How to lower CPA
Nine levers, ranked by effort. Start with waste (search terms, conversion hygiene, constrained targets), then structure (segments, copy), then the conversion stage where a 25% CVR lift cuts CPA 20%, then the offer itself. Each lever acts on one funnel gate.
1.Cut the worst search terms and placements
2ClickHoursSort spend by conversions, exclude the zero-conversion tail. The oldest trick still moves accounts that have not done it in a quarter.
2.Fix conversion-action hygiene
4AcquisitionHoursRemove soft events from the optimized conversion set so bidding buys customers, not page-views. Definitional gap: cost per action vs acquisition.
3.Raise a constrained Target CPA
4AcquisitionHoursCounterintuitive: an unrealistically low target throttles volume and starves learning. Reset to history, then step down.
4.Tighten geo, schedule and device splits
1ImpressionDaysBid down segments that convert poorly instead of paying one blended price for unequal auctions.
5.Rewrite ads against the query, not the brand
2ClickDaysHigher CTR lowers effective CPC in auction ranking, and pre-qualifying copy lifts downstream conversion rate at the same time.
6.Landing-page speed and message match
3ConversionWeeksThe highest-leverage stage: CPA = CPC ÷ CVR, so a 25% CVR lift cuts CPA 20% with zero media changes.
7.Shorten the form / reduce friction
3ConversionWeeksEvery field trades volume for quality. Price the trade with lead-to-close data, not opinion.
8.Creative refresh cadence
1ImpressionOngoingFatigued creative pays rising CPMs for falling CTRs on social. Refresh before the decay, not after.
9.Offer and pricing tests
4AcquisitionQuarterThe ceiling-raiser: a stronger offer lifts CVR and AOV together, lowering CPA while raising what a good CPA even is.
Patience
The lag between change and number
Two lags stack. Attribution windows book conversions days after their clicks, so this week's CPA partially reflects last week's account. And automated bidding re-learns after significant edits; Google's Target CPA documentation is explicit that the strategy optimizes across auctions over time, not instantly. Practical rule: change one lever, wait one full conversion cycle plus a week, then read. Detail on Target CPA and the formula page's attribution section.
The trap
When lowering CPA is the wrong goal
CPA falls fastest by buying only the easiest conversions, which caps volume. If the business needs growth, the right question is the marginal CPA of the next customer against the break-even ceiling, not the average. A rising average CPA with rising volume can be the plan working. Judge against your vertical median and the benchmark table.