Google Ads Smart Bidding: tCPA vs. tROAS Explained

Google’s Smart Bidding options get recommended constantly without much explanation of what each one is actually optimizing toward, and picking the wrong one for a store’s actual goal produces a campaign that’s technically “working” while quietly optimizing against the wrong number. Target CPA and Target ROAS are the two most commonly confused, and they solve genuinely different problems.

What Smart Bidding Actually Automates

Manual bidding means setting (and adjusting) a bid per keyword by hand. Smart Bidding hands that decision to Google’s algorithm, which adjusts bids in real time per auction based on signals a human can’t practically react to fast enough — device, location, time of day, and dozens of other factors evaluated for every single auction. The strategy chosen tells the algorithm what outcome to optimize those bid adjustments toward.

Target CPA: Optimizing for Volume at a Cost Ceiling

Target CPA (tCPA) tells Google’s algorithm to get as many conversions as possible while averaging around a specified cost per conversion. This fits a business where every conversion is worth roughly the same amount — a lead-gen form, a single-product store, a service business quoting similar-value jobs — since the algorithm has no concept of one conversion being more valuable than another under this strategy.

Target ROAS: Optimizing for Revenue, Not Just Volume

Target ROAS (tROAS) tells the algorithm to hit a specified return on ad spend, factoring in the actual value of each conversion rather than treating them as equal. This is the right choice for a store selling multiple products at different price points, where a $15 accessory and a $200 main product shouldn’t be optimized toward with the same weight — tROAS can chase the higher-value conversions more aggressively, while tCPA structurally can’t tell the difference.

The Data Requirement Both Strategies Share

Neither strategy works without accurate conversion value data feeding it — tCPA needs a real conversion count to optimize against, and tROAS specifically needs accurate revenue values passed with each conversion, not just a conversion firing. Our conversion tracking setup guide covers getting the Purchase event to pass real order value, which is a prerequisite for tROAS specifically — running tROAS on an account passing generic or missing values gives the algorithm nothing accurate to optimize revenue toward, and it’ll perform little better than tCPA while looking more sophisticated on paper.

Why Switching Too Early Backfires

Both strategies need a real volume of conversion data before they can optimize effectively — Google generally recommends at least 30 conversions in the last 30 days before trusting Smart Bidding to have enough signal, and switching a low-volume account into tCPA or tROAS too early usually produces a rocky, underperforming learning period similar to what a brand-new Performance Max or Advantage+ campaign goes through for the same underlying reason: an algorithm with too little data to learn from is guessing, not optimizing.

Setting the Initial Target Without Guessing

A tCPA or tROAS target set from a hopeful number rather than actual account history usually forces the algorithm into an unrealistic constraint — set the initial target close to what the account’s actual average cost-per-conversion or ROAS already is (checked from real account data, not a wished-for number), then adjust it gradually over time as performance data comes in, rather than setting an aggressive target from day one and wondering why delivery drops.

When Manual Bidding Still Makes Sense

A brand-new account with no conversion history, a campaign testing an entirely new product or audience with no baseline yet, or an account where conversion tracking isn’t fully trustworthy are all situations where manual bidding (or Google’s simpler Maximize Conversions strategy, which doesn’t require a target) is the more honest starting point. Layering tCPA or tROAS on top of shaky tracking data compounds the problem rather than fixing it — the Quality Score guide covers a related principle: some things genuinely need groundwork in place before a more advanced lever is worth pulling.

Signs the Target Is Set Wrong

A target set too aggressively (a tCPA too low, or a tROAS too high) usually shows up as a sudden, sharp drop in impressions and spend — the algorithm would rather show the ad less often than miss the target, which reads as the campaign “stopping working” when it’s actually just refusing to bid competitively enough to spend the full budget. The fix isn’t more budget, it’s loosening the target closer to what the account’s real performance history supports, then tightening gradually once volume stabilizes again.

Negative Keywords Still Matter Under Smart Bidding

Switching to automated bidding doesn’t reduce the need for a clean negative keyword list — if anything, it matters more, since the algorithm is now making faster, higher-stakes bidding decisions based on the same conversion data a poorly maintained negative list can distort. Smart Bidding optimizes whatever it’s given; a clean input is still the account manager’s job.