Target ROAS Is Not the Return You’re Asking For

I used to tell clients to set Target ROAS to the return they wanted. I was wrong. Last year I watched a store move its target from 300% to 500% expecting the same revenue at better efficiency. Spend fell by half, revenue fell from about $18k to $11k for the month, and the report still showed a higher ROAS.

The number you type is not an average Google promises to deliver. It changes how much Google will bid for the auctions it predicts are valuable. Raise it, and fewer bids are competitive. That is how a more ambitious-looking target can produce less revenue.

Calling the target a floor is useful shorthand, but only if you know where the analogy breaks. Google aims for the target across the bidding strategy; it does not impose a literal minimum return on every click. Google describes the mechanism: predict conversion value for an auction, then adjust the bid in pursuit of the target. Start there, and the rest of the account behavior makes more sense.

Step 1: Give Google a value it can distinguish

Before Google bids, it estimates what a click in that auction could be worth. In a simplified model, predicted value equals the chance of a conversion times the expected value of that conversion. A click with a 4% chance of becoming a $50 order has an expected value of $2.00. The same query five minutes later, from a different person on a different device, might have an expected value of $0.40 or $6.00. Google describes this prediction step before the bid is set.

That estimate depends on the values you send back. I learned this on a home services account where every lead was passed back as $100, whether it was a $29 filter replacement or a $900 install. The system could see which clicks produced leads. It could not see which leads were worth pursuing.

The same problem appears when a store sends a fixed 50 euro value for every purchase. A 20 euro item and a 300 euro order then look identical. Hand-entered lead values and purchase tags that fire twice can distort the picture too; one audit found that duplicate tracking could lift reported ROAS from 6.0 to 7.0 with no extra sales. Practitioners report the problem with flat values, and that lead-value audit shows why a prettier ratio need not reflect a better business.

Fix the ruler before you argue about the target. Otherwise, you are asking Google to sort auctions using values that cannot tell a good order from a bad one.

Step 2: See what a higher target does to the bid

Once Google has a predicted value, the target affects what it is willing to bid. The useful working model is bid equals predicted value divided by Target ROAS. Under that model, a $50 predicted value at a 500% target produces a $10 bid. Actual Smart Bidding is more involved than one division, but the model shows the direction: hold predicted value steady, raise the target, and the bid falls.

Take a search with $12 in predicted value. At a 300% target, the model gives you a $4.00 bid. At 500%, it gives you $2.40. At 800%, it gives you $1.50. If winning requires a $3.00 bid, the first is competitive and the other two are not. The search did not become less valuable between those calculations. You changed what you were willing to pay.

That is the auction filter behind the word floor. A higher target does not formally bar an auction at the door; it can make your bid too low to win it. Practitioners making 300% versus 600% comparisons see that distinction in the traffic they can capture. I used to describe the higher setting as smarter bidding. For the same predicted value, it is cheaper bidding. The target does not raise the value of a click. It changes the price you will offer for it.

Step 3: Watch the filter reach your revenue line

Now apply those lower bids across a campaign. Auctions with enough predicted value can still be won. Others become harder to win, so spend may fall even when the budget has room. Google warns that a target set too high can limit traffic and keep a strategy from spending its full budget. Practitioners see it as fewer competitive auctions and underspending.

This is why the dashboard can congratulate you while the business gets smaller. The remaining traffic may lean toward searches the system predicts will convert well, such as brand searches and past buyers. A practitioner teardown describes an account left buying only its safest clicks as volume collapses. The reported ratio improves partly because the campaign has stopped buying some of the harder traffic.

Say you spend $20k a month and record $64k in revenue: 320% blended ROAS. You raise the target to 500%. Spend falls to $11k, and the report shows about 460% ROAS on roughly $50k in revenue. The ratio is higher. Revenue is about $14k lower. That may be an acceptable trade if the removed spend was unprofitable, but the ratio alone cannot tell you that.

A higher ROAS on a smaller base is not automatically a win. Before you celebrate, ask what the spend you cut was earning. That question takes us from the average on the screen to the return at the edge of your budget.

Step 4: Judge the next dollar, not the average dollar

Your account screen shows blended ROAS: total recorded value divided by total spend. It answers a backward-looking question about all the auctions you bought. A target change asks a different question: what will the next stretch of spend earn, or what did the spend you just removed earn? That is marginal ROAS.

The last $1,000 you spend need not earn your blended return. As you try to buy more volume, you may reach auctions with weaker predicted returns than the ones you already win. In the other direction, tightening the target may remove those auctions first and leave a handsome average behind. Google cautions against relying on average CPA and ROAS by segment, because an auction in a lower-average segment can still offer the stronger marginal opportunity.

A Campaign A versus Campaign B example makes the trap plain. Campaign A shows 1.5x average ROAS but only 0.8x on its next $20k of spend: that additional $20k brings in $16k of revenue. Campaign B shows a less attractive 1.1x average yet still returns 1.1x on its next increment. The averages point toward A. The marginal revenue points toward B. Neither revenue figure, on its own, establishes profit; product costs still matter. I have made the blended mistake myself, moving budget to the prettier average and watching total profit slip.

So where should the target sit? First establish what an additional sale leaves you after the costs that vary with it. Break-even ROAS equals 1 divided by contribution margin. At a 45% contribution margin, that is about 2.22x. If you want to retain 10 percentage points of sales as profit after ads, the allowable ad-cost share falls to 35%, making your profit-preserving threshold about 2.86x.

Those are different thresholds. At 2.22x, the ad spend has used up the 45% contribution. At 2.86x, the simplified calculation leaves the desired 10 points. A margin breakdown for new and blended customers shows why the threshold can differ within the same account. Discounts, returns, and customer mix affect what you keep.

The math does not hand you a magic target to type in. It tells you what the marginal auctions must earn for the volume to be worth buying. Use margin to set the economic threshold; use a marginal test to find where the next tier sits. Setting 400% because you would like to see 400% in a report does neither.

Step 5: Use the model when delivery changes

Once the prediction, bid, and margin layers are in place, two apparently different problems become easier to read.

A campaign shrinks after you raise its target. Say a campaign spends $15k and records $48k in revenue at a 350% target. Move the target to 550%, and the simplified bid for any unchanged predicted value falls by roughly a third. If spend then settles near $9k and revenue near $47k, the dashboard shows about 520% ROAS. The ratio improved while revenue slipped. I have watched owners celebrate that kind of result before looking at the dollars. Raising the target again to recover revenue only tightens the same filter; practitioners describe the resulting retreat toward brand searches and past buyers. Check whether the auctions you lost were profitable before deciding whether to reverse course.

A new campaign barely spends. A fresh Performance Max or Shopping build with a 600% target and 12 conversions in the last 30 days gives the system little history to learn which auctions carry value. Guidance cited for Search and Shopping puts the figure at around 15 conversions in 30 days before value bidding steadies; Display and Demand Gen have their own minimums. Thin history paired with a demanding target can leave few auctions where the system bids confidently enough to compete. Before changing the number, check both the data going in and the margin you need out.

The reverse mistake is assuming a low target must be unprofitable. At 200%, bids can compete for a wider set of auctions. If a 220% threshold reflects only the first order but measured repeat value changes what those customers are worth, some of that extra volume may pay. Do not assume a repeat rate will rescue it; count the value you can actually measure. A margin-aware bidding rebuild works through that shift from revenue to what the store keeps.

In each case, the target controls bids against predicted value; the business decides whether the resulting volume is worth buying. To connect those two, you need a test rather than another glance at blended ROAS.

Step 6: Buy a small slice of the next tier

You cannot read marginal ROAS directly from the account’s blended column. You have to change the target, observe the additional or removed spend, and compare the value that moves with it. I prefer a controlled experiment over changing an entire account at once. It gives the target change a comparison and makes a shift in traffic mix easier to inspect.

For a campaign with steady volume, use a 50/50 experiment. Keep the control at its current target and lower only the test side by 10% to 15%. Keep the other settings aligned and allow a full reading window. A four-week read on conversion value and impression share is a practical starting point; the point is to avoid treating an immediate delivery change as the final result.

1. Before the test: Check value tracking and conversion volume. The roughly 15-conversion, 30-day Search and Shopping reference point is a useful check before asking value bidding to learn from a thinner record.
2. During the test: Leave the control at its current target and run the test side 10% to 15% lower. Keep the split and other campaign settings steady rather than editing several things at once.
3. At the readout: Compare the difference in recorded value with the difference in cost. Calculate (test value minus control value) ÷ (test cost minus control cost). Treat that as an estimate of marginal ROAS for the spend this change exposed, then check whether the traffic mix and impression share support the reading.

Return to the store spending $20k a month at 320% blended ROAS. Rather than jumping to a 500% target because 320% looks disappointing, test a lower target near 290% first. Suppose the additional $4,000 of spend in the comparison returns $9,200 in value. That tier earns 230% marginal ROAS. Against the 2.86x profit-preserving threshold for a 45% margin and a 10-point after-ads profit goal, it does not meet the goal. Reject the lower target. The test has not proved that 500% is right; you would need to examine the spend removed by a higher-target change before making that call.

Now change only the test result: suppose the extra $4,000 returns $13,600. Marginal ROAS is 340%. Under those same simplified margin assumptions, the additional tier clears the 2.86x threshold. The lower target earns a serious look even though it may make the blended ratio less impressive. This is the work I hand to groas now: keep the bid and budget pressure moving while the actions remain readable. A monthly agency review that stops at the blended average can miss the decision entirely.

That gives you a way to reason through a new case without guessing a fashionable percentage. If a campaign’s ROAS rises after you tighten its target but both spend and revenue fall, first ask which auctions the lower bids stopped winning. Then measure what the removed spend was earning against your margin threshold. If it was below the threshold, the smaller campaign may be healthier. If it was above, the prettier report cost you profitable revenue. Stop asking what average you want Google to show. Ask what the next dollars earn.