A safe path from manual CPC bidding to Smart Bidding: tracking checks, conversion thresholds, realistic first targets, protecting the learning phase and when to roll back.
Moving from manual cost per click (CPC) bidding to Smart Bidding, where Google sets the bid for each auction, is one of the biggest changes you can make to a Google Ads account. It is also often botched: a team flips the switch, sets an ambitious target, sees costs jump in the first week, makes three more changes, switches back and concludes that automated bidding doesn't work.
The safe route is simple, if not quick. Fix your conversion tracking first, make sure each campaign has enough conversions to learn from, start with a target close to what you achieve today, then leave the campaign alone through its learning period. Better still, run the switch as a 50/50 experiment against your current setup, so you compare like with like.
Getting this wrong costs money twice: in the weeks of poor results after a rushed switch, and in the efficiency you give up if you retreat to manual bids for good. This article covers the preparation, choosing a strategy and a first target, running the switch as a test, protecting the learning phase, and clear rules for when to tighten or roll back.
Smart Bidding can only be as good as the conversions you feed it. If your tracking double-counts form submissions, treats newsletter sign-ups as leads or misses phone enquiries, the system will faithfully optimise towards the wrong thing. Fix this before you switch, not after.
None of this is glamorous, but a week spent here saves months of optimising towards noise. It also gives you a trustworthy baseline for comparing manual and automated bidding later.
Smart Bidding can only be as good as the conversions you feed it.
Automated bidding learns from outcomes, so volume matters. Google doesn't set one hard minimum for every strategy, but the fewer conversions a campaign has, the slower and noisier the learning. As a working rule, aim for around 30 conversions per campaign in the last 30 days before you set a target cost per acquisition (CPA), the average you want to pay per conversion, and expect steadier results above 50.
If you're below that, consolidate first. Five campaigns with six conversions each give the system little to learn from, while one campaign with thirty gives it something. Where campaigns must stay separate for budget or reporting reasons, a portfolio bid strategy lets several campaigns share one strategy and pool their data.
You can also bid towards a step closer to the click, carefully. A business-to-business (B2B) company that closes three deals a month can't bid on deals, but it can bid on qualified enquiries. What doesn't work is the shortcut often suggested for small accounts: counting time on site, scroll depth or video views as conversions. As secondary conversions they have no effect on bidding, and as primary ones they teach the system to buy cheap clicks from people who scroll but never get in touch.
If none of this gets you near the threshold, Maximise conversions without a target, which simply tries to win as many conversions as your budget allows, is a reasonable middle step. Staying on manual bidding a little longer is also a legitimate decision, not a failure.
There are two main Smart Bidding strategies for most advertisers. Maximise conversions tries to win as many conversions as possible within your budget and can be given a target CPA. Maximise conversion value aims for the most revenue or value and can be given a target return on ad spend (ROAS), the value you want back for every euro spent. Enhanced CPC, the old halfway house between manual and automated bidding, is being phased out, so it is no longer somewhere to park an account.
Match the strategy to how you measure success. If every conversion is worth roughly the same to you, such as a quote request, use Maximise conversions; if values differ widely, as in an online shop with €20 and €400 orders, use Maximise conversion value.
The most common mistake is setting the target you wish you had instead of the one you have. Say your manual campaign produced 45 leads at an average of €70 over the last 60 days. A €50 target from day one tells the system to bid cautiously in almost every auction, and the result is often fewer conversions rather than cheaper ones; start at your real figure or slightly looser, around €70–80, or run without a target for two weeks and see where cost per conversion settles.
Value-based bidding works the same way in reverse, because a higher ROAS target means more restraint. If your manual campaign returned €400 in revenue for every €100 spent, a ROAS of 400%, start around 350% rather than 450% and raise it in steps once the campaign is stable.
Budget behaves differently too. Without a target, Maximise conversions aims to spend your full daily budget, while a tight target may leave money unspent. On a single day Google can spend up to twice your average daily budget, but over a month it won't charge more than about 30.4 times the daily amount, so one expensive day is not by itself a sign that something is broken.
Google Ads lets you run experiments on Search campaigns: you create a trial version with the new bid strategy and split traffic between the original and the trial, usually half each. Both run at the same time, with the same seasonality, competition and market conditions, so the comparison is fair. Plan for at least four to six weeks, longer if conversions typically arrive days after the click.
If an experiment isn't practical, migrate one campaign at a time, starting with the one that has the most conversions and the most stable history. Keep the others on manual bids as a reference point, and don't switch the rest until the first has finished learning and held its results for a few weeks.
After a switch or a significant change, the bid strategy shows a learning status while it calibrates. Expect this to take one to two weeks for a busy campaign, and longer for a quiet one or one with a long gap between click and conversion. Results in this period swing in both directions and tell you little, so avoid the changes that restart or unsettle learning:
Two traps make normal learning look like failure. Google reports conversions against the day of the click, not the day of the conversion, so the most recent days always look worse until late conversions arrive; judge performance on data that has had time to mature. And a single expensive day proves nothing, so look at rolling seven- or fourteen-day figures instead.
If something outside the campaign distorts the data, use the controls built for it rather than editing the campaign. Data exclusions tell Smart Bidding to ignore days when conversion tracking was broken, and seasonality adjustments warn it about short events, such as a three-day sale, when you expect conversion rates to change sharply.
Once the campaign has been out of learning for two to three weeks and is close to target, tighten in steps of about 10–20% and allow a full learning period between steps. Bigger jumps tend to cut volume faster than they cut cost.
Roll back only when the evidence is clear. After four to six weeks of matured data, the Smart Bidding version should be delivering clearly fewer conversions at a clearly higher cost than your manual baseline, and you should have ruled out three other explanations: a tracking problem, an unrealistic target and a change in the market. If any of those explains the drop, fix the cause and give the strategy another cycle.