Beyond ROAS: the Google Ads KPIs a business owner should actually watch

When a client asks "how is Google Ads doing?", the easiest answer would be to read out the ROAS from the platform. It is also the most dangerous one. ROAS tells you how much revenue Google credits itself with for every pound spent: it does not tell you whether that revenue carries any margin, whether it comes from new customers or from people who would have bought anyway, or whether yesterday’s figures are even final. This article lines up the KPIs we use to answer the real question: is advertising making the business money?

ROAS and break-even ROAS: the number that is almost always missing

In Google Ads, ROAS is conversion value divided by cost ("conv. value / cost" in the interface), often shown as a percentage: 400% means £4 of value for every £1 spent. On its own it says nothing, because the same 400% is excellent on a product with a 60% margin and disastrous on one with 20%.

The benchmark that matters is break-even ROAS:

Break-even ROAS = 1 ÷ contribution margin

Contribution margin is what is left of each pound of sales after variable costs: product cost, shipping, payment fees and an allowance for returns. Not the margin on the price list. An illustrative example with an average order of £80:

At a 35% margin, break-even ROAS is 1 ÷ 0.35 ≈ 2.86, or roughly 286%. Below that, every sale Google "brings in" costs more than it earns. If you also want a 10% profit on revenue after advertising, ad spend can take at most 25% of revenue (35% − 10%), and the target ROAS becomes 1 ÷ 0.25 = 400%.

The problem shows clearly when an account sells categories with different margins. Three product lines, same spend (illustrative figures):

The category with the highest ROAS is the only one making nothing. An algorithm running a single target ROAS across the account will push hardest there, because that is where revenue per pound is highest.

POAS and profit reporting

The next step is to stop measuring revenue and start measuring profit. POAS (profit on ad spend) is margin generated divided by spend: in the example above it is 1.0 for category A, 1.75 for B and 1.4 for C. Break-even is always 1, whatever the margin, which is exactly what makes it readable by anyone.

Google Ads can do part of this work, provided two things are in place:

1. Conversions with cart data: the purchase tag sends the products in the order, not just the total.
2. Cost of goods sold in the feed: the cost_of_goods_sold attribute is filled in for each product in Merchant Center.

With both, campaign, ad group and product tables gain columns such as revenue, cost of goods sold, gross profit and gross profit margin. Without COGS, the last two stay blank. Google has also introduced a profit optimisation goal for Performance Max and Standard Shopping built on the same data: at the time of writing it is described as a beta and we do not see it in every account, so check availability case by case.

CPA, target CPA and allowable CPA

If you optimise for conversions without a value — leads, sign-ups, first orders — three numbers get confused all the time:

• Actual CPA: what you paid per conversion on average. The platform tells you.
• Target CPA: what you ask Smart Bidding to pay. It is an instruction, not a result.
• Allowable CPA: the most you can pay for a customer without losing money. Your margins tell you, and no platform knows it.

An illustrative example: a new customer’s first order leaves £50 of contribution margin; over the following twelve months that customer places another 1.5 orders on average, at £30 margin each. Allowable CPA is £50 if you want to recover the cost on the first order, roughly £95 if you think in twelve-month terms. Target CPA should sit below whichever threshold you choose, to leave room for profit: £40 in the first case, up to £70 in the second if cash flow allows and repeat rates are measured rather than hoped for. Working back to these numbers from a revenue goal follows the same logic as our budget guide.

New versus returning customers

A £38 CPA looks comfortably under a £40 target. But if half of those conversions are customers who had already bought from you, the cost per new customer is double. This is where platform ROAS misleads most: brand campaigns and anything that reaches existing customers report very high returns on sales that would largely have happened anyway.

Google Ads tackles this with customer lifecycle goals. The new customer acquisition goal has two main modes: bid higher for new customers (an extra value is added to the first purchase) or only bid for new customers, which limits traffic. Google identifies existing customers through the account’s conversion data, uploaded customer lists, or a parameter in the conversion tag that marks each purchase as new or returning. The measurement documentation also includes a "new vs returning customers" segment and a customer acquisition cost column. According to trade press reports, an option that switches on new-customer reporting alone, without changing bids, has been rolling out since August 2026.

An illustrative example of how the campaign ranking changes:

The bottom row is deliberately crude (it loads all the spend onto new customers alone), yet it turns the ROAS ranking upside down. And even that row tells you what you paid, not what Google actually added: for that you need an incrementality test, as we explain in our guide to experiments.

MER: the number nobody can inflate

Every platform counts the conversions it can tie to its own ads, under its own attribution rules. The upshot is that adding up the revenue claimed by Google, Meta and the rest often gets you close to, or beyond, the shop’s total revenue. The simplest check is MER (marketing efficiency ratio): total business revenue for the period divided by total marketing spend.

If the platforms claim 96% of revenue and you know organic search, email and word of mouth account for at least a third, someone is counting twice. MER will not tell you which channel works, but it cannot be inflated: we read it every month alongside the platform numbers, and when the two diverge that is the cue to dig deeper.

Lead generation: cost per qualified lead and per customer

In lead generation, cost per lead is the most widely used KPI and the most misleading. Optimising for the number of enquiries pushes Google towards whoever fills in forms most readily, who is often not who buys. Two campaigns, same spend (illustrative figures):

The "expensive" campaign brings in customers at half the price. Seeing this takes two things: a CRM that records the status of every lead, and that status flowing back into Google Ads as an offline conversion, so that bidding also learns from quality rather than volume. We cover this in detail in our article on value-based bidding for lead generation. For the business owner, the right KPI is cost per customer acquired, with cost per qualified lead as the leading indicator.

CTR, CPC and Quality Score: useful, but not KPIs

CTR, CPC, Quality Score, impression share, conversion rate: we look at all of them every week, but they explain why a result moved, not whether it was good. A rising CPC is not a problem if cost per customer holds; a falling CTR may simply mean a campaign has expanded into new searches. Google itself states that Quality Score is a diagnostic tool, not a key performance indicator, and should not be optimised or aggregated with the rest of your data.

Our rule of thumb: if a metric is not expressed in margin, in customers or as a ratio to spend, it goes in the appendix.

Conversion lag: why yesterday’s numbers lie

In the standard "Conversions" column, Google Ads records a conversion against the day of the click, not the purchase. If someone clicks on Monday and buys on Thursday, the sale appears on Monday, but only from Thursday onwards. Yesterday’s spend, on the other hand, is already all there. The result: the most recent days always look worse than they will turn out to be.

The figures are illustrative, but the shape of the curve is the one we usually see. To find yours, use the Days to conversion segment on a date range that ended at least 30 days ago. There is also a set of columns that report by conversion date, such as "Conversions (by conv. time)", which is useful for reconciling Google Ads with your back office, since that works by sale date. In practice: do not judge a change on the last few days, and close the monthly report a few days after month end, flagging that the figures may still rise.

How to structure a monthly report people actually read

The report that works with our clients fits on one page, and runs from the business to the platform, not the other way round:

1. Business outcome: total revenue or margin, new customers, MER, against target and against the same month last year.
2. Google Ads contribution: spend, ROAS against break-even ROAS (or POAS, or gross profit), cost per new customer or per customer acquired.
3. What changed and why: three points, not thirty. This is where diagnostic metrics come in, and only if they explain something.
4. Decisions for next month: what we will do, what we need from the business (prices, stock, CRM data), what we are testing.
5. Data notes: extraction date, conversion lag, changes to tracking or value rules.

If you want to know which of these numbers your account can already measure, and which need work on the feed, tracking or CRM, our free audit starts exactly there. You can see how we work in our process, which services include reporting, and the answers to the most common questions in the FAQ.