Ecommerce ROAS benchmark: a Google Ads study of 38 accounts and 41 audits

Ecommerce ROAS benchmark and core metrics from 38 accounts

Most ROAS benchmarks you find are either platform averages across every business model or survey answers. This one comes from live Google Ads data: 38 accounts in the Harucon manager account, managed or under audit, each with at least €1,000 in spend and conversions between May and mid-August 2026. Conversions are counted as each ad account reports them.

A quarter of accounts sit below the 25th percentile, a quarter above the 75th. The median is the account in the middle. It describes the market, and it says nothing about whether that account makes money.

Why euros: 30 of the 38 accounts bill in euros. The rest bill in US dollars, Swiss francs or pounds, which we converted to euros at the daily rate. ROAS and conversion rate are ratios and read the same in dollars. Click prices differ from market to market, so treat CPC and CPA as orientation for a US store.

For a US reference point, Triple Whale reports a median Google Ads ROAS of 3.27 across more than 21,000 ecommerce brands from August 2025 to July 2026. That sample is far larger and covers a different period, so the two numbers sit side by side as context, not as a head-to-head comparison.

ROAS by campaign type: Search, Shopping, Performance Max, Demand Gen

We included a campaign type for an account only if it spent at least €500 there, and we report a type only if at least 5 accounts ran it. Search includes brand campaigns.

Shopping and Performance Max tie at a median of 2.30. Performance Max has the higher lower quartile (1.71 vs. 1.29) and drops below 1 in fewer accounts (6% vs. 11%). Demand Gen barely returns what it costs at the median, and 42% of accounts running it saw a ROAS below 1.

Search leads because it includes brand searches, people typing your store name. Those clicks are cheap and convert well, so they lift the Search median. Do not use 4.30 as the bar for prospecting keywords. If you need a refresher on what a ratio like 2.30 does and does not say, read up on ROAS meaning.

Hold the benchmark against your break-even ROAS

A benchmark tells you where other stores land. Your break-even ROAS tells you where you start making money. For a store whose ad account counts orders before refunds and without sales tax, break-even is 1 ÷ ((1 − return rate) × contribution margin).

The same median ROAS makes one store money and costs another. An apparel store with 30% margin and 25% returns needs 4.44 in its account, above the 75th percentile of Shopping and Performance Max in this study. A brand with 50% margin and 10% returns breaks even at 2.22 and makes money below the median.

Compare your campaigns with your own break-even first. Use the benchmark second, to see whether your gap comes from the account or from your margin.

Primary conversion actions: what Smart Bidding optimizes for

Smart Bidding bids toward whatever the account marks as a primary conversion. Several primary actions can be legitimate, for example one purchase action per country or store, but each one needs checking. If add-to-carts or page views count as primary, Google chases them and reports a ROAS your store never sees.

In our analysis of 41 audits, at least 19 of 41 accounts had several primary conversion goals or misconfigured tracking. Examples from individual audits:

• Add to cart set as the primary goal instead of purchase, in 2 accounts.
• 76 active conversion actions, with a page view marked primary.
• More than 130 conversion actions, including at least 6 parallel purchase actions.

Before you compare your ROAS with any benchmark, open your conversion settings and count your primary actions. If more than one purchase action is primary, one order can count twice.

Campaign count, PMax share and budget do not predict ROAS

You may have heard that more campaigns give you more control, that Performance Max lifts efficiency, or that bigger budgets run better. In the 38 accounts, none of the three relates meaningfully to ROAS.

An r of 0 means no relationship, 1 means the two move in lockstep. Anything under 0.2 counts as weak. Restructuring into more campaigns or moving budget into Performance Max will not raise ROAS on its own. The audits point to where the differences sit instead: tracking, brand share and how budget spreads across products, devices and regions.

What 41 Google Ads audits found most often

The second data set is 41 Google Ads audits of ecommerce accounts in Germany, Austria, Switzerland, the US and the UK, with annual budgets from five to seven figures, from recent years through 2026. Counts are floors: a finding only counts if the audit documents it.

In 3 audits with quantified values, 90% to 98% of products never produced a profitable sale through Google Ads. One account put 87% of its budget into products without a single profitable conversion. Further examples from individual audits:

• Channels: the best channel at ROAS 13.7 got 0.8% of budget, the worst at ROAS 1.5 got 51.5%.
• Devices: desktop reached ROAS 13.0 on 31% of budget, mobile ROAS 3.4 on 62%.
• Break-even: one account ran twelve months at ROAS 1.16 against a real break-even of 1.85. No month reached it.

None of the top four findings needs more ad spend to fix. Product-level cuts are the core of what a Google Shopping agency should handle, and device, rule and script gaps are the first checks a Google Ads agency should run.

Brand share: how much revenue comes from your own name

In 8 audits with quantified values, 58% to 94% of measured revenue came through searches for the store's own brand. Those buyers were already looking for you. A strong account ROAS can mostly reflect them while generic campaigns lose money.

Performance Max serves on brand searches unless you exclude them. The PMax figure comes from a separate sample: 13 accounts over the 90 days to September 20, 2026. In 5 of 11 accounts running Performance Max, 25% or more of PMax conversion value came from brand searches.

Split brand and non-brand into separate campaigns and judge them separately. If you run search on Google and Microsoft, a PPC agency should apply the same split on both networks.

Methodology and limits

• Currency: 30 of 38 accounts bill in euros. USD, CHF and GBP were converted roughly at the daily rate. ROAS and conversion rate do not depend on currency. CPC and CPA carry the rounding of that conversion.
• Medians, not averages: one account at ROAS 20 would drag an average up. The median shows the account in the middle, and the 25th and 75th percentiles frame the middle half.
• Medians do not multiply out: each column is its own median, so CPC divided by conversion rate will not give you the median CPA.
• Campaign types: only types with at least €500 spend per account and at least 5 accounts. Search includes brand.
• Audit counts are floors: a finding only counts where the audit documents it. Data was anonymized at the source.
• Limits: all values come from the ad account and depend on each account's tracking. The sample is small, not random and mostly European. It describes stores Harucon manages or has audited, not the whole market.

We publish ad account numbers because only they compare across accounts. We judge our own work in the store backend, and that is also where we measure our guarantee.