How To Improve ROAS Google Ads Without Shrinking The Account

Almost every request to improve return on ad spend arrives with a method already attached: put the target up. Fastest thing to do, visible within a fortnight, and almost never the thing that needed doing.

The distinction I keep coming back to is between improving the ratio and improving the account. Raising a target improves the ratio by declining auctions; the account underneath is identical, just buying less of everything including the profitable part. Six weeks later revenue is down, the return looks better, and nobody is happy.

So this article covers how to improve roas google ads by changing what the campaign buys rather than how much of it. Four levers, in the order I actually pull them, with the accounts where each one mattered and the mistakes that taught me the sequence.

Why raising the target is not improving the return

The mechanics are worth stating plainly because the confusion here is structural, not careless. Target ROAS is a constraint on which auctions the system enters. Google's documentation on Target ROAS bidding says it directly: a target that's too high may limit the traffic your ads receive, and the recommended target should sit at or below what the campaign has historically achieved. It also advises lowering the target when you want more volume and raising it when you want more value per unit of spend which is the trade named out loud.

Raise the target and the system stops bidding on auctions it predicts will fall short. Some were genuinely unprofitable; many were merely below-average and still comfortably above break-even. Both disappear together, because the bidder forecasts a ratio, not a margin. What you get is a campaign with the same waste rate spending less money. Nothing was fixed. Something was hidden.

The test I use before touching a target

If raising the target by fifteen percent costs more revenue than the efficiency gain is worth, the target was never the problem. That's a question about which constraint the business is under, not an account setting. Which is why the levers below share one property: they change the ratio at constant or increasing spend.

The google ads target roas limited by budget mistake

This is the single most misread status I encounter. A campaign runs Target ROAS, the status column reads limited by budget, and the reaction is almost always to raise the target so the campaign "spends more carefully within its budget".

That's backwards, because two constraints are being confused. The label means the campaign is hitting its daily ceiling while the bid strategy still sees auctions it wants at a predicted return above your target. It found profitable demand it cannot afford.

What the status is actually reporting

Two constraints can bind a value-based campaign. The target constrains which auctions are worth entering. The budget constrains how many of the worth-entering auctions get paid for. When you see google ads target roas limited by budget, the second one is binding and tightening the first does nothing to relieve it. You end up with a campaign restricted twice, delivering a fraction of the volume available to it.

The three responses that make sense

Raise the budget, if the reported return is comfortably above break-even. The boring correct answer, and the one nobody wants. A campaign clearing 5x against a 2.5x break-even is refusing profit at the door every day it stays capped.

Lower the target and raise the budget together, if you're growth-constrained. Counterintuitive, and it works: more auctions entered, more data for the bidder, more conversion volume. The ratio falls, the profit rises. Whether that's the right trade depends on the margin math I cover in the ROAS fundamentals piece.

Leave both alone and fix the waste, if the budget genuinely can't move. A capped campaign wasting a third of its budget on the wrong queries has a third more budget than it looks, and recovering it costs nothing.

The variant that catches people out

A campaign can read capped while the target is also unreachable: few auctions entered, expensive wins, daily cap hit early. Here lowering the target genuinely is the fix. You tell the two apart by comparing reported return against average target, and by checking whether daily spend is pinned to the cap or merely spiking.

The four levers that move the number

These change the account rather than its size. I work them in this order, and the order is the point.

Lever 1 Feed quality, on any account with products

Highest-leverage work on ecommerce accounts and the least glamorous. Titles leading with a brand nobody searches. Missing GTINs suppressing eligibility on the highest-intent queries. No product types, so the system has no structure to reason about. Half the catalogue disapproved and nobody in the diagnostics tab.

Fix it and the campaign matches to buying intent instead of browsing intent: same spend, better traffic, higher return. The systematic version is the feed optimisation work most accounts have never had done. I check the feed before a single bidding setting now.

Lever 2 Structure, so the bidder can tell things apart

A campaign mixing a seventy-percent-margin category with a fifteen-percent one asks the bidder to optimise a blended fiction. It hits your target on average by overspending on the cheap products and underspending on the profitable ones, and no target value fixes that because the problem is the averaging.

Splitting by margin band gives you targets that correspond to real break-evens. On Shopping and Performance Max that means separate asset groups or campaigns by product tier; on Search it means separating brand from non-brand, because brand's inflated return otherwise masks everything else. Those buyers were coming anyway, and the bidder can't tell.

Lever 3 Negatives, and the query report nobody reads

The unglamorous one that keeps paying. Every account I inherit spends on queries that were never going to convert: research phrases, job seekers, "free" and "how to" modifiers, and on ecommerce an entire vocabulary of parts and repairs on accounts that only sell whole units.

The point isn't that this traffic converts badly it converts at zero while consuming budget that would otherwise reach the auctions the bidder wanted. On a capped campaign that's a budget increase you don't pay for. Building a proper negative keyword list is a week of work then a monthly habit, and it moves limited campaigns faster than anything else. One caveat learned expensively: below about thirty conversions in the window, judge on spend and query intent rather than conversion count.

Lever 4 Bidding, last and smallest

By now the campaign is buying different traffic and its achievable return has moved, so the settings finally matter. Set the target to observed performance after the changes settle, and move it in ten to fifteen percent steps, one per conversion cycle. The wider choice of bid strategy is worth revisiting once the account has changed shape. What I no longer do is treat bidding as the first move: the bidder optimises within the constraints you hand it.

Three accounts, three different constraints

The homeware retailer that was capped and didn't know it. Shopping campaign at 3.4x against a 2.8x break-even, permanently capped, and six months of target increases from the previous manager. Return had climbed to 3.9x; revenue had fallen by roughly a third. I reset the target to observed performance and cut the parts-and-spares queries eating a meaningful share of spend, leaving the budget alone because the client wouldn't move it. Return reached 4.1x on the same budget within two cycles and revenue came back with it the recovered spend went where the bidder had wanted it all along.

The B2B supplier where the feed was the whole story. Return below break-even, every conversation about bidding. The feed had no product types, brand-first titles on a catalogue nobody searched by brand, and a material share of items disapproved on image policy. Rewriting titles to lead with the category and clearing the disapprovals moved the return more in three weeks than a year of target adjustments had.

The account where I was wrong. A lead-gen client reporting 6x, wanting better. A month on structure and negatives moved almost nothing: the lead values were estimates set eighteen months earlier and never reconciled against closed deals, so I'd been optimising the quality of someone's guesses. Feeding closed-won values back reframed everything, and several campaigns the estimates had flattered turned out to be the weak ones.

The order I work in, and why

Verify the conversion values are real and differentiated if they're estimates or flat placeholders, nothing downstream can be trusted and this is the only task. Then read the status column and diagnose the constraint: capped, target-restricted, or both. Most accounts get misdiagnosed right here.

Fix the feed if there is one. Separate what shouldn't be averaged brand from non-brand, high margin from low. Cut the waste through query reports and negatives, as a recurring habit rather than a project.

Then, and only then, set the target to what the changed account is achieving. Everything before this raised the ceiling; this step only decides how much of the new headroom you take as efficiency and how much you leave as volume. It's a business decision wearing an account-settings costume.

The practical takeaway

If the campaign reads limited by budget and its return clears break-even, the answer is more budget, not a higher target. Everything else is a workaround for a decision somebody didn't want to make, and the workarounds cost real revenue.

If the budget can't move, spend your effort on what the campaign buys: feed, structure, negatives, in that order, before bidding. Those levers raise the achievable return; the target field only decides how much of it you collect.

The whole of how to improve roas google ads reduces to one question, asked honestly: am I making the account better, or making it smaller? Both move the number in the right direction. Only one of them survives contact with the revenue line six months later.