Target CPA

Target CPA no longer exists as its own standalone bid strategy in Google Ads - it now lives as an optional target field inside Maximize Conversions. With it set, the algorithm bids up on auctions it predicts will convert near or below your target and pulls back on ones it expects to cost more, so individual conversions land above and below the number while the account average tracks toward it over time. That averaging is the whole mechanism: a target set well below what the market actually supports does not make the algorithm work harder, it just makes it buy less.

Target CPA can be applied at the individual campaign level or shared across several campaigns through a portfolio bid strategy, which pools conversion data so campaigns that individually do not have enough volume to learn well can borrow signal from others in the same portfolio. That is a useful option for smaller accounts running several campaigns each too thin on conversions to support its own target on its own, though it also means a sudden change in one campaign's conversion pattern, such as a promotion or seasonal spike, can shift bidding behavior across every campaign sharing that portfolio, not just the one where the change happened.

It is most useful once an account has steady conversion volume and a real baseline CPA to work from - typically after a stretch on plain Maximize Conversions with no target, long enough to see what cost the market actually supports. It is less useful for ecommerce accounts where individual order values vary a lot, since Target CPA treats every conversion as equally valuable; Target ROAS is the better fit there because it accounts for that variation directly. It is also worth noting the target applies as an account-level or campaign-level average, not a per-conversion cap, so a handful of expensive conversions mixed with several cheap ones can still average out to the target even though none of them individually cost that exact amount.

The single most common mistake is setting an aggressive target well below the account's demonstrated average CPA and expecting the algorithm to find a way to hit it. It cannot manufacture cheaper conversions the market does not support - it responds by bidding down and skipping pricier auctions, which shows up as reduced volume and an underspent budget rather than the cost savings someone was hoping for. A second mistake is changing the target too often; each meaningful change resets a learning period of roughly one to two weeks with noisier results, so an account that adjusts its target CPA weekly is essentially never out of the learning phase.

Compare Cost/conv against the target on a trailing basis rather than day to day - daily numbers bounce around even in a healthy campaign. If actual cost/conv sits persistently far from the target, check Impr. share lost (rank) alongside Impr. share lost (budget): a low target that is causing the account to lose auctions on rank, meaning bids too low to compete, points to a target that needs raising, while a budget-capped account with room to spend more points to a target that is actually working and could be scaled.