Target CPA bidding

Target CPA is Google's automated bid strategy: you name an average cost per conversion, it sets a bid in every auction to hit that average▸Google Help · checked 2026. It manages the mean, not each conversion. Set it from history, not hope.

Per auction, the system predicts conversion probability from signals you cannot see at bid time (device, hour, audience, query context) and scales the bid so the campaign's spend-per-conversion averages your target. Google's own documentation frames modern Smart Bidding as per-auction AI bid setting▸Google · checked 2026. Two consequences follow. First, individual conversion costs scatter widely around the target; judge weeks, not days. Second, the strategy needs conversion volume to learn: thin accounts get noisy predictions and noisy spend.

• Constrained target: target well below achievable CPA, so the algorithm sits out auctions. Symptom: impressions collapse, CPA holds. Fix: raise the target to history, step down gradually.
• Learning resets: big edits (target swings, budget jumps, conversion-action changes) restart learning. Batch changes; move targets 10-15% at a time.
• Polluted conversion actions: the strategy optimizes whatever counts as a conversion. Include soft events and it buys cheap actions, not customers. The action-vs-acquisition gap applies at bid time, not just reporting.
• Uniform-value assumption: orders vary in value, but Target CPA treats each conversion the same. Wide AOV spread wants value-based bidding instead; conversion math on CPA vs ROAS.

The target is a business decision wearing a bidding UI. Ceiling: break-even CPA from AOV, margin and repeat rate. Market context: your vertical's survey median on the benchmarks table. Then the flip side: at your target and required volume, the max-spend mode tells you what budget the plan implies. Platform-level context sits on Google Ads CPA and blended CPA.