting it reduces conversions at the same CPA. Look at the CPA before the spend: if it's good, the campaign that spends a lot is the one to feed.
3. Raising the budget of the one that «won't take off»
If it doesn't convert on €20, it won't convert on €60: you will just buy three times more of the same. Diagnosis first (search terms, ad, landing page, tracking); money afterwards.
4. Pausing at weekends «because they don't convert»
Smart Bidding already bids less when the probability is low. Pausing breaks the continuity and the learning, and often the data was false (people search on Sunday and buy on Monday: the conversion is attributed to Sunday's click). If you genuinely don't want weekends, use the ad schedule, not the pause button.
5. Pausing and reactivating campaigns like a switch
Every long pause restarts part of the learning and loses auction history. A campaign is paused to stop existing, not to «save a few days».
6. Device/time bid adjustments with Smart Bidding
They do nothing (except -100%): the algorithm already bids on those signals. Touching them gives a feeling of control and zero effect. With manual bidding they do work.
7. Accepting Google's recommendations wholesale
The Recommendations tab (and the «optimisation score») mixes useful suggestions with others that raise your spend or widen your reach without asking (adding broad match, raising budgets, switching expansions on). Read every one and accept only what you understand. The optimisation score is not a mark for your account: it is a list of things Google would like you to do. And check that auto-apply recommendations is switched off.
8. Aggressive automated rules
«If the CPA goes above X, lower the bid by 30%» looks prudent and is a chaos generator: it reacts to daily noise, fights against Smart Bidding and restarts learning periods. Automated rules are for alerting and for safety actions (pausing if daily spend goes haywire because of an error), not for optimising bids.
9. Judging with short time windows
A bad day is not a trend. Nor are three days, with few conversions. And conversions arrive late (today's click converts next week): the last 3-7 days always look worse than they will end up being. Compare complete weeks against complete weeks, and exclude the most recent days from the analysis.
10. Changing several things at once
Target + budget + ads + negatives on the same day: if it gets better or worse, you will never know why. One change, one measurement.
The change rule
Nearly all the mistakes above disappear with a single discipline:
One relevant change per campaign per week, small in size (10-20%), noted down with the date and the reason, evaluated over at least two complete weeks and excluding the last 3-7 days.
Keep a change log (a spreadsheet will do): date, campaign, what was touched, from what to what, why, and a column to note the result three weeks later. It is the cheapest learning tool there is, and the one that separates people who manage from people who fiddle.
What you should do every week
So that «don't touch» doesn't turn into «don't look»:
- Search terms report and negatives (Module 3).
- Spend against plan: is any campaign limited by budget and profitable? Is any spending without converting?
- Conversions arriving normally (if they drop to zero all at once, it is almost always broken tracking, not the market).
- Ads: weak pieces with volume (Module 4).
- Google alerts: ad disapprovals, billing problems.
💡 Ninja trick: the change rule is hard to keep by hand because the temptation to touch lives in the interface. Scripts have no temptations: Guardian watches spend and anomalies every day without touching anything and emails you only when there is something real; Smart Bidding (SBNS) applies gradual, spaced-out budget and target changes, with an automatic record of every move; and both respect the learning period. Automating is not giving up control: it is putting discipline into it.
What you should remember
- The damage is done by whoever touches without data: targets, budgets and pauses by eye.
- High spend with a good CPA = feed it; spend without conversions = diagnose, don't water.
- Recommendations and automated rules: read, don't accept wholesale; alert, don't optimise.
- The rule: one change, small, per week, noted down, evaluated over two complete weeks.
End of Module 5. In Module 6 comes the mother of all levers: conversion tracking — what to count, how to install the tag and consent.