If your account generates fewer than 50 conversions a month, run Maximize Conversions. If it generates more, move to Target CPA or Target ROAS. That is the short version of smart bidding strategies explained - the rest is knowing why that number matters and how to set a target that does not choke your delivery.
Google names four Smart Bidding strategies. In practice there are two modes of operation, and confusing them is the most consistent way to burn budget on a campaign that looks fine in Ads Manager until you read the actual data.
This is the Bid Maturity Ladder - two rungs, and most accounts try to skip the first one.
| Dimension | Maximize mode | Target mode |
|---|---|---|
| Strategies | Maximize Conversions, Maximize Conversion Value | Target CPA, Target ROAS |
| Optimizes for | Volume or value, within your budget | Efficiency, against your cost or return target |
| Data required | None - works from day one | 50+ conversions/month minimum |
| Your input | Budget only | Budget + a specific CPA or ROAS number |
| Failure mode | Unconstrained spend | Chases a target it cannot reach, starves delivery |
| Best fit | New campaigns, low-volume accounts | Mature accounts with stable conversion history |
What actually matters in this decision
Most accounts land on Target CPA or Target ROAS because those names feel like control. They are control - but only once the algorithm has enough signal to honor them meaningfully.
The 50-conversion threshold is not Google's arbitrary cutoff. It is the point at which the algorithm has enough data to make directional decisions instead of guessing. Below that number, any target-mode google ads bidding strategies will behave erratically - either spending slowly while the system hunts for auctions it can enter within the cost constraint, or entering bad ones just to keep spending.
Running Target CPA on a 20-conversion-per-month account is asking the algorithm to hold a precise cost limit on a road it has only driven twice.
The two modes of Google smart bidding
Maximize mode (Maximize Conversions, Maximize Conversion Value) tells Google to spend your budget getting as many conversions, or as much conversion value, as possible. No cost constraint applied. You control the budget ceiling. Google optimizes within it.
Target mode (Target CPA, Target ROAS) adds a constraint. Target CPA says: get conversions, but only at or near this cost per acquisition. Target ROAS says: get conversion value, but only at or near this return multiplier. Google now has to find conversions AND hold your constraint simultaneously.
The difference between them is not which is "better" as a google ads bidding strategies decision. It is which is appropriate given how much conversion signal the algorithm has to work with. One is a free optimization. The other is a constrained one - and constraints only help when the system has enough history to know which auctions to enter.
Rung one: Maximize Conversions
Maximize Conversions is underused because it looks like giving up control. It is the right first rung for every new campaign, for two reasons.
First, it builds the conversion history the algorithm needs before any cost constraint is applied. Second, it surfaces your real baseline CPA before you assign a target to it. Most Target CPA inputs are back-calculated from a break-even number - useful as a goal, but not the same as what the account can actually deliver right now against current creative, landing-page conversion rate, and audience quality.
Running Maximize Conversions for four to six weeks surfaces that real number. Then Target CPA becomes a genuine constraint against a known baseline, not a wish against an imagined one. Accounts that hold Maximize Conversions until conversion volume is stable tend to transition to Target mode with far less learning friction than accounts that rush the switch.
Maximize Conversion Value follows the same logic applied to ecommerce: let Google maximize total revenue within budget, then layer Target ROAS once you know what return rate the business can profitably sustain.
Rung two: Target CPA and Target ROAS
Set a Target CPA at your break-even number, not your aspirational one. The algorithm can optimize downward from a viable starting point. It cannot pull a CPA below a level the account's conversion rate can structurally support.
Target CPA has one reliable failure mode: the target is set too aggressively from day one. A target below what the account can structurally achieve starves delivery. Google sees bidding opportunities it cannot take because they would push over the cap, so it stops entering those auctions. Spend drops. Impressions fall. The instinct is to check the budget - the budget is fine. The target is the constraint.
The fix: start with a target 10-20% looser than the ultimate goal. Tighten in 10-15% increments every two weeks as data confirms the system can hold each new level.
Target ROAS carries the same trap. A 5x ROAS target on an account that has historically returned 3.2x is not a goal. It is a constraint that blocks most auctions before they start.
The input that makes any strategy work: negative keywords
Smart bidding optimizes distribution. It does not filter queries. A campaign running broad match without a solid negative keywords strategy will optimize its spend efficiently - on the wrong traffic.
The algorithm learns from conversion signals in your account. If unqualified queries are triggering low-intent clicks that occasionally convert - or converting on actions that do not translate to real revenue - the system learns toward those patterns. Broad match in 2026 is genuinely broad. Without a maintained negative list, you are asking smart bidding to make better decisions from worse inputs.
Run a search terms report weekly. Build a negative keywords strategy at the campaign level before layering any smart bidding mode on top. The Performance Max best practices guide covers query exclusions on PMax, where the negative keyword mechanics work differently. On standard Search campaigns, the principle is the same: data quality in, quality decisions out.
BAVai scans search term patterns every morning across every BAVS account - not because constant changes are needed, but because one week of unchecked broad match traffic can corrupt the conversion signal that smart bidding is learning from.
Where each mode wins
Run Maximize Conversions or Maximize Conversion Value if:
- The campaign is new or recently restructured
- You have fewer than 50 conversions in the last 30 days
- You do not yet know your realistic CPA or ROAS at current creative and landing-page quality
- A major change reset the learning phase - significant budget shift, new creative batch, campaign restructure
Run Target CPA or Target ROAS if:
- You have 50+ conversions/month, stable for at least four weeks
- You know the CPA or ROAS your business can profitably hold
- You are running Performance Max alongside Search campaigns and need consistent unit economics across both
- You are scaling and need predictable cost per acquisition as budget increases
The case for staying on Maximize longer
There is a clean argument for staying on Maximize Conversions longer than feels comfortable. Running Target CPA on a low-volume account does not just underperform - it corrupts the learning signal the algorithm needs to eventually make Target mode work. Every week of constrained, thin-data learning makes the eventual transition harder.
The accounts that hold Maximize Conversions until conversion volume is genuinely stable - then move to Target mode with a realistic starting target - tend to reach their CPA goal faster than accounts that impose a number early and fight the algorithm for months.
Verdict: Run Maximize Conversions until the account can feed the constraint. Move to Target CPA or Target ROAS at 50+ conversions/month with a target set 10-20% looser than your goal. Tighten gradually, maintain negative keywords as a data hygiene layer, and treat the two rungs as sequential - not optional.
The honest version of smart bidding strategies explained is that the strategy matters less than the account's readiness to use it. A mature account on Maximize Conversions outperforms an immature one on Target ROAS, consistently.
If you removed the CPA target on your most important Search campaign today and ran Maximize Conversions for six weeks, what do you think the real number would be - and would it change the target you have been trying to hit?