Target CPA vs Target ROAS: Google Ads Bidding for Ecommerce (2026)
Most ecommerce owners are not really stuck on whether to use Smart Bidding. They are stuck bouncing between Target CPA and Target ROAS — today tROAS feels like it spends too little, tomorrow tCPA feels like it only buys cheap conversions. In 2026 Google split these back into standalone options (as of June, "Maximize conversions with a Target CPA" is now just Target CPA, and "Maximize conversion value with a Target ROAS" is now Target ROAS). The underlying bidding behavior is unchanged, but the cost of picking wrong has not gone anywhere.
This post lays out how the four Smart Bidding strategies divide the work so you can tell which box you belong in. For the broader shifts, start with 2026 Google Ads bidding strategy changes.
First, what each strategy actually optimizes for
All four use machine learning to set bids at auction time, but they optimize toward different goals — and that is the first fork in the road.
- Maximize Conversions: get as many conversions as possible within budget, regardless of what each is worth. It spends the full budget with no efficiency constraint.
- Target CPA: get as many conversions as possible under a cost-per-acquisition ceiling you set. It thinks in units, and does not distinguish a 20 dollar order from a 500 dollar one.
- Maximize Conversion Value: pull total conversion value as high as possible within budget, leaning toward higher-value orders. It also spends the full budget.
- Target ROAS: maximize conversion value while hitting a return target you set. A 400 percent target means you want 4 dollars back for every 1 dollar spent.
For ecommerce the deciding question is: do your conversion values vary a lot? If order values differ widely — which is typical in ecommerce — value-based strategies (Target ROAS or Maximize Conversion Value) usually fit the business better than unit-based ones.
Which bid strategy when (table)
| Situation | Suggested strategy | Why |
|---|---|---|
| New campaign, not enough conversion data | Maximize Conversions / Max Conversion Value | Build data first, no efficiency constraint, use as a transition |
| Order values vary a lot, want to protect return | Target ROAS | Common ecommerce default, prioritizes high-value orders |
| Values vary but data is still thin | Maximize Conversion Value | Value-based without getting throttled by too-high a target |
| Order values roughly equal, want to control CPA | Target CPA | Most direct way to cap cost per acquisition |
| Leads or form-fills with similar value | Target CPA | Fits conversions of uniform value |
| Chasing volume, budget must be spent | Maximize Conversions | Unconstrained, fills the budget |
One-line memory aid: values vary widely, use the ROAS column; values are uniform, use the CPA column; not enough data yet, use a Maximize strategy as a bridge.
Data thresholds: do not rush the target on
Smart Bidding learns from the conversion data you feed it, so too little data makes it swing wildly. Google's rough guidance is around 30 conversions in the last 30 days for Target CPA and at least 50 for Target ROAS (more is always better, and these are ballpark ranges, not hard switches). Accounts under 30 to 50 conversions a month with variable order values often see erratic results on tROAS.
A steadier rhythm is: run Maximize Conversions or Maximize Conversion Value without a target for two to four weeks to build data, then layer in a tCPA or tROAS based on actual performance. That way the target is set against real numbers rather than a hopeful guess.
The three most common mistakes
1. Setting the target too aggressively. The most common 2026 mistake is setting a ROAS or CPA target that is too aggressive. If you are hitting 4x and jump straight to 6x, the algorithm sharply throttles spend chasing volume that does not exist at that threshold, and only bids on the cheapest, lowest-quality traffic. Adjust in steps of no more than 15 to 20 percent at a time, and give each change at least two weeks before judging it.
2. A messy conversion setup. If page views, scroll depth, video plays, newsletter signups and actual purchases are all set as primary conversions, the algorithm sees "50 conversions a day" when only 5 are purchases, and optimizes for the easy, low-value actions. Target ROAS is only as good as the conversion values you feed it — if values are missing, inconsistent or inflated, bidding optimizes toward the wrong outcome. For fixing tracking, see the complete Google Ads for ecommerce guide.
3. Confusing strategy with channel. A bidding strategy is how you bid, not which campaign type you run. Performance Max also runs on these same Smart Bidding targets, so understanding the bidding logic matters there too — details in the Performance Max guide for ecommerce.
A practical decision path
If you are still on the fence: not enough data, start with Maximize; enough data and clearly variable order values, default to Target ROAS; selling something with highly uniform value where you just want to cap acquisition cost, consider Target CPA. Start conservative and tighten in small steps.
Whether the whole effort is worth it for your store is a fair question to answer first — see is Google Ads worth it for ecommerce.
Frequently asked
Can I use Target CPA and Target ROAS in the same campaign at once?
No. A campaign runs one bidding strategy at a time. Most accounts use different strategies across different campaigns — for example Maximize on a prospecting campaign and Target ROAS on the core conversion campaign.
Can I set Target ROAS on a brand-new campaign?
Not recommended. With thin conversion data, target-based strategies tend to swing. It is steadier to run a Maximize strategy for two to four weeks to build data, then layer the target on.
What Target ROAS number should I set?
There is no universal number. The sound approach is to look at your current actual ROAS and nudge up from there (no more than 15 to 20 percent at a time), rather than setting an ideal figure. Too high and it strangles volume.
My conversion values are similar — which one?
If each order is worth roughly the same, Target CPA is usually more direct; it caps cost per unit and suits uniform-value conversions. Wide value variation is what makes Target ROAS or Maximize Conversion Value worth it.
How long before a target change shows results?
Give it at least two weeks. Smart Bidding re-enters a learning phase after a change, and judging too early lets short-term noise trick you into constant back-and-forth tweaking.
GrowthGPT uses multi-source data to plan budget, bids and scaling — a campaign plan you can execute today.
Leads EshopPick's product-research and data desk. Focuses on TikTok Shop US sourcing frameworks, fee-and-profit math, and platform comparisons. Every take is grounded in our weekly real-sales data and Opportunity Score — practical calls, not chart-chasing.