Smart Bidding in 2025: When to Use It and When to Override It
Smart Bidding is not a shortcut. It is a tool that works exceptionally well under specific conditions and fails badly when those conditions are not met. Most Google Ads guides tell you to switch to automated bidding and let the algorithm do the work. That advice is missing half the picture.
The missing half is what Smart Bidding actually needs to function, when you should take manual control back, and how to transition between modes without blowing up your performance in the process.
What Smart Bidding Actually Uses
Google's Smart Bidding uses real-time auction signals to set bids at the individual auction level. The signals include device, location, time of day, browser, operating system, search query, remarketing list membership, ad characteristics, and hundreds of micro-signals Google does not publicly document.
The algorithm is adjusting bids in fractions of a second, evaluating probability of conversion against your target metric. It does this better than any human can manually once it has enough data to work from.
That phrase "enough data" is where everything hinges. Target CPA and Target ROAS need a minimum of 30 to 50 conversions in the last 30 days to produce reliable optimisation. Maximise Conversions can work with less but still needs a meaningful conversion history. Below those thresholds, the algorithm is essentially predicting outcomes from a small, statistically unreliable sample. It will optimise confidently toward the wrong conclusions.
Where Smart Bidding Performs Well
When the conditions are right, Smart Bidding consistently outperforms manual bidding. The right conditions look like this: a campaign with 50 or more monthly conversions, stable conversion tracking, consistent business goals, and no major account disruptions (new tracking, restructured campaigns, changed landing pages).
Established eCommerce accounts running Target ROAS with solid conversion value data are a natural fit. A restaurant booking campaign that has been running for six months and is consistently generating form submissions or call conversions is a good candidate for Target CPA. A lead generation account with consistent CPL history can use tCPA effectively once the volume is there.
The other context where Smart Bidding genuinely adds value is cross-device and cross-session journeys. A user might click on desktop, research on mobile, and convert on tablet. Manual bidding cannot account for that path. Smart Bidding can factor in the likelihood of conversion across the whole session path using Google's logged-in user data.
Where Smart Bidding Fails
New accounts are the clearest failure case. An account with two weeks of data and 12 conversions does not have the signal base for Smart Bidding to do anything useful. Start with Manual CPC or Maximise Clicks with a CPC cap.
Seasonal businesses with dramatic fluctuation cause problems. Smart Bidding learns from historical patterns. If your business has a four-week peak season followed by nine months of low activity, the algorithm will be perpetually miscalibrated. It learns what happened; it cannot predict what this year's pattern will look like until the data arrives.
Low-conversion-volume accounts, even mature ones, stay on manual longer. If you are generating eight bookings a month from Google Ads, tCPA does not have enough signal. You are better off with Manual CPC and well-structured bid adjustments by device, location, and time of day.
Campaigns with unreliable tracking are the worst scenario for Smart Bidding. If your conversion tracking is misfiring, double-counting, or tracking the wrong events, you are training the algorithm on garbage. The bidding will optimise toward whatever the system thinks it sees. Fix the tracking first, always, before switching to any automated strategy.
tROAS vs tCPA: Choosing the Right Target
These two strategies are not interchangeable. Choosing between them depends on whether your conversions have consistent value.
Target CPA treats all conversions as equal. You set a target cost per conversion and the algorithm works toward it. This is the right strategy when every conversion is worth roughly the same: a restaurant booking, a form submission, a phone call. You are buying actions, and each action has a similar value to the business.
Target ROAS requires conversion value data. It is designed for accounts where conversions have different values, like eCommerce where one order might be $30 and another $300. The algorithm optimises toward return on the specific ad spend, adjusting bids to capture higher-value conversions more aggressively. For ROAS bidding to work, your conversion tracking needs to pass revenue values back to Google. Without that, tROAS defaults to treating all conversions as equal value and you would have been better off on tCPA.
For most local service businesses and restaurants, tCPA is the right destination. For eCommerce and anything with variable transaction values, tROAS becomes relevant once the volume supports it.
Set your targets based on actual data, not aspiration. If your real CPA over the last 90 days has been $45, setting a tCPA of $20 will cause the algorithm to restrict spending so severely that impressions collapse. Start within 15-20% of your actual historical performance and tighten from there as volume allows.
When Manual CPC Still Wins
Manual CPC is not a legacy strategy. There are specific situations where it is the correct choice, not a fallback.
New campaigns need Manual CPC while they build conversion history. This is not optional. Running Maximise Conversions with zero data often results in the algorithm spending the entire daily budget in one burst, on clicks that never convert, because it has no idea what a converting click looks like.
Restructured campaigns reset the clock. If you significantly restructure an account, whether that means new ad groups, new landing pages, or new conversion events, the data the algorithm was using to make decisions may no longer apply. Treat the restructured campaign like a new one.
Highly specific or niche campaigns where query control matters more than volume benefit from manual control. If you have a campaign targeting a very specific search behaviour and cannot afford to let the algorithm broaden its interpretation, manual bidding with tight exact match keywords gives you precision that Smart Bidding can undermine.
Budget-constrained campaigns where you cannot afford to overspend on the algorithm's exploratory phase also benefit from manual control. Smart Bidding sometimes has a learning period where it deliberately tests different approaches. On a $20 per day budget, that exploration can eat several days of spend before finding its footing.
Transitioning Between Bidding Modes
The transition from manual to Smart Bidding requires patience. The learning period is real and it produces temporary performance dips.
Do not make this switch during a high-stakes period. Do not change your bidding strategy the week before Mother's Day if you are a restaurant. Do not switch tCPA targets by more than 20% in one move. Do not change the bidding strategy and the campaign structure at the same time. One variable at a time.
When moving from Manual CPC to Maximise Conversions, give the campaign two to four weeks with no major adjustments. The learning phase officially lasts about a week according to Google, but real stabilisation often takes longer, especially in lower-volume accounts.
When moving from Maximise Conversions to Target CPA, set your initial tCPA target at approximately 10-15% above your actual recent CPA. This gives the algorithm headroom to operate while you gradually tighten the target. If you set the target below historical performance immediately, the campaign will throttle heavily and volume will drop.
Moving back to manual from Smart Bidding is occasionally necessary, and it is not a failure. If the algorithm consistently misses targets, if performance has deteriorated over multiple weeks without explanation, or if you are facing a period of unusual market conditions, manual control is reasonable. Monitor for the first two weeks after the switch because manual CPC after Smart Bidding can produce initial volatility as the system recalibrates.
The Practical Approach We Use
For new client accounts, we start on Manual CPC with capped bids set based on the target CPA and an estimated conversion rate from the landing page. We let campaigns run until they hit 30 conversions in a rolling 30-day period before considering the switch.
For accounts that qualify, we move to Maximise Conversions first. We hold that strategy for four to six weeks and confirm the CPA it delivers. If the CPA is acceptable and volume is sufficient, we then move to Target CPA set at the actual delivered CPA plus 10%.
For eCommerce clients with variable revenue data, tROAS is the endpoint. Getting there requires conversion value tracking to be working cleanly first, which means verifying that purchase events are passing accurate revenue figures before any bidding change.
The biggest mistake we see in accounts inherited from other agencies is Smart Bidding running on accounts that never had the data to support it. The strategy looks sophisticated. The results tell a different story.
If your account is not hitting targets and you are on Smart Bidding with fewer than 30 monthly conversions, that is the first thing to fix. Learn more about how conversion tracking underpins all of this at our Google Ads conversion tracking guide.