Money mistakes have something in common: nearly all of them are made by somebody trying to save money. And nearly all of them end up costing more than the problem they were trying to avoid. Five cases.
Mistake 1 · The wished-for target
Case: a law firm with a real CPA of €95 per lead set a target CPA of €40 "because that is what we can afford". The campaign went from 40 leads a month to 4. The client's conclusion: "Smart Bidding is useless". The reality: Google only entered the auctions it estimated at €40 or less — which was almost none.
How to spot it: spend far below budget after changing a target; permanently "learning"; a target a long way from the real CPA of the previous 30 days.
How to fix it: target = real −10-20% (€85), lowered in steps every 3-4 weeks while you improve what actually brings the CPA down (search terms, ads, landing page). If the business cannot take €95 per lead, the problem is not the bid: it is the close rate or the product.
Mistake 2 · Cutting budget from the campaign that works
Case: an ecommerce with two campaigns: "Catalogue" (spends €1,500/month, ROAS 5) and "New arrivals" (spends €400/month, ROAS 1.8). To "balance things out", the manager cut Catalogue to €900 and raised New arrivals to €1,000. The result: total sales fell by 30% on the same spend.
How to spot it: budget changes moving in the opposite direction to ROAS or CPA. The change history next to the chart shows it.
How to fix it: money goes where the marginal return is better: raise Catalogue while it stays limited by budget and the ROAS holds; New arrivals gets diagnosed (product? ads? landing page?) before you water it.
Mistake 3 · Itchy hands
Case: an account with 47 target and budget changes in 60 days (change history). The manager reacted every morning to the previous day's figure. The strategy never came out of "learning"; the CPA was 40% worse than in an earlier period when nothing was touched.
How to spot it: a change history with almost daily bid/budget modifications; strategies permanently in learning.
How to fix it: the change rule from Module 5 — one a week, small, written down, evaluated over two full weeks. And take away edit access if you have to: some people need not to be able to touch.
Mistake 4 · Recommendations applied in bulk
Case: a client switched on "apply automatically" in every Recommendations category to push their "optimisation score" to 100%. Within three weeks: 200 new broad match keywords, three budgets raised by 30%, Display expansion turned on and "automatically created assets" with copy the brand had not approved. Spend +55%, conversions +4%.
How to spot it: a change history with "Google Ads" or "recommendation applied automatically" as the author; a suspiciously high optimisation score; keywords nobody recognises.
How to fix it: switch off automatic application in every category, revert what was applied from the history, and treat Recommendations as reading material (Module 7).
Mistake 5 · The Display expansion nobody turned on
Case: a Search campaign for an accountancy firm with a CTR of 0.3% and expensive conversions. Segmenting by network: 45% of the spend was going to the Display Network through the expansion checkbox, ticked by default when the campaign was created. Nobody knew. The Search conversions, on their own, were profitable.
How to spot it: segment by network. Any Search campaign with "Display Network" rows has expansion switched on.
How to fix it: untick the expansion (Module 2), and if you do want Display, put it in its own campaign with placement monitoring.
The five-minute money audit
- Current target vs real 30-day CPA/ROAS → Mistake 1.
- Recent budget changes vs each campaign's performance → Mistake 2.
- Number of bid/budget changes in the last 60 days of history → Mistake 3.
- Automatic application of recommendations: on? → Mistake 4.
- Network segment on Search campaigns → Mistake 5.
💡 Ninja trick: all five mistakes are decisions made by eye. The Suite's Smart Bidding (SBNS) replaces them with rules: it judges each campaign against its real business objective, moves budgets towards where the marginal return is better, changes targets in small, spaced-out steps, and leaves a record of everything. Guardian, for its part, flags anomalous spend the same day. If you have itchy hands, you had better have a script.
What you should remember
- The target is set from the real figure, not the wished-for one.
- Money follows the marginal return: you feed what works.
- One change a week; the change history gives you away.
- Recommendations: never on automatic.
- Segment by network: Display expansion comes ticked as standard.