Maximise Conversion Value Google Ads vs Target ROAS

The question I get asked is almost always framed wrong. People ask which of the two smart bidding options is better, when the two aren't really competitors one is the other with a constraint attached. Once that lands, the decision stops being a preference and becomes a question about what the account can currently prove.

I've had this conversation most often with ecommerce clients who switched to a target because a case study told them to, then watched volume collapse and blamed the algorithm. Google s algorithm did what it was told. The instruction was just written by someone who didn't have the number yet.

What follows is how I actually decide: the structural difference in one paragraph, the specific conditions under which I leave value uncapped, the conditions under which I clamp it, and the failure mode that neither setting protects you from an account happily buying revenue that loses money on every order.

Maximize conversion value vs target ROAS: what actually differs

Google's own documentation makes the relationship explicit rather than leaving it to interpretation. In About Maximize conversion value bidding Google states that without a target ROAS set, the strategy tries to fully spend your average daily budget, treating that figure as an instruction rather than a limit, warns that this can significantly increase spend if you were previously underspending, and notes plainly that the same strategy with a target return on ad spend attached behaves like a Target ROAS strategy. From June 2026 the labels themselves are converging: "Maximize conversion value with a Target ROAS" is simply being renamed Target ROAS, with no change to the underlying behaviour.

So there is one bidder and two operating modes, and which mode you pick decides what the strategy focuses on when it goes to optimize bids in each auction. Uncapped, the budget is the binding constraint and value is what gets maximised inside it. Capped, efficiency becomes the binding constraint and spend is what flexes the system keeps adjusting bids downward until it only buys what it thinks it can buy at your ratio, which is why volume drops the instant a target is set too high.

Why that reframing changes the decision

If the two settings were different algorithms you'd pick one on philosophy. Because they're one algorithm with an optional ceiling, the question is only ever: do I have a credible number to put in the ceiling? A target derived from the account is information. A target derived from what the business wishes its return were is a constraint on nothing but volume the bidder can't conjure efficiency that isn't available in the auction, it can only decline to buy, and declining to buy is exactly what you'll see.

The volume-versus-efficiency trade is not symmetrical

This is where most comparisons of maximize conversion value vs target roas go soft. They present a clean trade: accept less volume, get better return. In practice tightening a target usually improves reported ROAS while shrinking total profit, because the campaign gives up marginal orders that were still contributing above variable cost. Whether that trade is good depends entirely on your margin structure, which sits outside the roas goals you can express in the interface. Under maximise conversion value google ads without a cap you at least see the whole demand curve before deciding where to cut it.

Both modes need the same foundation

Neither works on binary conversions. If every conversion is worth the same to you one lead, one call, one booking you have nothing to optimise value against, and the number of conversions is the better objective: max conversions, or target cpa if you also need an efficiency ceiling. The value strategies earn their keep only when your outcomes genuinely differ in worth. It's worth reading how the count-based equivalent handles its budget, because the spending mechanics are identical and the failure modes rhyme.

When I leave value uncapped

Three situations, and in all three the reason is the same: I don't yet have a number I'd be willing to defend.

New accounts, or accounts that just started sending values

If conversion tracking only started passing revenue last month, no historical ROAS exists that means anything. Google's guidance is to wait roughly four weeks or three conversion cycles after changing how value is reported before adopting a value strategy and I'd add that setting a target during that window is worse than useless, because you're constraining the campaign against a figure produced by incomplete data. I run uncapped for three to four weeks specifically to generate the number.

Accounts where the target was invented and volume died

The recovery pattern is consistent. A campaign sitting well under budget with a target it never reaches isn't being efficient, it's starved: it can't learn because it isn't buying, and it isn't buying because the target says the available inventory is too expensive. Removing the target for two or three weeks restores spend, restores data, and produces an observed ROAS I can then set a target from. The learning period is far shorter when nothing is throttling entry in the first place.

Fixed budget commitments and promotional windows

Retainer accounts with a monthly spend commitment, seasonal pushes, clearance periods. Here full budget utilisation is the objective and return is the tiebreak, which is precisely what uncapped value does by design. A target in this context can leave money unspent at month end, which for these clients is the actual failure.

What I watch while it runs uncapped

Not daily ROAS that's noise for the first fortnight. I watch the distribution: which products, which value bands, which query groups the spend is migrating toward. That distribution is the real output of the exploration phase, and the conversion rates behind each band tell you more about what the target should be than the headline ratio does.

When I attach a target ROAS

I constrain when the account has produced a number, and when the business has a genuine ceiling rather than an aspiration.

The readiness test

Three conditions together, not any one alone: at least three or four weeks of stable value reporting; enough conversion volume that the weekly ROAS figure isn't swinging by half; and a break-even ratio the client can actually articulate from their own margins. If someone can't tell me their contribution margin, we're not ready to set a target, we're ready to calculate one.

How I set the first number

I set the first roas target at or slightly below observed ROAS, never above. If the campaign has been returning 340% uncapped, I'll set 320-340% and let it stabilise the goal of the first target is to hold current performance while adding a floor, not to improve anything. Then I tighten in roughly ten percent steps, one step per conversion cycle, and stop the moment volume starts falling faster than efficiency improves. That stopping point is the real answer, and it's usually lower than what the client hoped for.

The cases where I constrain immediately

Thin margins where a bad fortnight has real consequences. Inventory-constrained businesses that can't fulfil unbounded volume. Accounts where the client checks daily and reacts the uncapped exploration period is technically survivable and politically frequently not, and a campaign that gets paused on day five because of a scary number has learned nothing either way.

Portfolio-level constraints

When several campaigns share an efficiency goal, a portfolio bid strategy lets the system move spend between them toward the shared ratio, rather than forcing each campaign to hit it in isolation. On accounts with one strong campaign and two mediocre ones this usually beats setting three separate targets, because the strong campaign is allowed to carry the average.

How I stop the algorithm chasing value without profit

This is the failure neither setting protects you from, and it's the one that costs real money. The bidder maximises the number in the value field. If that number is revenue and your margins vary across the catalogue, it will systematically buy the wrong orders and both uncapped and target modes do this equally, because the input is wrong in both.

What it looks like in the data

Reported ROAS holding steady or improving while the business gets no better. I've audited accounts hitting a comfortable target where the actual contribution was negative, because spend had concentrated on high-ticket low-margin lines and drifted away from the accessories that carried the margin. Nothing in the interface flags this. The campaign is doing exactly what was asked.

Fix one: send margin, not revenue

The structural solution is to feed profit rather than turnover the documentation is explicit that the value you define can be profit margins rather than sales revenue. Passing revenue multiplied by product-level margin, or a category-level approximation where per-SKU margin isn't available, changes what the bidder optimises for at the root. Even a rough three-tier approximation beats raw revenue, and it's the single highest-leverage change I make on ecommerce accounts.

Fix two: adjust for what happens after the sale

Returns, cancellations, subscription churn, and first-order-versus-repeat differences all sit outside the platform. Discounting values for known return rates by category, or uploading corrected values later, keeps the model from optimising toward a segment that looks excellent at checkout and terrible sixty days out. Where the real outcome lands offline entirely, feeding it back with offline conversion values is the only way the bidder ever learns which of those clicks were worth buying.

What doesn't fix it

Raising the target. If the value definition is wrong, a tighter ratio just makes the campaign more efficient at buying the wrong thing you get a better ratio on a worse mix. I've watched accounts tighten three times chasing profitability that the target field cannot deliver, because the problem was never the constraint. It was the number being constrained.

The diagnostic I run

Export spend and revenue by product or category, apply real margins outside the platform, and compare contribution against platform ROAS. If the ranking of your best segments differs between the two views, your value definition is misleading the bidder and no bidding setting will correct it.

The sequence I use

Order matters more than any individual step. Get the value definition right first margin-adjusted where possible, returns accounted for because everything downstream optimises against it. Then check the budget is defensible: uncapped, the strategy will spend it, so if you can't say roughly what revenue that budget should produce, you're not ready to hand it over.

Run uncapped for three to four weeks. Don't cut the budget mid-learning; that restarts exploration on worse data and is the most common way these campaigns get broken. Read the observed ROAS and, more importantly, the spend distribution across value bands.

Then decide. If the business has a real efficiency ceiling, attach a target at or just below observed, and tighten in small steps with a conversion cycle between each. If the business is budget-constrained rather than efficiency-constrained, leave it uncapped and manage the mix through the value data instead. Revisit quarterly margins move, and a target set against last year's cost structure is just as invented as one set from a guess.

The practical takeaway

They aren't two strategies. They're one bidder with an optional efficiency ceiling, and the only question worth asking is whether you have a number honest enough to put in it. Uncapped when you're learning what the account can do; capped when the business has a real constraint and the account has produced a real figure.

And underneath both: the strategy maximises whatever you called value. If that's revenue and your margins vary, it will find you the least profitable version of a good-looking ROAS, on schedule, exactly as designed. Fix the value definition before you touch the target field that's the change that moves profit rather than the ratio you report.