ROAS 5.55. CPC SEK 4.58. Conversion rate 3.76%. CPA SEK 114.

Ask any Swedish e-commerce business what they benchmark against and you get the same four numbers: ROAS, CPC, conversion rate and CPA. Then ask where their benchmark comes from, and it goes quiet.

There simply is no good Swedish source. You find international numbers that don't apply to Sweden, or agency pages with ranges and no method. PostNord's E-barometer and Svensk Handel's E-commerce Indicator are excellent primary sources for revenue and category growth — but they don't measure advertising efficiency.

So we ran the numbers on our own portfolio.

The answer

Median values for Swedish e-commerce accounts in Google Ads, twelve months (1 August 2025 – 31 July 2026):

• ROAS: 5.55 — the middle half sits between 3.16 and 7.08
• CPC: SEK 4.58 — middle half SEK 2.17–4.63
• Conversion rate: 3.76% — middle half 2.09–5.22%
• CPA: SEK 114 — middle half SEK 65–175

Average order value in the same sample is SEK 452.

The full distribution

Look at the last row. It is calculated on exactly the same data as the median row, but weighted by media spend. ROAS goes from 5.55 to 9.10 and CPC from SEK 4.58 to 2.50 — purely by changing the calculation method.

If you're at ROAS 3 you're in the bottom quartile. At 7 you're in the top one. The spread between the lowest and highest account is a factor of 6.5 — and it is mostly explained by business model and order value, not by how well the account is managed.

Why “average” is the wrong question

This deserves its own heading, because it is the most important insight in the whole analysis. Three entirely legitimate ways to calculate average ROAS on the same data:

• Median per account: 5.55. Each account counts once, regardless of size.
• Spend-weighted aggregate: 9.10. Total conversion value divided by total spend.
• Spend-weighted aggregate excluding the largest account: 3.33. Same calculation, one account removed.

From 3.33 to 9.10 — almost three times the difference, from the same underlying data. The explanation is that a spend-weighted aggregate lets the largest accounts dominate the result. The median doesn't — every account weighs the same.

That leads to two conclusions. For you as a reader: benchmark against the median, never against an aggregate. For the industry: any source that states an average without disclosing the calculation behind it is useless as a benchmark. We have not found a single Swedish source that discloses it.

A warning on the same theme: the medians in the table above are calculated independently of each other, account by account. They don't hold together mathematically. You cannot divide median AOV by median CPA and get median ROAS — the numbers come from different accounts. That applies to every percentile table of this kind, including the international benchmarks you may be comparing against.

Search wins on all four metrics — and takes 11% of the budget

Same sample, split by campaign type:

Search has the highest ROAS, the lowest CPC and the lowest CPA. Performance Max has the lowest ROAS of the three and takes four fifths of the budget.

Don't conclude that you should move budget from PMax to Search. That is exactly the fallacy this table invites. Search largely captures demand that already exists, including brand searches that would have converted anyway — which is why both CPC and CPA are low there. Performance Max works broader and creates part of that demand. The comparison is descriptive, not causal. To know what is actually incremental you need geo experiments or lift studies, not this table.

What the table does say: if you report a single account-level ROAS, it hides a spread of several hundred percent. That is not something you can steer on.

We exclude Display, Demand Gen and Video from the table. The sample is too thin, and CPC is a meaningless metric for video advertising priced on impressions.

Clicks got 17.8% more expensive in a year

This is the clearest trend in the data. Median values per half-year:

Median CPC rose 17.8 percent from the first half of 2025 to the first half of 2026. Conversion rate improved at the same time, from 3.05 to 3.77 percent. ROAS was effectively flat, between 5.01 and 5.57 across all three periods, and median AOV barely moved: SEK 466, 453 and 455.

Remember the warning above — these medians don't hold together mathematically, so don't derive one from the other. But each pattern on its own is clear: clicks get more expensive, traffic converts better, and the net effect on ROAS is roughly zero.

If you expect ROAS to erode dramatically from rising CPCs, it isn't visible here. What is happening is that conversion optimisation and bidding have so far absorbed the price increase. Anyone who doesn't improve conversion rate in step with rising CPC will lose ground.

Q4 does not lift your ROAS

Median ROAS in November–December was 5.77 versus 5.43 in other months. A six percent difference.

Christmas trading lifts volume, not efficiency. If you budget on the assumption that Q4 delivers a better return per krona, recalculate — you get more transactions at roughly the same ROAS. That is still a good argument for scaling up, but it is a different argument.

And now the objection: ROAS is a weak metric

We publish a ROAS benchmark, and we want to be equally clear that ROAS is the worst of the four metrics above to steer on. The reason is that ROAS measures revenue, not profit — and revenue without margin is just turnover.

The classic worked example, taken from Funnel's write-up on POAS: two campaigns cost USD 10,000 each and bring in USD 200,000 each. Identical ROAS of 20. But one product costs USD 100 to produce and the other 190. POAS — profit divided by ad spend — becomes 10 for the first and 1 for the second. Same ROAS, a tenfold difference in profitability.

This is not a theoretical problem. It is what happens every time an algorithm optimises towards conversion value in a range with varying margins: budget drifts towards the products that look best in ROAS terms, which need not be the ones making money. The broader your range and the more your margins differ, the more misleading a single aggregate ROAS becomes.

POAS (profit on ad spend) counts profit instead of revenue: revenue minus cost of goods, shipping and payment fees, divided by ad spend. The formula is trivial. The hard part is the data — you need per-product margin into the ad platform or into a data layer between the systems, and that is where most implementations stall.

Our experience is that a hybrid works best over time:

• POAS or contribution margin as the steering metric for budget allocation between campaigns and product groups. This is where decisions are made, and where margin has to be present.
• ROAS as an operational signal in day-to-day work, within a product group where margins are reasonably uniform. There, ROAS is fast, available and good enough.
• MER or blended ROAS — total revenue divided by total marketing spend — as a sanity check against attribution drift. If channel ROAS looks stellar but blended stands still, attribution is lying.
• Incremental measurement a couple of times a year, through geo experiments or lift studies, to calibrate how much of the attributed value is actually new.

Picking a single metric and steering hard on it is the most common reason we see for accounts that look good in the report and worse in the P&L.

How you should use these numbers

• Benchmark against the median, not an aggregate. An aggregate is dominated by the largest account in any sample.
• Set your own ROAS target from your gross margin, not from an industry average. The average tells you where others are, not where you make money.
• Break it down by campaign type before drawing conclusions. An account-level ROAS hides a spread of several hundred percent.
• Expect rising CPC. Almost 18 percent in a year in our data. Anyone not improving conversion rate at a similar pace loses ground.
• Add a profitability metric. If you can only add one thing to your reporting this quarter, add margin.

FAQ

Method and sample

We think this part is the whole point of publishing the numbers.

The portfolio: Kvantic manages a Google Ads portfolio in the order of SEK 200 million in annual media spend, across several markets and currencies. This analysis is based on the subset of the portfolio that meets three criteria: Swedish market, e-commerce with revenue tracking in place, and reporting in Swedish kronor so that CPC and CPA are directly comparable. B2B, SaaS and service accounts are excluded — one SaaS account in the portfolio reports a ROAS above 90, which says everything about why that filter is needed.

Sample volume: SEK 66.3 million in media spend, SEK 603 million in attributed revenue, 31.3 million clicks, 907,217 conversions, 1,666 million impressions, 2,553 account days.

Period: 1 August 2025 – 31 July 2026 for the headline figures. January 2025 – July 2026 for the half-year comparison.

Data source and definitions: Google Ads API, pulled 17 August 2026. ROAS is metrics.conversions_value / metrics.cost_micros, i.e. on primary conversions. We deliberately do not use all_conversions_value, which includes micro conversions — in one account it produces a value more than 300 times higher. Conversion rate is conversions divided by ad clicks, not session-based conversion rate in Google Analytics. The two numbers are not interchangeable, and the difference is a common source of confusion when comparing against international benchmarks that often measure the latter.

Anonymisation: all figures are aggregated or reported as percentiles. No client names, and no individual client can be identified from the numbers.

What the numbers don't say

• Conversion value is defined by the advertiser. What counts as a conversion and what value it is assigned is set per account. That is a fundamental limitation in any comparison of ROAS and CPA between companies, and it cannot be calculated away. One account in the sample has an unusually low average order value, which likely reflects how conversions are counted there rather than actual order values.
• The sample is not representative of Swedish e-commerce as a whole. These are merchants who hired a performance agency, which reasonably means they are more mature than average. Expect a broad sample to sit lower, not higher.
• The data isn't sufficient for an industry breakdown. We can give a robust median for Swedish e-commerce, but not separate figures for fashion, consumer electronics or pharmacy.
• Attribution is Google Ads' own. Other attribution models give other numbers. This is not incremental ROAS.
• Google has flagged changes to its bidding system from 17 August 2026 that may temporarily affect campaigns using Target ROAS and Target CPA. Our measurement period ends 31 July and is therefore unaffected — but if you compare against your own numbers from autumn 2026, that may explain deviations.

We update these numbers every quarter. If you have your own data and want to compare method — or think we've got the maths wrong — get in touch.

Sources and references

• Google Ads API, account data pulled 2026-08-17
• PostNord E-barometern, in collaboration with HUI Research — primary source for revenue and category growth in Swedish e-commerce, but does not measure advertising efficiency
• Svensk Handel's E-commerce Indicator — monthly revenue indicator, also does not measure advertising efficiency
• Google Ads: About Target ROAS bidding — Google's own documentation, including conversion thresholds per campaign type
• Funnel: POAS — a metric all e-commerce marketers should know — source of the worked example
• Channable: What is POAS? — definition and formula, plus a review of the limitations of ROAS

Read next

• SEM strategies in Stockholm 2026
• The best digital analytics tools 2026
• Our SEM service

Linus Andersson, CEO, Kvantic — a performance marketing agency in Stockholm and Google Premier Partner.