Definition
Target ROAS bidding is an automated strategy in which you tell Google Ads the return you want from a campaign and the system sets each bid to pursue it, raising bids in auctions where it predicts high conversion value and lowering them where it predicts little. It optimises toward an average across the campaign rather than a guarantee on any single auction, so results land above the target on some days and below it on others.
Setting a target from the store's own numbers
- Contribution margin rate on the products in scope
- 0.38 of net revenue
- Break-even ROAS
- 1 ÷ 0.38 = 2.63
- Overhead the channel is expected to cover
- A further share of each sale, decided by the owner rather than by the platform
- Working target entered in the campaign
- 3.5, which leaves room above break-even for overhead and profit
- What the campaign optimises toward
- Conversion value divided by cost, averaged across the campaign over time, not per order
- What the store watches alongside it
- Order volume, impression share and new customer count, because a target set too high buys efficiency by buying less
An illustrative setup. The point is that the target is derived from margin and overhead rather than copied from another account, since the same figure can be comfortable for one catalogue and unreachable for another.
Why it matters
Target ROAS bidding matters because it is how a store expresses commercial intent to a system that would otherwise optimise for whatever is easiest to buy. Once conversion values are accurate and the target reflects real margin, the algorithm can evaluate signals no manual bidding process could reach: the device, the time, the query, the audience and the specific product in the auction. That is a genuine advantage on large catalogues. It is also where profit thinking enters an ad account, because sending gross profit rather than revenue as the conversion value turns the same setting into a profit target while the platform still calls it ROAS.
Where it goes wrong
- Setting the target above break-even without checking whether the auction can deliver it, which quietly throttles spend until the campaign serves almost nowhere
- Moving the target repeatedly in small steps, which restarts learning and prevents the strategy from ever settling into a stable pattern
- Feeding it revenue when margin varies across the catalogue, so the system optimises toward the cheapest revenue rather than the most valuable orders
- Running one target across products with very different economics, when custom labels exist precisely so that thin margin items can carry a different goal
- Ignoring conversion tracking quality: duplicated purchases, missing checkout events or an inconsistent order total all teach the model the wrong lesson before any bidding decision is made
Questions about target ROAS bidding
How much data does target ROAS bidding need before it works?
Enough recent conversion value in the campaign for the model to find a pattern, which in practice means a steady flow of orders rather than a handful. Campaigns with sparse conversions tend to behave erratically under a value target. Where volume is thin, consolidating campaigns so the data pools in one place usually helps more than lowering the target does.
What happens when the target is raised?
The system becomes more selective, bidding only where it predicts a strong return, so impressions and spend fall while the reported ratio improves. Total profit can fall at the same time, because the auctions given up were still contributing. Change the target in meaningful steps, then judge the result on orders and profit rather than on the ratio alone.
Can target ROAS be used in Performance Max?
Yes. Performance Max campaigns run on maximise conversion value, with an optional target ROAS, and the same principles apply: accurate values, a target derived from margin, and enough conversion volume to learn from. Because the campaign spans several surfaces at once, structure and asset quality carry more weight there than in a Standard Shopping campaign with the same target.