Smart Bidding in practice: tROAS vs tCPA and when each wins.
The bidding strategy debate is mostly noise. The real question is what your conversion data can support, and whether the account is clean enough for the algorithm to learn from at all.
Should you use target ROAS or target CPA. It is one of the most asked questions in Google Ads and one of the most dogmatically answered. The truth is less exciting than the debate. Both are good strategies. The right choice depends on your business model, and the strategy matters far less than the quality of the data you feed it.
The simple version of the decision
Target CPA optimizes toward a cost per conversion. It treats every conversion as roughly equal. Target ROAS optimizes toward a return on ad spend, weighting conversions by their value. The choice mostly comes down to whether your conversions are worth meaningfully different amounts.
- Lead generation where every lead is worth roughly the same: target CPA is usually the cleaner fit
- Ecommerce with a wide spread of order values: target ROAS lets the algorithm chase the valuable baskets
- Lead generation where lead quality varies a lot: target ROAS with values passed back from the CRM beats target CPA
- Early accounts with thin conversion volume: start simpler, often with target CPA or maximize conversions, until there is enough data
The part nobody wants to hear
Smart Bidding is only as good as the conversion data underneath it. If your conversion actions are misconfigured, if duplicates are inflating counts, if values are flat when they should reflect margin, then it does not matter which bidding strategy you pick. The algorithm will optimize precisely toward the wrong outcome, efficiently.
Switching bidding strategies to fix a performance problem that is actually a data problem is the most common wasted month we see in audits. Clean the conversion layer first.
Setting targets that the algorithm can hit
A target is an instruction, not a wish. Set a target ROAS far above what the account has ever achieved and Smart Bidding will simply throttle delivery, chasing a number it cannot reach, and your volume collapses. Set it too loose and you leave efficiency on the table. We set targets close to recent actual performance and move them in small increments, giving the system room to learn between changes rather than yanking the wheel.
Give it room to learn
Every change to a bidding strategy or target resets some of the learning. Accounts where someone adjusts targets every few days never stabilise. The discipline is to make a considered change, then leave it alone long enough to read a real signal. Patience is an underrated bidding skill.
So tROAS versus tCPA is the wrong first question. The right first question is whether the account is feeding the algorithm clean, correctly valued conversions, and whether the targets are grounded in reality. Get those right and either strategy performs. Get them wrong and neither will.
Written by Sophie Mills, performance marketing strategist. If this resonated and you want to apply it to your own account, you can book a strategy call or run a free audit.
How we research, source figures, and handle corrections: editorial policy.