About 30 conversions in 30 days is the practical floor a Google Ads Smart Bidding strategy needs to honor a Target CPA reliably in 2026, and a strategy starved below it is the most common reason your actual CPA sits stubbornly above target. A Target CPA that overshoots is one of the most misread problems in paid search: the instinct is to lower the target to force a lower cost, which usually collapses conversion volume and triggers a fresh learning phase that makes everything worse. The durable fix is almost always the opposite — re-anchor the target on reality, gently, without resetting learning.
This guide explains exactly why Smart Bidding overshoots a Target CPA or Target ROAS, how to tell a real problem from normal learning-phase noise, and the step-by-step way to re-anchor the target. To check your account against the most common Smart Bidding drift problems automatically, run our free 5-axis Google Ads audit.
Updated 2026-05-20 with current Smart Bidding learning-phase and Target CPA behavior observed across US, UK and European accounts.
- Target CPA is an average, not a ceiling — individual conversions land above and below by design. 2. A sustained gap over 20-30% past learning is the real signal, not single-day noise.
- Three root causes: an unrealistic target, too little conversion data, or destabilizing edits. 4. Lowering the target rarely helps — it collapses volume and restarts learning. 5. Re-anchor gently toward your true historical CPA, one 15-20% step at a time.
Why is my actual CPA higher than my Target CPA?
The first thing to internalize is that Target CPA is an average goal, not a hard ceiling. Smart Bidding aims for an average cost per conversion across all auctions, so individual conversions will land above and below the target by design. A handful of expensive conversions in a week does not mean the strategy is broken.
What matters is the sustained average. A few percentage points above target is normal. A gap of more than 20 to 30 percent, holding steady past the learning window, is the signal that one of three root causes is in play: the target is unrealistic versus history, the data is too thin, or recent edits keep restarting learning.
The wrong first move — lowering the target to force a cheaper cost — is also the most common one. It tells the algorithm to bid only on cheaper, lower-intent auctions, which usually collapses volume faster than it lowers cost. Diagnose the cause before touching the dial. For the broader choice between target-based and volume-based bidding, see our Maximize Conversions vs Target CPA guide.
Is the target unrealistic versus historical performance?
This is the single most common cause of a persistent overshoot. If you set a Target CPA far below what the account has ever achieved, you are asking Smart Bidding to do something the auction will not allow.
Pull your historical CPA — Look at the trailing 30 to 90 days for this campaign, or a comparable one, before the current target was applied. That average is your realistic anchor. A target set 40 or 50 percent below it is an aspiration, not a target, and the algorithm cannot honor it without starving itself of auctions.
Anchor, then improve — The honest path to a lower CPA is not a lower number in the box. It is a better Quality Score, stronger landing pages, and cleaner conversion tracking — the levers that actually lower the cost of a conversion. Set the target where the data says you are, then earn your way down. If you are weighing CPA against value-based bidding, our Target ROAS vs Target CPA guide covers the trade-off, and you can model a realistic figure with our CPA calculator.
Is there enough conversion data to hit it?
Smart Bidding is a statistical model, and a model with too few data points makes noisy, unreliable bids. A thin strategy will drift above target simply because it cannot optimize confidently.
The conversion floor — Google's guidance points to roughly 30 conversions in the trailing 30 days for Target CPA, and about 50 for Target ROAS, measured per bid strategy rather than per campaign. Below that floor the actual CPA swings widely and the average tends to sit above target.
Consolidate the signal — The fastest way to feed a starved strategy is structural: merge fragmented campaigns, route low-volume campaigns into a portfolio bid strategy, and keep your primary conversion action consistent so one strategy sees more data. This is the same root issue behind a stuck status — see our Learning limited fix guide. A strategy with 80 conversions a month hits its target far more reliably than four campaigns with 20 each.
How target changes and edits destabilize bidding
Even a realistic target on a data-rich account will overshoot if you keep editing it. Every large change can restart the learning phase, throwing the strategy back into noisy early behavior.
What restarts learning — A Target CPA or Target ROAS change larger than roughly 15 to 20 percent, a switch in bid strategy type, a dramatic budget change, or a change to which conversion actions count. Each of these resets the calibration the algorithm has already done.
What does not — Small budget nudges, adding a handful of keywords or negatives, minor ad edits. These rarely trigger a reset.
The discipline — Make one meaningful change at a time, then wait the full learning window before the next. Stacking several target edits in a single week is the most common self-inflicted reason a CPA never converges. Seasonality is a legitimate exception, but handle it with planned adjustments rather than reactive edits — our seasonal bid adjustments guide shows how to do it without destabilizing the model.
Should you switch to Maximize Conversions with a cap?
When an account is simply too thin to honor a Target CPA, forcing the target is the wrong tool. A volume-first strategy often stabilizes faster.
The case for it — Maximize Conversions has no explicit target to satisfy, so it has more freedom to enter auctions and gather signal. With a Target CPA cap applied, it still respects an upper bound on cost while building the conversion volume a thin strategy needs. For a starved account stuck above target, this combination frequently exits the noisy phase sooner than a strict Target CPA ever could.
The sequence — Start with Maximize Conversions plus a sensible cap to build a stable baseline of conversions. Once the volume is there and the cost has settled, reintroduce a clean Target CPA anchored on the new, reliable average. You get the volume first and the control second, instead of fighting for both at once on too little data.
The tCPA-miss fix table
Work this table top to bottom — it is ordered from the fastest cause to confirm to the deepest structural fix.
Lowering the Target CPA to chase a cheaper cost is the most common mistake here. It tells Smart Bidding to skip higher-intent, higher-cost auctions, which usually collapses your conversion volume far faster than it lowers your CPA — and a change over 15 to 20 percent restarts learning on top of that. Re-anchor toward your true historical CPA instead, in small steps.
How to re-anchor the target step by step
Re-anchoring is a sequence, not a single edit. Done gently, it preserves the learning the algorithm has already accumulated and walks the target toward your goal without a reset.
Confirm the gap is real — Rule out learning-phase noise first. Only treat the overshoot as a problem if it persists past the window and exceeds target by more than 20 to 30 percent on a sustained average.
Set the anchor — Pull the trailing 30 to 90 days of CPA as your realistic baseline. This is where the target should live today, even if your goal is lower.
Move in one modest step — Adjust the Target CPA toward the goal by no more than 15 to 20 percent in a single change, so you stay inside the existing learning rather than restarting it.
Lock and wait — Leave the strategy alone for about 7 days and 30 conversions, make no other large edits, then read the sustained average to decide whether another small step is warranted.
Audit before you scale. A strategy that overshoots quietly burns budget on the wrong auctions. Run the SteerAds free 5-axis audit to surface Smart Bidding drift automatically, and use our CPA calculator to set a target the auction can actually honor.
Sources
Official sources consulted for this guide:
-
support.google.com — About Target CPA bidding
-
support.google.com — About the learning period for Smart Bidding
-
support.google.com — About Maximize conversions bidding
-
blog.google — Google Ads and Commerce updates
FAQ
Why is my CPA higher than my target?
An actual CPA above your Target CPA almost always traces to one of three causes: the target is unrealistic versus your historical CPA, the strategy has too little conversion data to model bids reliably, or recent edits keep restarting learning and resetting the algorithm to noisy early behavior. Target CPA is an average goal, not a ceiling, so individual conversions will land above and below it by design. A persistent gap of more than 20 to 30 percent above target, sustained past the learning window, is the real signal that one of those three root causes is in play rather than normal variance.
Should I lower my Target CPA to force a lower cost?
Usually not. Lowering the target tells Smart Bidding to bid only on cheaper, lower-intent auctions, which can collapse your conversion volume far faster than it lowers your CPA. The result is often fewer conversions at a barely-improved cost, plus a fresh learning phase that makes the next two weeks noisy. If your actual CPA is genuinely above where the account can profitably operate, the durable fix is improving Quality Score, landing pages and conversion tracking — not squeezing the target. Re-anchor toward your true historical CPA instead of forcing an unreachable number.
How often can I change my Target CPA without breaking it?
Change it in small steps and space the changes out. A single adjustment larger than roughly 15 to 20 percent can restart the learning phase, throwing the strategy back into noisy early behavior for several days. The practical rule is one meaningful target change at a time, then wait the full learning window — about 7 days and roughly 30 conversions — before judging the result or making the next move. Stacking several target edits in one week is the most common self-inflicted cause of a strategy that never settles near target.
Target CPA versus Maximize Conversions with a cap — which should I use?
Target CPA gives you explicit control over average cost but needs enough volume to honor it; Maximize Conversions chases volume and, with a Target CPA cap set, still respects an upper bound while having more freedom to enter auctions. For thin accounts stuck above target, Maximize Conversions with a sensible cap often gathers signal faster and stabilizes sooner. Once a stable conversion baseline exists, you can reintroduce a clean Target CPA. Choose by your binding constraint: control versus volume.
How long does Target CPA take to stabilize and hit target?
Expect roughly the learning window — about 7 days and around 30 conversions for Target CPA — before delivery and cost settle, and longer for low-volume accounts that accumulate conversions slowly. During learning, the actual CPA is noisy by design and frequently sits above target; judging the strategy before it exits learning leads to premature edits that restart the clock. If the CPA is still well above target two to three weeks after settings stabilized, the cause is structural — target, data, or tracking — not a matter of waiting longer.
Does editing the target reset Smart Bidding learning?
A large target change does. Moving the Target CPA or Target ROAS by a big margin, switching the bid strategy type, changing the budget dramatically, or altering which conversion actions count all restart learning. Minor nudges — a small budget change, adding a few keywords — generally do not. The safe practice is to treat the target as a dial you turn gently: one modest move, then a full learning window of observation before the next. Frequent large edits are why many strategies never converge on their target at all.
Is a CPA above target during the learning phase normal?
Yes. During the learning phase the algorithm is still calibrating bids against your conversion data, so the actual CPA swings above and below target and the average is unreliable. This is expected behavior, not a fault, for the first 7 days and roughly 30 conversions. The mistake is reacting to learning-phase noise with target cuts or bid edits, which restarts learning and prolongs the instability. Wait out the window with stable settings, then assess the sustained average rather than any single noisy day.