How *target CPA* steers the auto bidder

What the bidder actually does#

Maximize Conversions replaces hand-set keyword bids with automatic bids steered by your target CPA, using Search Match to find queries beyond any list you supply. It runs on search results only, and there are no per-keyword bids to tune: the system prices each opportunity against the install goal you declared.

Apple own best practices for Maximize Conversions define target CPA as the ideal amount per install and stress judging results on average over time. Their campaign help on target CPA adds that daily cost may swing above or below while the average is steered home. Frame the choice from the Apple Ads hub and Basic versus Advanced before committing.

Set the target CPA from your maths#

Your target CPA must come from unit economics, never from a benchmark table. Start from revenue per install divided by your return target, open at that figure, and tighten only after the learning window closes with volume behind it.

A target CPA set below what the auction can bear buys silence, not efficiency, so sanity check the target against category medians first. CPT to CPA maths derives the figure cleanly, and conversion values keeps revenue per install honest for subscription and trial flows.

Budget it like learning#

Fund the bidder with a daily budget that can buy several target CPAs of spend per day, because a learner starved of trials learns slowly. Apple own guidance ties budget sizing to the target CPA and the daily conversions you want, and the Advanced monthly ceiling of daily budget times 30.4 still bounds the month.

Size the month through budget maths, protect the first fortnight with learning budgets, and resist splitting one learner into three thin campaigns that each starve; consolidation explains why fewer, fatter learners win.

Honour the two-week window#

Apple guidance points to roughly two weeks for the system to learn user behaviour against your target. During that window, hold budget, target and structure still: edits restart learning and corrupt the read you are paying to collect.

Track tap through and conversion rate separately while waiting, so you can tell a matching problem from a pricing problem on day fifteen. Search Match shows what the matching arm reaches for, and TTR diagnostics separates weak fit from weak bids.

When Manage Bids wins instead#

Choose Manage Bids where control pays: brand defence, exact harvests and tight CPA guardrails all need keyword level hands. Auto bidding explores; hand bidding exploits what exploration already proved.

A common settled pattern pairs both: Maximize Conversions discovers, Manage Bids harvests the proven terms into exact ad groups with negatives guarding the split. Structure the pair with discovery versus brand and price the harvest side via max CPT ceilings.

Reading auto-bid results#

Judge the bidder on average CPA against target plus conversion volume, never on a single costly day. Apple aims at the average, so weekly CPA at target with rising installs is success even when Tuesday stung.

Reconcile reported installs to device truth monthly; reconciliation sets the routine, and AdAttributionKit explains the signal underneath. If the average will not settle, our team reads the learning curve free with every engagement.

Keep a simple log of target changes with dates, because every edit restarts the averaging the bidder is trying to settle. Two quiet weeks teach more than two months of twitching.

Frequently asked questions#