Target ROAS vs Target CPA: Order-Value Spread Decides It Before You Do

Target ROAS optimizes for revenue at a return target. Target CPA holds a flat cost-per-conversion regardless of what each conversion is worth. The deciding factor isn't your industry or your spend, it's the variance in your average order value. If your products swing more than 2-3x in price, ROAS recovers profit that CPA spends bidding on cheap conversions. Below that spread, CPA usually wins.

Most "which bidding strategy should I use" advice ends up reading like the menu at a chain restaurant: technically a list, but everything tastes the same. Pick CPA for lead gen. Pick ROAS for e-commerce. Done.

That framing is roughly correct and almost entirely useless. Optmyzr's performance breakdown of Target CPA vs Target ROAS rollouts found something more interesting: ROAS won in roughly 54% of the accounts they tested. CPA won in 46%. The split didn't track to vertical or budget. It tracked to whether the account had genuine variance in conversion value, and whether that variance was being measured cleanly.

Which means the real question is not "ROAS or CPA?" It's "do my conversions actually have different values, and is Google seeing those differences?"

The order-value spread test

Pull your last 90 days of revenue data from Google Ads. Sort by conversion value. Look at the top 20% of conversions versus the bottom 20%. If the top is roughly 2-3x the bottom, Target ROAS has something to optimize against. If the top is closer to the bottom (within 1.5x), CPA is doing the same job with less complexity.

A few examples I've seen play out repeatedly:

• A B2B SaaS account with one $99/mo plan and one $499/mo plan: 5x spread. ROAS, every time.
• A mortgage lead-gen advertiser where every form fill was roughly the same value to sales: 1.0-1.2x spread. CPA, no contest.
• A boutique apparel store with $45 socks and $380 jackets: 8x spread. ROAS, but only after the conversion values were actually being passed through.

That last point is where most accounts fall apart, which we'll get to in a second.

One more wrinkle to flag while we're here. The spread test only works if Google Ads is actually receiving the per-conversion value. A surprising number of accounts I audit are passing a single static value (revenue averaged across all SKUs, or worse, a flat placeholder like $1) which makes the Top 20% / Bottom 20% calculation look perfectly compressed even when the underlying basket is wildly variable. If your Conversions report shows every single conversion at the same value, your tracking is the problem, not your bid strategy.

Target CPA's quiet tax on e-commerce accounts

Here's the failure mode that trips up the majority of e-commerce accounts running on Target CPA: Google optimizes for the cheapest conversion it can find. If your $35 phone case has a 9% conversion rate and your $400 jacket has a 1.8% conversion rate, CPA will reliably deliver phone case sales because they're cheaper to win. The jacket auctions look expensive on a per-click basis. CPA sees them, decides they don't pencil, and walks away.

You hit your CPA target. Your revenue line goes flat. Profit follows.

The Google Ads team documented this dynamic without quite saying it out loud: Target ROAS uses predicted conversion value per impression, while Target CPA only sees conversion probability. ROAS sees a $400 jacket auction and says "this is worth chasing even at a higher CPC because the predicted return is higher." CPA sees the same auction and says "too expensive."

This is why most performance teams I respect have moved their entire Shopping/PMax stack to Target ROAS by default and only fall back to CPA when conversion volume is so thin that ROAS can't learn. We covered the broader version of this in our pillar guide on running Google Ads without letting Google run you. Bid strategy is one of the few levers you have left after Google has automated the others.

Where Target ROAS quietly fails

ROAS is not free. It needs three things to work, and I see two of them ignored constantly.

Reliable conversion values. If you're sending a flat $100 value for every purchase to "estimate" revenue, congratulations, you've built a CPA campaign with extra steps. Target ROAS will optimize toward whatever values you feed it. Garbage in, expensive garbage out.

Enough volume. Google's official guidance says 15 conversions in 30 days for Search. Practitioners recommend 30-50, and an account-level threshold was added recently. 15 conversions across the conversion action in the past 30 days lets any campaign in that account use the strategy. From what I've seen, accounts under 30 monthly conversions struggle no matter what Google's minimum says. The bid system needs enough data points to recognize patterns, and 15 is barely "patterns" at all.

Conversion attribution that doesn't lie. Search Engine Land's deep dive on Target ROAS makes the same point I keep making in audits: if you have stacked conversion actions (form fills, calls, downloads) all flowing into "Conversions," and they have wildly different values, ROAS can optimize against a fiction. Pull the "include in Conversions" and "Primary action" settings before you trust the bid strategy. We unpacked the full version of this in Smart Bidding Is Optimizing for the Wrong Conversions.

The 30-day check before you switch either direction

If you're considering a switch, run this audit before you touch the bid strategy panel.

1. Volume check. Conversions in the last 30 days, by campaign and at account level. Below 30 monthly: don't switch yet, fix volume first.
2. Value spread. Top 20% conversion value divided by bottom 20% conversion value. Above 2.5x? ROAS is the lever. Below 1.5x? CPA. Between, run a portfolio test.
3. Conversion hygiene. Open the conversion actions panel. Anything you don't trust (engagement-based form fills, micro-conversions like add-to-cart, calls under 30 seconds) should not be in "Conversions." If you find junk, fix that before changing the bid type. CPA campaigns running on bloated conversion sets misallocate budget faster than any bid strategy mistake.
4. Tracking method. Server-side conversion or enhanced conversions for leads enabled? If not, Target ROAS is operating with a 20-30% blind spot on attribution and you're going to get inconsistent results.

A specific benchmark: when accounts pass all four checks and switch from Target CPA to Target ROAS on Shopping or PMax, Adalysis's case data and our own audit work suggest a 12-25% lift in revenue at the same spend within roughly 4-6 weeks. Below the four checks, the same switch produces flat or negative results.

What to do if you don't have the data yet

New campaigns or accounts under 15 conversions in 30 days have a third option that most tutorials skip: Maximize Conversions (no target) or Maximize Conversion Value (no target). These don't try to hit a specific CPA or ROAS, they just try to spend the budget on the most, or most-valuable, conversions Google can find.

Run that for 3-4 weeks. Once you have around 30 conversions, look at the actual CPA or ROAS the campaign settled into. That's your target floor. Set Target CPA or Target ROAS at roughly that level, not the aspirational number you wish it would hit. Setting an aggressive target on a campaign that doesn't have the data to support it is the fastest way I know to watch impression share collapse to zero.

This is what most "set Target CPA at $X" advice gets wrong. The number isn't a goal you set. It's a level the data tells you the auction supports.

If you genuinely have to start with a target on day one (often a CFO requirement, sometimes a portfolio constraint), set it at roughly 25-30% above what your category benchmarks suggest and tighten it weekly. Setting the target loose and tightening is recoverable. Setting it too tight from the start is not. The system reads early no-spend signals as "this campaign isn't viable" and it can take 2-3 weeks to dig out once it's spiraled.

Common questions I get asked about this

Can I use both at the same time? Not on the same campaign. You can run Target CPA on lead-gen Search campaigns and Target ROAS on Shopping or PMax in the same account, and that's actually the most common setup I see in healthy accounts.

What about Maximize Conversion Value with a target ROAS? That's the same thing as Target ROAS now. Google merged the two strategies in 2022. Same with Maximize Conversions + Target CPA = Target CPA. Don't get distracted by the dropdown labels.

My agency wants to use tCPA on PMax. Bad idea? For most e-commerce accounts, yes. PMax with Target CPA punishes higher-AOV products in exactly the way described above. The exception is single-SKU advertisers or services with flat pricing. If that's not you, ask them to defend the choice with the value-spread numbers from your own data.

The line that actually matters

Target ROAS vs Target CPA isn't a strategy question. It's a measurement question. The accounts that get this right have done the order-value math and the conversion hygiene before they touch the bid strategy dropdown. The accounts that get it wrong are picking based on whichever option their last consultant recommended, and Google is happily optimizing toward whatever number they put in the box.

The decision is sitting in your conversion data. Pull the spread, fix the conversions, then pick.

Notice Me Senpai Editorial