Your tROAS campaign spends the whole budget and misses the target. You raise the target, spend drops to nothing. You lower it, and suddenly you're paying for sales you don't want. It feels broken. Here's what's actually happening.
What the target actually does: it throttles the bidder
When you set a target ROAS, Google's smart bidding tries to hit that number, not maximize your volume. Read that again, because most people never do. The bidder has one job, and it isn't spending your budget.
Set the target too high relative to what the account can actually deliver, and Google gets conservative. It finds a small pocket of conversions that hit the target, then stops spending because pushing further would break the target. Your budget goes unspent, or gets spent on a trickle of expensive clicks. I've seen accounts with Rs.20,000 a day budgets sitting unused because the target demanded a return the auction couldn't reproduce. The money isn't gone. It was never bid.
Set the target too low and Google goes the other way. It spends everything chasing volume and your efficiency drifts. Worse, it drifts gradually, so you don't notice for weeks. Revenue looks big, margin quietly dies.
The target is a governor, and it only works well when it's set near what the account realistically produces. That's the whole mechanic. People call it broken when it's really set at a fantasy.
Why it feels more volatile now
Google's conversion modeling has gotten more aggressive. Modeled conversions, privacy-driven signal loss, and the shift toward PMax all make the reported numbers noisier. A tROAS target that worked eighteen months ago can behave erratically now because the conversions feeding it are less complete. It's not that the setting broke. The data underneath it changed.
Here's why that matters to the bidder. tROAS learns what your expected return per click looks like from historical conversion data. Feed it complete, observable data and it picks targets with confidence. Feed it a mix of measured and modeled conversions, where a chunk of purchases arrive as estimates, and every auction decision carries more uncertainty. The same target that produced steady delivery before now whipsaws, because the input got fuzzier.
On a thin-data account this gets ugly fast. Fifty conversions a month was already borderline. Fifty conversions a month, half modeled, means the bidder is steering from a blurry picture, and tROAS amplifies every blur because it's the strictest bid strategy in the account.
When Max Conversion Value makes sense
Switch to Max Conversion Value when your priority is spending the full budget and growing revenue, and you have room above break-even. It lets Google spend toward the ceiling of what it can find rather than stopping at a target. You trade some efficiency control for volume.
Concretely: your break-even ROAS is 3x, the account runs a blended 4x, and you want to grow. Max Conversion Value lets the bidder take the additional spend even where the return dips toward 3.5x, which is exactly what you want. With a hard tROAS target of 4, that same spend never happens, and "growth" stays theoretical.
Keep tROAS when you have a hard efficiency floor you cannot cross, a specific break-even ROAS that must hold, or a fixed budget that has to return a known amount. tROAS on a tight-margin business is a guardrail, and Max Conversion Value removes guardrails.
The move that fixes both
Set your tROAS target near what the campaign is already achieving, not where you wish it would go. A target slightly above current real performance pushes Google to grow into it without shutting off spend. A target far above reality just starves the account.
Then walk the target in small steps. If the campaign delivers 3.2x and you want 4x, set 3.4. Let it stabilize for two to three weeks, watch volume hold. Move to 3.6. When a step shrinks spend or volume hard, you've found the ceiling the account can support, and you stop there no matter how much you wanted the next number.
And judge both, tROAS and Max Conversion Value, against your break-even ROAS from your gross margin, not a number you saw online. Google will hit whatever target you set, the question is whether that target makes you money. A campaign that spent its full budget at 3.1x on a 2.5x break-even beat a campaign that spent half its budget at a 4x target, and the second one looked better in every screenshot.
Should PMax just replace it?
One question always follows. If tROAS misbehaves on modeled data, why not run Performance Max everywhere and stop fiddling? Run it where you have nothing better and the answer is sometimes yes, but know what you're trading. PMax has the loosest controls of anything in the account. It decides placements, formats, and audiences on its own, and you hand it every channel at once. If and you can't inspect what it's spending on, you can't steer it away from the junk as it drifts, and it does drift, because its objective is conversion value and nothing else.
The workable middle: keep search tROAS on your brand and core non-brand terms where the data is dense and the intent is known, use PMax for the long tail and discovery where you'd otherwise be guessing, and judge all of it against the same break-even. PMax should earn its budget like any other campaign, which means you need somewhere to see its real incremental revenue, brand search overlap included. Look at the search terms report, run the brand-exclusion tests, and only then decide. Handing a whole account to a black box because precise bidding got harder is a bigger swing than any tROAS setting change.
I'm Hirdesh Matai, founder of HDM GRO-ads, the science of e-commerce advertising.