ROAS Google Ads: Formula, Benchmark And How I Use It

The question I get asked most often about return on ad spend is whether a given figure is good. Someone reports 4.2 and wants to know if that's a win. I can't answer it, and neither can anyone else who doesn't know what the business makes on each order.

Return on ad spend ROAS is the ratio between the revenue generated by your ads and your total ad spend. That definition is the easy part, and it's where most explanations stop which is precisely why the metric ends up decorating slide decks instead of shaping a marketing strategy anyone can defend.

So this piece does three things. It defines roas google ads properly and shows where the figure lives in the interface. It replaces the benchmark question with the margin calculation that actually answers it. And it covers the part I care most about: which decisions this number is qualified to inform, and which ones it quietly corrupts when you let it drive them.

What roas google ads means and where to find it

Return on ad spend is conversion value divided by cost. Spend €1,000, record €4,000 of conversion value, and your return is 4 or 400%, depending on which convention you're using. Google reports it as a ratio in the interface and expresses targets as a percentage in the bidding settings, which trips up more people than it should. A target of 400% and a reported figure of 4.0 describe the same thing.

The roas formula is that simple, and it stays simple at every level of the account. What changes is the denominator you choose: one keyword, one google ads campaign, or the total ad spend across everything running.

In your google ads account, the column is Conv. value / cost. Add it from the columns menu on any table campaign, ad group, keyword, product group, asset group and you have return at that level. There's also an Average target ROAS column that shows what your bid strategy was optimising toward rather than what it delivered, and comparing the two is one of the quickest diagnostics available. Google's documentation on Target ROAS bidding is explicit that the system predicts future conversions and their values from the values you report through conversion tracking, then sets maximum CPC bids to chase your target. It also notes that the average target ROAS column is traffic-weighted across the changes you've made worth knowing before you compare it to a simple period average.

Worth separating two things that share a name. The reported column is a measurement of what happened. Target ROAS in Google Ads is an automated bidding strategy: when you set a target ROAS, the system treats that number as an instruction and declines auctions it predicts will break it. A target ROAS bid is therefore a constraint on future behaviour, not a summary of past performance, and confusing the two is how people end up arguing about a figure that describes two different things.

The dependency nobody mentions first

Everything above rests on conversion values arriving in the account, correctly, per transaction. If your ecommerce tag passes the actual order total, the column is meaningful. If it passes a fixed number and I've found this in more accounts than I expected, usually a placeholder set during setup and never revisited then what you're reading is conversion count wearing a currency symbol.

The same applies to anything derived from that data: conversion rates that look implausible are a tracking symptom, not a triumph, and they distort every downstream figure including this one.

The test takes five minutes: sort your conversion actions by value and look at the distribution. If every conversion carries an identical value in a business where order sizes obviously vary, stop reading reports and go fix the tracking. There's no interpretation of a broken input that produces a valid conclusion, and the whole conversion tracking setup deserves an audit before any return figure gets quoted to anyone.

Revenue is not margin, and the column knows nothing about the difference

What is a good ROAS for Google Ads?

The honest answer: whatever exceeds your break-even, by enough to cover the costs the platform can't see and leave a profit you're satisfied with. Every published benchmark I've seen is a distribution of accounts with margins you don't share, and treating one as a target is how businesses end up chasing a number that's either trivially easy or structurally impossible for them.

All useful roas calculations start from the same input: your profit margin. Without it there is no such thing as a good roas for google campaigns, only a number floating free of the economics it's supposed to describe.

Calculate break-even first

Divide 1 by your gross margin as a decimal. Forty percent margin gives a break-even of 2.5 below that you're paying to lose money. Twenty percent margin gives 5.0. Seventy percent gives roughly 1.43.

Sit with what that implies. A business at twenty percent margin needs 5x just to stand still, so the "disappointing" 4x it reported all quarter was a slow leak. The figure was never comparable across businesses. Only the ratio to break-even is.

Then subtract what the platform can't see

Why the industry benchmark doesn't help

Every industry benchmark I've seen reports an average roas across businesses whose cost structures have nothing in common with the one in front of me. Retail averages blend seventy-percent-margin accessories with fifteen-percent electronics. Adopting one as a target means importing someone else's break-even and calling it your own.

There's also a framing problem. This metric is not return on investment it ignores every cost outside the ad account, including the salaries of the people running it. A campaign can clear its break-even and still be a poor investment once fulfilment, tooling and management time are counted. I keep the two questions separate and answer them for different audiences.

The blended trap

One clarification that saves arguments: the account-level figure is a weighted average dominated by whatever spends most. Brand campaigns typically report spectacular returns because those buyers were coming anyway. Prospecting reports far less because it's doing the actual acquisition work.

Average them and you get a number that flatters the account and hides the campaign doing the real job. When people ask what is a good ROAS for Google Ads and quote an account-level figure, my first question is what share of that spend was brand and the answer usually reframes the conversation entirely.

Using it for decisions instead of reporting

A vanity metric is one you report but never act on. Return on ad spend becomes a vanity metric the moment it appears in a monthly deck and nowhere in a decision. Here are the four decisions I make from it, and I try not to make others.

Decision 1 Budget allocation between campaigns

This is where it works best. Comparing two campaigns with the same margin structure, the higher-returning one should get more budget until its return degrades to match the other. That's it. Marginal return, not average return, is what you're chasing and it degrades as you scale, always, because you're buying progressively less qualified traffic.

The mistake I made for a long time was reallocating on averages. A campaign at 6x doesn't stay at 6x when you double its budget it buys the next tier of traffic, which is worse by definition. Move budget in increments, watch what the return does, and stop when the incremental spend stops paying for itself. The number tells you which direction to move; it never tells you how far.

Decision 2 Setting and adjusting a target

This is the decision where the metric feeds directly back into the bid strategy driving the campaign. Set the target to what the campaign is already achieving, then move it in ten to fifteen percent steps, one step per learning cycle. An aspirational target doesn't make the system work harder it makes it bid on fewer auctions, which cuts volume and starves the data it learns from.

Raising a target always costs volume. That trade is fine when the account is efficiency-constrained and wrong when it's growth-constrained a business decision rather than an account one. The mechanics of improving return without simply shrinking the campaign are a longer topic, and the choice between value-based and cost-based optimisation sits inside the wider framework for picking a bid strategy.

Decision 3 Product and category prioritisation

On Shopping and Performance Max campaigns, segmenting return by product category surfaces things nobody expected: high-margin lines quietly subsidising low-margin ones, a category that has never once returned its spend, a single SKU carrying the account.

Decision 4 When to stop optimising the account

If a campaign sits comfortably above break-even and has been stable for months, the highest-value work is usually not in the bidding settings. It's in feed quality, landing pages, offer competitiveness or the margin itself. I've stopped expecting target tuning to rescue accounts where the constraint is somewhere else, and saying so out loud is more useful than another fortnight of adjustments.

What I refuse to decide from it

Whether to pause a campaign inside its learning period. Whether a week was good. Whether a keyword deserves to exist, when it sits early in a path that closes elsewhere. Anything at all in a lead-gen account where the values are estimated rather than transacted there the figure measures the quality of your value estimates, not the quality of your advertising, and I've seen teams optimise their estimates without noticing.

What I infer from the reported figure

When I open an inherited account, the return column is a read on the tracking underneath it before it's a read on performance. Each of these is a pattern I've seen repeatedly, not a rule.

An implausibly high figure 20x, 40x on a non-brand campaign: something is double-counting. Usually the same purchase recorded by two tags, or a thank-you page visitors reload. I check value per conversion against a plausible average order value before anyone gets excited.

Identical values across every conversion: value-based bidding in name only. The account is running Target ROAS while optimising cost, and the label has been misleading everyone for months. Nobody lied. Somebody just never checked what was arriving.

Account-level return healthy, non-brand return underwater: brand is carrying the account. Extremely common, rarely visible in the reporting the client sees, and it means the acquisition engine isn't working while everyone believes it is.

Reported return far below the average target: the target is aspirational. The campaign is almost certainly limited by something other than budget, delivering a fraction of available volume while everyone concludes the market is expensive.

A figure that swings violently week to week: the campaign is too thin for value-based bidding. A handful of large orders is moving the average. There isn't enough data for the system to learn from, and there isn't enough for you to conclude anything either.

Return improving while revenue falls: someone raised a target and called it optimisation. Efficiency bought by shrinking the campaign. Occasionally the right call, but it should be a stated decision rather than a discovery three months later.

Nobody in the business can state the gross margin: the number has never been used for anything. This is the one that tells me most. If the margin is unknown, break-even is unknown, and every judgement made about performance so far has been vibes.

What I stopped doing

Quoting industry benchmarks. They average businesses whose economics have nothing to do with the one in front of me. I ask for the margin instead, and if nobody knows it, that's the first task.

Judging campaigns on the account-level average. It's dominated by brand and it hides the campaign actually doing the acquiring. Segmented or nothing.

Reallocating budget on average rather than marginal return. Doubling the budget of a high-returning campaign does not double the returns. It buys the next tier of traffic, which is worse by definition.

Reading weekly figures in low-volume accounts. Below roughly thirty conversions in the window, one large order rewrites the story.

Optimising toward it in lead gen without auditing the value model. If the values are guesses, you're optimising toward the guesses. I now check how each value was derived before trusting a single figure built on them.

Treating it as the goal rather than a constraint. Maximising return is trivial spend nothing but brand budget and report 15x. The actual goal is profit at volume, and this metric is the efficiency boundary you operate inside, not the thing you're trying to make as large as possible.

The practical takeaway

Get the margin before you get the benchmark. One divided by your gross margin is your break-even, everything above it is profit, and no published average can tell you what any of that is for your business. That single calculation resolves most of the arguments people have about this metric.

Then segment before you judge. Brand and non-brand are different businesses sharing one account, and averaging them produces a figure that describes neither. Confirm the values are real, confirm they're differentiated, and only then read the column.

And use it to move something. A roas google ads figure that appears in a report and never changes a budget, a target or a product decision is decoration. The number earns its keep when it's the input to a decision somebody makes and when the decision is one it's actually qualified to inform.