Why Your Google Ads ROAS Is Lying To You - And What To Measure Instead

ROAS is the most widely reported metric in ecommerce Google Ads. It is also, in many accounts, the most misleading. A 4x ROAS sounds like a success story. At a 20% gross margin, it is a loss. This is not a fringe case. It is happening in accounts across the industry right now - while the monthly report says "strong performance."

The problem with ROAS

Return on Ad Spend is calculated as revenue divided by ad spend. If you spend £10,000 and generate £40,000 in revenue, your ROAS is 4x. Simple. Clean. Easy to report.

It is also, in isolation, commercially meaningless.

ROAS measures the efficiency with which your ad spend generates revenue. What it does not measure - what it has never measured - is whether that revenue is profitable. It contains no information about your cost of goods. No information about your fulfilment costs. No information about your return rate. No information about the margin that sits between the revenue number and the cash that actually reaches your business.

This distinction matters enormously. Consider two brands, both running Google Ads at 4x ROAS.

Brand A operates at 40% gross margin. Their break-even ROAS - the point at which ad spend breaks even against the gross profit it generates, using gross margin as the simplified proxy - is 2.5x. At 4x ROAS, they are comfortably profitable on their ad spend. They have room to scale, room to discount, room to absorb a spike in CPCs without the economics collapsing.

Brand B operates at 20% gross margin. Their break-even ROAS is 5x. At 4x ROAS, they are losing money on every single order generated by Google Ads. The revenue is going up. The cash position is getting worse. And the monthly report is saying "strong performance."

Both brands are hitting 4x ROAS. One is building a business. The other is funding a very expensive illusion.

Where does break-even ROAS come from?

Break-even ROAS is the ROAS at which the contribution generated by ad spend exactly covers the ad spend itself. Below it, you lose money on every Google Ads conversion. Above it, you generate contribution. The simplest version of the formula uses gross margin as the proxy:

Break-even ROAS = 1 ÷ Gross Margin %

• At 40% gross margin: break-even ROAS = 2.5x
• At 30% gross margin: break-even ROAS = 3.3x
• At 25% gross margin: break-even ROAS = 4.0x
• At 20% gross margin: break-even ROAS = 5.0x
• At 15% gross margin: break-even ROAS = 6.7x

These figures use gross margin as a first proxy. An operating break-even uses pre-ad contribution instead: gross profit minus the other variable costs that move with each retained order, such as payment fees, pick, pack, delivery and returns handling. On that stricter basis the formula is 1 ÷ pre-ad contribution margin, and the result is always higher than the gross-margin proxy. The worked example later in this article uses the stricter basis.

This number should be the foundation of every Google Ads target you set. It is the commercial floor beneath which your advertising strategy loses money regardless of how good the ROAS looks in the report.

Plenty of ecommerce brands running Google Ads do not know this number, because it has never been calculated for the account. Their ROAS targets were set in an onboarding call based on what sounded commercially reasonable - or, more commonly, what the client said they needed to hit to justify the spend - and have not been revisited since.

This is not a minor oversight. It is one of the central failures in Google Ads management.

The discount problem

The gap between ROAS and profitability becomes most dangerous - and most visible - during sale periods and promotional events.

Consider a brand with a standard 40% gross margin running a 20% discount during a promotional event. The discount compresses their margin from 40% to 25% overnight. Their break-even ROAS moves from 2.5x to 4.0x.

If their agency's ROAS target is set at 3x - which looked healthy under normal trading conditions - they are now running a loss-making campaign during their highest-spend period of the year. CPCs spike during sale periods as every competitor enters the auction simultaneously. So their cost per conversion goes up at precisely the moment their margin per conversion goes down.

The account hits 3x ROAS. The report is sent. The agency calls it a successful peak period.

In January, the brand wonders where the cash went.

Why agencies don't tell you this

The question worth asking is: if this is so foundational, why do so few agencies ever raise it?

Part of the answer is structural. Many agencies charge a percentage of ad spend. Their fee increases when the budget increases. Their fee is unaffected by whether the spend is profitable. The metric that governs their commercial success - client retention - is served by monthly reports that look good and clients who feel the account is being managed. ROAS is a metric that can be made to look good without any connection to the client's actual unit economics.

Part of the answer is competence. Calculating break-even ROAS and managing to it requires understanding the client's P&L, building it into the account structure, revisiting it when margin changes, and applying it at product level rather than as a blended account target. This is harder than moving a ROAS slider. Many agencies have not built the capability to do it properly.

Part of the answer is the pitch dynamic. Agencies win business by presenting confident strategies and ambitious targets. "We need to understand your gross margin before we can tell you what a profitable ROAS target looks like" does not win pitches the way a slide deck with projected revenue figures does. So agencies stop asking. And the question disappears from the conversation.

What remains is a very detailed report about a metric that does not tell you whether your advertising is making you money.

POAS: the metric that actually matters

Profit on Ad Spend (POAS) is the metric that connects your advertising directly to your commercial performance. Where ROAS divides revenue by ad spend, POAS divides contribution by ad spend. In its simplest form the contribution figure is gross profit, which is a proxy; the stricter version is pre-ad contribution after other variable costs such as payment fees, fulfilment and returns handling.

POAS = Contribution ÷ Ad Spend

A POAS above 1.0 means your ad spend is generating more contribution than it costs. A POAS below 1.0 means every pound spent on advertising is destroying margin. A POAS above 1.0 is still not net profit: overheads, salaries, agency fees, software and tax all sit below this line.

This is the number your agency should be targeting. Not ROAS. Not blended revenue. Not impression share or CTR or quality score. The ratio of contribution to ad cost, measured at campaign level, at product category level, and ideally at SKU level for your highest-spend lines.

The transition from ROAS to POAS requires one input that agencies often never ask for: your margin data. Not estimated. Not blended across the entire business. Actual contribution margin by product or product category, updated when your costs change, applied specifically to the campaigns and product groups that are being optimised.

With this data, the account can be structured around profit rather than revenue. High-margin products get aggressive targets. Low-margin products get tight floors or exclusions. Promotional periods get recalibrated targets that reflect the compressed margin of the sale. The account stops generating impressive ROAS and starts generating actual profit.

A worked example: two product groups at the same 5x ROAS

The argument above is easy to accept in principle and easy to ignore in practice. So here it is as arithmetic you can reproduce. Two hypothetical product groups, A and B, over one trading period. Both are given identical net revenue and identical ad spend, so both report exactly the same ROAS. These are illustrative figures chosen to be easy to follow. They are not client data, not an average, and not a benchmark.

Break-even ROAS is one divided by the contribution margin. For Group A that is 1 ÷ 0.45, or 2.22x. For Group B it is 1 ÷ 0.20, or 5.00x. Both groups report 5x. Group A is running at more than double its break-even and leaves £250. Group B is sitting exactly on its break-even and leaves nothing at all. A single blended ROAS target across the two would have shown one healthy-looking number and concealed this completely.

Post-ad contribution here is before overheads, salaries, agency fees, software and tax. It is contribution, not net profit, and it should never be reported as profit.

A sensitivity scenario, not a prediction

Hold the total ad spend at £400 and ask what the same period would look like under a different split, holding efficiency at 5x and both margin structures unchanged. Equal spend of £200 each produces £250 in total. A hypothetical split of £300 to A and £100 to B produces £375: Group A generates £1,500 net revenue at 5x, £675 of pre-ad contribution at 45%, leaving £375 after its £300 of ad cost, while Group B generates £500 net revenue, £100 of pre-ad contribution at 20%, and again leaves nothing. Same total spend. Same blended reported ROAS. £125 more contribution.

You can run the same arithmetic on your own product groups with the break-even ROAS calculator and the POAS calculator, or work through it on paper.

What this looks like in a published client case

The example above is hypothetical by design. The nearest published client account of the same problem is our Thermos case study, where an audit found conversion tracking that was double counting, overstating reported ROAS by roughly 40%. Correcting that is a reporting change rather than a commercial gain, and the case page separates the two. It also reports a 94% increase in contribution margin and an 86% increase in revenue against the preceding equivalent period after deduplication, across a five-month engagement.

Those are case-reported figures from a single client engagement, measured against a preceding period. They are not an independently audited result, not a controlled test, and not evidence of what any other account would do. Period-on-period comparison carries whatever seasonality and trading conditions sat in those months. The case page states its own measurement basis and limits; read them there rather than taking the headline numbers on their own.

What SKU-level profitability actually looks like

The most common version of this problem in practice: a brand with a catalogue of several hundred SKUs, managed through a single Shopping campaign or a consolidated Performance Max campaign, with one ROAS target applied across every product regardless of margin.

In this setup, Google's algorithm optimises toward products with the strongest conversion signal - typically the brand's best-selling SKUs, which have accumulated the most historical data. These products receive the majority of the budget. They convert well. The ROAS looks strong.

But best-selling SKUs are not always the highest-margin SKUs. In many catalogues, the products that sell most reliably are mid-range items with moderate margins. The high-margin products - often newer lines, more premium price points, or less established categories - have less historical data, receive less budget, and are progressively starved of signal by an algorithm that rewards conversion volume rather than profit contribution.

The result is an account that generates strong ROAS by systematically advertising the wrong products at scale. The revenue looks good. The margin quietly erodes. And the management fee arrives on the first of the month regardless.

Fixing this requires margin data applied at SKU or product category level, campaign segmentation that separates product tiers by profitability, and bid targets derived from break-even ROAS for each tier rather than a single blended target across the account.

The return rate adjustment

There is a further adjustment that many agencies never make: return rate.

If your account reports a conversion when an order is placed - as most do - your ROAS includes revenue from orders that will subsequently be returned. Return rates vary widely by category, price point and fit complexity, so use your own figure by product category rather than any industry average. To make the arithmetic concrete, take a hypothetical brand where 25% of booked revenue reverses through returns.

With a 25% return rate, a campaign showing 4x ROAS on booked revenue is performing at 3x on net revenue, because a quarter of the booked revenue never becomes cash. The two bases must not be mixed: a target expressed on net revenue has to be compared against ROAS measured on net revenue, and the same applies on booked revenue.

The correct calculation keeps the two revenue bases separate and never counts returns twice. Suppose the contribution margin on retained orders is 40%, and returned goods come back into saleable stock, recovering their full COGS with no handling loss. Then the economics of a kept order do not change: £100 of booked revenue becomes £75 of net revenue, and that £75 carries £30 of pre-ad contribution. The margin on net revenue stays 40%, so the break-even ROAS on net revenue stays 2.5x. What changes is the ratio expressed on booked revenue: £30 of contribution against £100 booked is 30%, so the break-even ROAS on booked revenue is 1 ÷ 0.30, or 3.33x. These are the same economics stated against two different revenue bases, not two different answers.

Returns handling, restocking and write-offs are real costs, but they are additional measured costs that belong in other variable cost, where they reduce the contribution margin itself. The error to avoid is deducting returns from revenue and then applying a second returns allowance inside the cost lines. That counts the same loss twice and overstates the break-even figure.

Making the transition

Transitioning from ROAS-based management to POAS-based management is not complicated. It requires three things that most brands already have access to, plus one technical step: the contribution values have to reach the conversion tracking the bidding system actually optimises against, which is covered in what profit-based bidding actually requires.

First: Your margin by product or product category. Gross margin is the starting point; contribution margin after other variable costs is the operating figure. This should come from your P&L or your Shopify cost data. It needs to be accurate, not estimated, and it needs to reflect your actual COGS rather than your retail margin. When supplier costs change, this number needs to change with them.

Second: Your return rate by product category. This is available in your Shopify analytics. It needs to be segmented by category because return rates vary significantly - your footwear line may return at 35% while your accessories return at 8%.

Third: Campaign segmentation that maps to your margin tiers. High-margin products in one group with an aggressive POAS target. Mid-margin products in another with a moderate target. Low-margin products either excluded or given a hard floor below which spend stops.

With these three inputs, the account can be managed to commercial reality rather than to revenue optics. The reports will look different. The ROAS figures will often be lower, because you have removed the low-margin conversions that were pulling it up. But the profit per order will be higher, and the cash position will improve.

That is the point of the exercise. Not impressive numbers in a monthly report. Actual profit from your advertising spend.

The honest conversation

Many agencies will not initiate this conversation. The reasons are structural, commercial, and - in many cases - a reflection of genuine gaps in capability. Asking the margin question, building it into the account, and reporting on profit rather than revenue requires more work, creates more difficult conversations, and generates reports that are harder to celebrate.

It also produces better outcomes for the brands involved.

If you have been running Google Ads without a break-even ROAS calculation, without margin data in your account structure, and without product-level profitability reporting - start there. Calculate the number. Share it with your agency. Ask them to recalibrate every target in the account from it.

Their response will tell you a great deal about the quality of what you have been receiving.