Ecommerce Ad Benchmarks in Italy: ROAS, CPA and Conversion Rates

“We’re at a 2.4 ROAS on Meta with a €38 CPA. Are those good numbers?” Some version of this question lands almost every week. It usually comes from someone running an ecommerce store turning over one or two million euros, looking at the dashboard and suspecting they’re burning money. The honest answer fits in a few lines: it depends on your average order value and on how much margin you keep on it. A 2.4 ROAS is a serious problem if you sell supplements at a 70% margin, and an excellent result if you resell consumer electronics at 12%. The “Italian average” you want to measure yourself against doesn’t exist as official data, however confidently blogs quote it to three decimal places.

This piece shows you how to read your own numbers without a made-up average. First comes the chain that links conversions, ROAS and CPA. Then a table of thresholds by price band, with the formula written out in front of you. Last comes the check almost nobody runs: whether your data has enough volume to mean anything at all. It’s the same method you’d use for digital advertising benchmarks in Italy in general, applied to ecommerce.

The average you’re looking for doesn’t exist, and that’s fine

You can’t find a reliable “average ecommerce ROAS” for Italy because a single figure throws incompatible businesses into the same bucket. Take two fashion stores. One sells €29 T-shirts, the other sells €420 coats. Their audiences differ, and so do their decision times, return rates and margins. Squeeze their ROAS into one average and you get a number that describes neither of them.

Aggregated platform benchmarks have the same problem. One compilation of 2026 data from ecommerce Meta accounts puts so-called blended ROAS between 2.0x and 3.5x, CPC between €0.50 and €1.80, and link CTR between 1.2% and 2.5%. Fashion and beauty sit at the top of those ranges, electronics and home goods at the bottom.¹ Those figures tell you whether you’re on the right pitch. Whether you’re actually making money takes a different calculation, and the same source says so plainly: your benchmark is your break-even ROAS plus a margin buffer, and the platform average is there for something else.¹

There’s a deeper reason too, and experimental research on online advertising backs it up. Businesses don’t all get the same return from the same ad spend. A large field experiment on real businesses shows that what a firm gains depends on the business itself more than on any universal curve.² For you, that means copying another store’s target ROAS is a methodological mistake, even when they sell the same kind of product.

Average ecommerce conversion rate in Italy: what we know and what we don’t

You’ve already seen the conversion rates that do the rounds in Italy: fashion 1.2%, beauty 2.1%, supplements 2.5%, electronics 0.9%. They bounce from one article to the next, and the further they travel the more official they sound. Trace them back, though, and none of them leads to a primary source you can check. No organisation publishes average conversion rates by sector for Italian ecommerce. The official observatories that do exist measure market size, the number of online shoppers and spending by category, which is something else entirely. On-site conversion is known only to whoever can open that site’s analytics.

Knowing this does you a favour, because it stops you comparing yourself to a ghost. The conversion rate that matters is yours, tracked over time and broken down by source. Cold paid traffic converts at one rate. People arriving from a branded search or a newsletter convert at another. Blend them and you get an average that can’t tell you where to step in.

Where platform benchmarks really help is on the intermediate signals of the ad funnel, more than on the final conversion. Link CTR tells you whether your creative and your audience are connecting. On Meta, anything under 0.8% usually points to a creative or targeting problem you should diagnose before raising the budget.¹ That signal is useful because it sits right next to the lever you can pull. On-site conversion depends on price, product page, shipping costs and trust, none of which a sector average can capture.

ROAS, CPA and conversions are the same equation

Almost no Italian page writes this part down, and it can spare you months of confusion. Conversions, ROAS and CPA are one thing seen from three angles. Keeping them in three separate dashboard tiles hides the connection.

Start with ROAS. Revenue is the number of conversions times the average order value. Spend is the number of clicks times the cost per click. The number of conversions is the number of clicks times the conversion rate. Cancel out the clicks above and below the line and you’re left with the core relationship.

ROAS = average order value (AOV) × conversion rate (CR) ÷ cost per click (CPC)

This leads to a practical point few people connect. Lift your conversion rate by 20% without touching CPC and your ROAS rises by exactly 20%. Work on the product page or the checkout moves the same bottom-line number as work on the campaigns.

CPA is the same equation turned around. Cost per acquisition is spend divided by conversions, which comes out as CPC divided by conversion rate. And since ROAS is AOV divided by CPA, the same relationship gives you CPA directly.

CPA = average order value ÷ ROAS

With a €90 AOV and a ROAS of 3, your CPA is €30. End of story. If someone promises you a lower CPA “without changing anything”, what they’re really saying is that they’ll raise ROAS or lower your order value, because there’s no third route. Keep this chain in your head and you’ll spot inconsistent targets in seconds, like “I want a ROAS of 4 and a €15 CPA” when the average order is €90.

Realistic ROAS by price band

Everyone segments by industry out of habit. Segmenting by price band tells you more, because average order value alone decides how much you can spend to win a customer. The table below is built on paper. I calculate the ROAS and CPA thresholds with the formula above and state the assumptions. The conversion rates are illustrative ranges, there to give you an order of magnitude, and for the reasons already covered they don’t come from a primary source.

Break-even ROAS is simply 1 divided by gross margin. At a 40% margin you break even at a ROAS of 2.5: every euro of revenue leaves you 40 cents, and you spend one euro on ads for every 2.5 in sales. The maximum CPA on a first order is your margin in euros per order, which is order value times margin.

Average order value	Expected on-site CR	Assumed gross margin	Break-even ROAS	Max CPA (first order)
under €50 (e.g. €35)	2.0–3.5%	45%	2.2	~€16
€50–150 (e.g. €90)	1.2–2.2%	40%	2.5	~€36
€150–500 (e.g. €280)	0.8–1.5%	35%	2.9	~€98
over €500 (e.g. €750)	0.4–0.9%	30%	3.3	~€225

Find the row closest to your business and compare your actual ROAS with the break-even figure, not with an average. If your average order is €90 at a 40% margin and you’re sitting at 2.4, you’re losing a few cents per order before you count returning customers. That’s manageable, as long as the margin on the second purchase bails you out. A 2.4 ROAS on a €35 average order, on the other hand, puts you in healthy territory, because your break-even is 2.2.

The max CPA column is the one that changes how people think. At a €750 average order you can pay €225 to win a customer and still break even on the first purchase. A store selling at €35 has €16 of margin, so it can’t chase that CPA even if it wanted to. That’s why an ecommerce CPA benchmark in the abstract means nothing until you tie it to order value.

From ROAS to CPA: the same threshold seen from the cost side

The max CPA in the table is the one that breaks you even on the first order. Plenty of stores make their money on the second order, the third, or the subscription. If a customer orders three times a year on average, the margin you can invest in winning them is their twelve-month margin, which is far bigger than a single order’s. Separating first-purchase CPA from the CPA you can sustain on customer value is what separates a store that scales from one that hits a budget ceiling and stays there.

Breaking CPA into its parts shows you where the cost is hiding. A high CPA can come from an expensive CPM, a weak CTR or a low on-site conversion rate, and each one needs a different fix.¹ Before you decide “ads don’t work”, check which of the three has drifted from your own history.

This is the first place where AI saves you real time, provided you already have your numbers to hand. The right process is to work out break-even and max CPA from your actual margins. AI’s only job is to do it fast and without arithmetic slips. Open ChatGPT or Claude and paste in a prompt like this one, swapping my figures for yours.

Act as an ecommerce analyst. Here is my data:
- average order value (AOV): €90
- gross product margin: 40%
- shipping costs I absorb per order: €4
- average return rate: 8%
- average orders per customer over 12 months: 2.3

Calculate, showing your working:
1. my real net margin per order after shipping and returns
2. break-even ROAS on the first order
3. max CPA on the first order
4. max sustainable CPA based on 12-month customer value
Then tell me how much I can raise my target CPA if I accept
breaking even on the second order instead of the first.

The value comes from including shipping and returns in the inputs. The prompt itself is the easy bit. Anyone who calculates break-even on clean gross margin is kidding themselves about how much room they really have.

How many orders before you can trust your ROAS?

This check changes more decisions than any other, and Italian guides almost never mention it. Below a certain number of conversions a month, that month’s ROAS is statistical noise rather than signal. Comparing it with a benchmark is like judging a die after three rolls.

Experimental research on online advertising makes the point well. Display ad effects are small and need large samples to measure reliably, so much so that statistical power and sample size become the central problem of any serious test.³ If you go from a ROAS of 2.1 to 3.4 on 25 orders a month, most of the time you’ve simply drawn a lucky month. Nothing has actually improved.

The rule of thumb I give small ecommerce businesses: below 30–40 paid orders a month per campaign, don’t optimise on monthly ROAS and don’t compare it with any benchmark. Look at longer windows, such as the quarter, or merge campaigns to build volume. The more orders you collect, the more your average settles and the more it’s worth reading.

Here AI helps with the second piece of the process: building your own internal benchmark instead of hunting for an external one. Export weekly orders and spend for the last six months from GA4 or Google Ads, drop them into Google Sheets, and use a prompt like this.

I'm giving you a table with, for each week: paid orders,
ad spend, ad revenue, conversion rate.
1. Calculate the mean and standard deviation of my weekly ROAS.
2. Tell me how many weeks I need, at my average order volume,
   before a change in ROAS is probably real rather than noise.
3. Flag any unusual weeks and suggest whether they come from
   order volume that's too low or from a genuine shift in performance.
Don't make up data: use only the table I give you.

The last line of the prompt matters as much as the rest. Without that constraint the model tends to fill gaps with plausible-looking numbers, which is the last thing you need when you’re trying to work out which numbers are real.

Performance Max and the ROAS that isn’t all yours

People who write about Performance Max usually sell Performance Max setups, so the awkward part tends to get left out. Some of the ROAS you see in the dashboard comes from people who would have bought anyway: searches for your brand name, returning customers, traffic the campaign took credit for without creating it. The same applies when you compare it with standard Shopping, and to the cannibalisation of branded traffic.

The technical knot has a name, incrementality, and researchers have treated it for years as the real problem in ad measurement. Optimising for true incremental lift over attributed conversions is described as the right frontier, because platform attribution tends to claim credit it hasn’t earned.³ Your attribution model changes the number you read. Last-touch, the most common, hands the whole order to the final click even when earlier touchpoints did the work.⁴

Meta has a twin version of the problem. Prospecting ROAS plus retargeting ROAS almost never adds up to your real revenue, thanks to double counting across platforms and generously counted view-through conversions. The clean way to look at it is MER: total revenue divided by total ad spend, checked against what actually shows up in your accounts.

You can run a practical test tomorrow without expensive tools. For two weeks, exclude the campaigns that capture your brand name, or slash their budget, and watch total revenue rather than the ROAS of the individual campaign. If real sales don’t drop as much as those campaigns’ ROAS would suggest, you’ve just measured how much of your ROAS was traffic you already had. The same attribution logic applies when you compare channels against each other, as with Google Ads benchmarks in Italy or the reference average ROAS data across the cluster.

Raise your average order value and every threshold moves

Everyone lists average order value as a metric to watch. Few treat it as what it really is: the lever that shifts your sustainable ROAS threshold without touching a single campaign. Look at the formula again. The max CPA you can afford is order value times margin. Raise your average order by 20% and your sustainable CPA rises by 20%, while break-even ROAS stays where it was. You’ve just bought room to manoeuvre in the auctions without changing your offer or your creative.

You already know the levers. The numbers just make them clearer. A free-shipping threshold set slightly above your average order nudges customers to add one more item rather than pay for delivery. If your AOV is €42, a threshold at €59 pulls a share of baskets up towards that figure. Bundles and larger sizes or formats work the same way, lifting the value of each order rather than chasing more orders.

There’s also experimental evidence that these thresholds genuinely change buying behaviour, beyond just shifting the average on paper. The tricky part is setting the threshold against the real distribution of your baskets rather than a round number picked on gut feel. Set it too far above a typical order and you put people off instead of nudging them. So before you touch the campaigns, work on order value and recalculate the table’s thresholds with the new figures. Quite often the store that “can’t make ads pay” has an order value too low for the margin it keeps, and targeting has little to do with it.

The same reasoning carries over to models beyond the classic online shop. A DTC brand with repeat purchases thinks in terms of customer value. A local service selling online alongside a physical location also has to count orders that start with an ad and close in the shop. The method doesn’t change. Only what goes into order value and margin does.

Further reading

• B2B ads benchmarks in Italy: LinkedIn, Google and the channels that generate leads

Sources

Footnotes

1. Meta Ad Benchmarks for Ecommerce in 2026: CTR, CPC, CPM, CPA, and ROAS by Sub-Vertical. adlibrary.com ↩ ↩² ↩³ ↩⁴

2. Frontiers: Which Firms Gain from Digital Advertising? Evidence from a Field Experiment. dash.harvard.edu ↩

3. Inferno: A guide to field experiments in online display advertising. onlinelibrary.wiley.com ↩ ↩²

4. Return on investment and return on ad spend at the action level of AIDA using last touch attribution method on digital advertising platforms. doi.org ↩