Target CPA and target ROAS at a glance
Target cost per acquisition is a bid target for campaigns optimising towards conversions. The advertiser states an average CPACost of each conversion. it is willing to pay, and the bidding system adjusts bids auction by auction to get as many conversions as it can at or near that average. The Google Ads Help documentation describes it as a target for the average, so some conversions will cost more and some less. Every conversion counts the same, whether it is a small order or a large one.
Target return on ad spend works on value. The advertiser states the ROASRevenue generated per unit of ad spend. it wants, meaning conversion value per unit of spend, and the system bids higher where it predicts higher value and lower where it predicts less. The Help documentation explains that this requires conversion values to be recorded and enough valued conversions for the system to learn from. A large basket and a small one are treated differently.
Both are forms of automated bidding, covered in depth in the guide to Smart Bidding strategies, and both exist in similar forms in other platforms and in search management tools, as the guide to bid strategies in Search Ads tooling shows. The choice of whether to automate at all is a separate decision, covered in the manual versus automated bidding comparison.
Side-by-side comparison
Target CPA vs Target ROAS
| Row label | Target CPA | Target ROAS |
|---|---|---|
| What is optimised | Number of conversions at an average cost | Conversion value relative to spend |
| Data required | Reliable conversion tracking with enough volume | Reliable conversion tracking plus accurate values, with enough valued conversions |
| Value tracking | Not needed; every conversion counts as equal | Essential; wrong or missing values mislead the system |
| Best-fit businesses | Lead generation, sign-ups, bookings and stores with similar order values | Stores with varied basket sizes, travel, and lead businesses that pass back qualified values |
| Volume behaviour | Lowering the target reduces volume; raising it buys more | Raising the target reduces volume; lowering it buys more |
| Target setting | From the maximum affordable cost per conversion | From the minimum acceptable return, anchored on break-even |
| Response to change | Steady when conversion value is stable | Responds to shifts in value mix, promotions and pricing |
| Reporting | Cost per conversion and conversion volume | Return, conversion value and profit after costs |
| Failure modes | Buys cheap, low-value conversions; ignores basket size | Chases high-value but rare conversions; starves on thin value data |
The directions run opposite, which confuses people. Tightening a target CPA means lowering it; tightening a target ROAS means raising it. Either way, tightening protects efficiency and costs volume.
Neither target guarantees the stated figure. Both are averages the system aims for over time, and results drift above and below them as auctions, seasonality and conversion delays shift. Conversion lagThe time between an ad interaction and the conversion it leads to, which leaves the most recent reporting days incomplete. matters especially for value bidding: if orders or qualified deals are recorded days after the click, recent performance looks worse than it is, and teams that react to it by changing targets keep the system in a permanent state of relearning. Agree a review rhythm that respects the lag before anyone touches a target, and change one thing at a time so the effect can be read.
When to choose target CPA
Choose target CPA when conversions are worth roughly the same to the business. A lead form, a demo request, a free trial sign-up, an app install or a booking for a standard service usually fit. If a salesperson treats every lead the same and contract values are not known at the time of conversion, a cost target is the honest choice.
Target CPA also fits online stores where order values cluster tightly, or where value tracking is not yet reliable. Optimising towards value that is wrong can do more harm than optimising towards a simple count. Fix the values first, then consider switching.
Set the target from what a conversion is worth, not from history. For leads, work back from the share of leads that become customers, the value of a customer and the margin; the CPA calculator helps with the arithmetic. Then decide whether the campaign must break even on the first conversion or can recover cost over the customer's lifetime.
The weakness of target CPA is that it is blind to quality. A campaign optimising towards form fills will find the cheapest form fills, which are often the least qualified. Feeding back qualified stages as the conversion, where volume allows, keeps the target pointed at what matters.
When to choose target ROAS
Choose target ROAS when conversion values vary meaningfully and are tracked accurately. Online stores with a wide range of basket sizes, travel bookings, and businesses that send back deal values from a CRM are typical fits. The system can then bid more for the searches and audiences that tend to bring larger orders.
Target ROAS needs good data. Values should reflect what the business keeps, ideally after returns and cancellations, and should arrive consistently. Some advertisers send profit rather than revenue as the value, which aligns bidding with margin directly.
Set the target from margin. The break-even ROAS is the return at which ad spend is covered by the gross margin on the sales it brings; below it, every sale loses money. The guide to contribution margin and break-even explains the calculation, and the break-even calculator and the ROAS calculator help apply it. A target above break-even protects profit; a target closer to it buys growth at thinner margins.
Target ROAS is less forgiving of thin data than target CPA. A campaign with few valued conversions gives the system little to learn which searches bring large orders, so predictions swing and spend can stall. Grouping similar products into one campaign, or using a shared portfolio strategy across campaigns, gives the system more to work with.
An outdoor retailer sets its target from margin
An outdoor gear retailer runs target CPA at £25 on a shopping campaign. Orders range from £18 for socks to £420 for tents, with an average of £85 and a gross margin of 40%. Under target CPA, the campaign wins many small orders: 62% of conversions are under £40. The retailer calculates a break-even ROAS of 2.5 (one divided by 0.4), sets target ROAS at 3.5 to leave room for overheads, and fixes its value tracking to exclude returns. After a five-week learning period, conversions fall by 14%, but conversion value rises by 21% and gross profit after ad spend rises by 30%.
Decision table
| Situation | Choose | Why |
|---|---|---|
| Every lead or sign-up is worth about the same | Target CPA | Value data would add little |
| Order values vary widely and are tracked accurately | Target ROAS | The system can favour higher-value conversions |
| Value tracking is missing or unreliable | Target CPA, while values are fixed | Wrong values mislead value bidding |
| CRM sends back qualified deal values with enough volume | Target ROAS on those values | Optimises towards pipeline, not form fills |
| The campaign is new with little conversion history | Start with a looser target of either kind | The system needs data before tight targets work |
| Profit matters more than revenue | Target ROAS on margin-based values | Aligns bidding with what the business keeps |
The comparisons hub covers related bidding and measurement decisions.
Common mistakes
- Copying the historical average as the target. Targets should come from margin and customer value; an unprofitable average locks in losses.
- Switching to target ROAS with bad values. If values include returns, tax or test orders, the system learns the wrong lesson.
- Tightening targets abruptly. Large changes reset learning and can collapse volume; move in small steps.
- Judging on a single week. Automated bidding needs time to settle after changes; read results over a full learning period.
- Forgetting the direction. A higher target ROAS and a lower target CPA both cut volume; know which way growth lies.
- Ignoring conversion quality under target CPA. Cheap leads are not good leads; feed back qualified stages where possible.
- Setting one target across very different products. Margins differ by category; separate campaigns or value rules keep targets meaningful.
