Automated bidding in Google Ads: when to use smart bid strategies
Automated bidding isn't a 'turn on magic' button — it's a tool that needs data to function and often underperforms manual bidding if switched on too early.
How machine-learned bidding works
Automated strategies calculate an optimal bid for every auction in real time, factoring in dozens of signals — device, time of day, location, user behaviour history — that manual bidding physically cannot process at that volume.
That strength is also the weakness: the algorithm needs a volume of data to learn from, and without enough conversion history it either can't launch properly or acts almost randomly, spending budget on trial with no formed model.
The learning period after launch or a significant strategy change usually runs one to two weeks, during which results are unstable — a normal part of the process, not a sign something is broken.
Strategy types and when to apply each
'Maximise clicks' is the simplest automated strategy, suited to new campaigns with no accumulated conversions, when the early goal is gathering traffic volume and statistics for further optimisation rather than a specific conversion.
'Maximise conversions' shifts focus from clicks to conversions and needs a minimum of 15-30 conversions over the last 30 days for stable operation — below that threshold the algorithm lacks enough signal for a reliable forecast.
'Target CPA' and 'target ROAS' are the most data-demanding strategies, sensible only after accumulating 30-50 conversions and understanding the business's real economics — what lead cost or ad return the business can actually afford.
When it's too early to automate
A new account with no conversion history, where tracking has only just been set up, is the worst candidate for automated bidding — the system literally has no data to learn from, and the first weeks become unpredictable spend.
Low-demand-volume niches — narrow B2B, rare services — also fit poorly with CPA-based automation in the early months, since accumulating 30 conversions can take too long at low traffic.
A seasonal business at peak season, when audience behaviour deviates sharply from normal, isn't the best time to switch strategy, since an algorithm trained on off-season data doesn't immediately adapt to the demand spike.
Common mistakes with automated bidding
Changing the target CPA by more than 15-20% at once knocks the algorithm off its learning trajectory and sends the campaign into a new phase of instability — edits should be made gradually with pauses for adaptation.
Switching between strategies weekly in search of the best one is a guaranteed way to never let any single strategy reach a stable result, since every switch restarts learning from scratch.
Ignoring audience signals and remarketing lists while running an automated strategy — the system uses that data as additional context, and its absence lowers forecast accuracy even when base conversion tracking is set up correctly.
A hybrid approach as a sensible compromise
Start with manual bidding or maximise clicks, accumulate conversions over a month or two, then move to maximise conversions, and only once results stabilise, to target CPA or ROAS. Skipping steps almost always performs worse than a gradual transition.
Keep some new or experimental campaigns on manual bidding even after moving the core account to automation — it gives a control point for comparison and hedges against the algorithm suddenly losing effectiveness when the market shifts.
Revisit the strategy choice quarterly based on actual data rather than sticking with a once-chosen setting forever — growth in conversion volume or a change in business economics can make a more sophisticated automated strategy viable and worthwhile where it was premature before.