What’s a Good ROAS? Benchmarks & What They Really Mean (2026)

The rough benchmarks (and why they’re only a start)

You’ll see figures like: a 3:1 ROAS as a common baseline, 4:1 as ‘good’ for many brands, and 5x+ for lean, high-margin businesses. E-commerce prospecting often runs lower (2–3x) while retargeting and branded search run much higher. These are useful reference points, but they ignore the one thing that decides whether the number is actually good: your margin.

The number that matters: break-even ROAS

Your break-even ROAS is simply 1 ÷ gross margin. At a 50% margin, break-even is 2x, anything above 2x profits. At a 25% margin, break-even is 4x, so a ‘good-looking’ 4x ROAS is exactly break-even, making zero profit. This is why two brands with the same 4x ROAS can have completely different outcomes. Run your numbers in our ROAS calculator to see your break-even instantly.

Why ROAS alone can mislead you

ROAS ignores product cost, shipping, payment fees, returns and the cost of acquiring vs retaining. A campaign with a lower ROAS but higher-LTV customers can be more profitable than a high-ROAS campaign full of one-time discount hunters. That’s why serious brands manage to contribution margin (CM2) and LTV, not ROAS alone, more in Beyond ROAS.

How to actually improve ROAS

Better creative and testing volume, tighter audience and search-term targeting, landing-page CRO, and cutting wasted spend usually move ROAS more than bid tweaks. But always improve it against profit, not the vanity number. That’s how we run paid campaigns — measured to margin.

Break-even ROAS worked all the way through

Most people calculating break-even ROAS stop one step too early. They take the gross margin, invert it, and call it done. That answer is usually optimistic. A real order carries costs that never appear in the product margin line, and the gap between those two numbers is where a lot of D2C brands quietly lose money while their dashboards look healthy.

Take an illustration. A skincare brand sells a kit at ₹1,500. Cost of goods is ₹500, so the gross margin is ₹1,000, which suggests a break-even ROAS of 1.5. Now add what the order actually costs: ₹90 shipping, ₹40 packaging, ₹35 payment gateway charges, ₹60 in warehouse and pick-pack labour, and a returns provision. That is ₹225 before returns. Contribution per order drops to ₹775, and break-even ROAS moves to 1,500 divided by 775, which is roughly 1.94.

Then the returns provision. If one order in ten comes back and you recover nothing but the product, the effective contribution across ten orders is nine times ₹775 minus the shipping and handling burned on the tenth, so call it ₹6,750 across ten orders, or ₹675 each. Break-even ROAS is now about 2.22. The number moved from 1.5 to 2.22 without a single assumption anyone would call unreasonable, and every brand that has ever told you it is profitable at a 2 has almost certainly not run this arithmetic.

Two refinements make the number sharper. First, if a decent share of your buyers order again, a first purchase can be allowed to run below break-even because the second and third orders carry no acquisition cost at all, and brands with genuine repeat behaviour routinely accept a first-order ROAS that would terrify a single-purchase business. Prove the repeat rate before you use it as an excuse. Second, discount codes. A code that fires on ad traffic reduces the revenue in the numerator while the platform happily reports the pre-discount figure, so your realised ROAS is lower than reported by roughly the discount rate on every order that used one.

Do it for your own catalogue. One spreadsheet, one row per SKU, all six cost lines. The number you get is the only ROAS target that means anything, and it will be different for every product you sell, which is itself the most useful thing this exercise tells you.

Blended ROAS, platform ROAS, and which one pays salaries

Two numbers claim to be your ROAS and they rarely agree. Understanding why is most of the job.

Platform ROAS is what Meta or Google reports inside their own interface, calculated from conversions each platform attributes to itself using its own window and its own rules. Both platforms are counting the same sale when a customer saw an Instagram ad on Tuesday and clicked a search ad on Thursday, which means your reported revenue across accounts can comfortably exceed the money that actually arrived in your bank. That is not fraud. It is two systems each answering the question ‘did my ad contribute’ honestly and separately, and then somebody adding the answers together, which nobody should ever do.

Blended ROAS is brutally simple. Total revenue for the month, divided by total advertising spend for the month, across every platform. No attribution, no windows, no arguing. It undercounts nothing and overcounts nothing, and it is the number your accountant would recognise. Its weakness is equally simple: it cannot tell you which campaign to turn off, because it treats organic sales, repeat customers and word of mouth as though the ads produced them.

A quick way to see the size of your own gap. Take three months, add up every rupee of platform-reported revenue across all accounts, then put your actual order revenue for the same period beside it. The ratio between those two is your inflation factor, and once you know it you can read platform numbers sensibly rather than pretending they are cash. Recalculate it quarterly. It moves whenever your channel mix does.

Use both, for different decisions. Blended tells you whether the whole operation is working, so set your target there and review it monthly against your break-even figure. Platform numbers tell you which creative and which campaign to feed or kill, so use them relatively rather than absolutely, comparing one ad set against another inside the same account rather than trusting the headline multiple. A brand hitting 4 on platform reports and 1.8 blended does not have a 4. It has a 1.8 and an attribution problem, and the correct response is to fix the target rather than the spreadsheet.

The next rupee is the only rupee that matters

Average ROAS is a report. Marginal ROAS is a decision. Almost every scaling mistake in performance marketing comes from confusing the two, and once you see the difference you cannot unsee it.

Here is why. Your account is currently at ₹5 lakh a month returning ₹15 lakh, so a reported 3. That average includes your branded search terms, your retargeting pool and your warmest audiences, all of which would have converted at some rate regardless, and it is dragging the headline number upward while telling you nothing about what happens if you spend more. The question that matters is what the six-hundred-thousandth rupee returns, not what the average of the first five lakh returned.

Test it directly. Raise spend by a defined increment, say twenty per cent, hold everything else steady, and wait a full purchase cycle. If revenue rose from ₹15 lakh to ₹17 lakh on an extra lakh of spend, the marginal return on that lakh was 2, even though your average still reads 2.83. Compare that 2 against your true break-even from the first section. If break-even is 2.22, you just bought volume at a loss while every dashboard congratulated you.

Branded search is the clearest example of the distortion. People typing your brand name were already looking for you, so those clicks convert beautifully and report a high return, which lifts your account average while adding almost no customers you would not have got anyway. Nobody is suggesting you switch it off. Just report it separately, because leaving it inside the blended account number makes every scaling decision you take from that number slightly wrong in the same direction.

Marginal returns fall as you scale. That is the auction working normally, not a sign that something broke, because the cheapest and most interested people were always going to be reached first and everyone after them costs more to persuade. Your job is finding the point where marginal ROAS meets break-even and holding spend near it. Above it, scale. Below it, stop. The average will look worse as you approach that line, and the business will make more money, which is a conversation worth having with your founder before you start rather than after.

COD, returns and the Indian arithmetic nobody puts in the dashboard

Everything above assumes the money arrives. In a lot of Indian D2C, a meaningful share of it does not, and that gap does not appear anywhere in a ROAS calculation until you force it in.

Cash on delivery is the main culprit. A COD order counts as revenue in your ads manager the moment it is placed, but it converts to actual cash only when somebody opens their door and pays, and return-to-origin means you eat forward shipping, reverse shipping and handling on every one that fails. Prepaid orders carry a gateway fee and settle almost fully. So a catalogue running heavily on COD needs a materially higher target ROAS than the same catalogue running prepaid, and averaging the two into one blended target hides which half of your business is subsidising the other.

The fix is boring and it works. Track your COD share and your RTO rate by campaign, by creative and by pin code, because the variation across those three cuts is larger than most operators expect and the worst-performing slices are usually identifiable within a month. Then act on it: prepaid discounts, partial-advance collection, order confirmation over WhatsApp before dispatch, and simply excluding the pin codes where the numbers never work. Each of those raises your realised ROAS without touching the ad account at all.

Put a number on it. Hypothetically, so the scale is visible. Say 60 per cent of your orders are COD and one in five of those never delivers. On 1,000 placed orders you have 600 COD, of which 120 come back, so 880 orders actually earn anything while your ads manager reported 1,000. Reported ROAS of 3 is a delivered ROAS of about 2.64 before you count the shipping burned on the failures. Run that on your own COD share tonight.

Returns are the second layer. Apparel behaves nothing like supplements. GST treatment, discount codes stacking on top of ad-driven traffic, and the marketplace orders you cannot attribute all move the real number further from the reported one. Build one monthly reconciliation sheet: ad spend, reported revenue, delivered revenue, contribution after all costs. Four lines. It will be the most useful document in your business.

Mistakes to stop making, and a 30-day repair list

A short list follows. These are the things that reliably wreck this metric.

Optimising for a purchase event that fires on the thank-you page while counting COD orders that never delivered. Comparing this month’s ROAS against a last month that quietly contained a sale. Judging a campaign after four days because the number looked bad, which mostly measures your own patience rather than the campaign. Setting one ROAS target across a catalogue where margins differ by a factor of three. Adding Meta and Google revenue together. Cutting spend the moment ROAS dips, which raises the average by removing the incremental volume that was actually paying the rent. And measuring the whole thing on a seven-day window when your considered purchase takes a month to close.

Now the repair. Week one, build the contribution sheet by SKU and calculate a genuine break-even ROAS for each. Week two, set up blended tracking, one number, revenue over spend, reviewed weekly, and stop quoting platform ROAS in leadership meetings. Week three, run a single marginal test: one campaign, twenty per cent more budget, everything else frozen, and write down what you expect before you start so you cannot reinterpret it afterwards. Week four, cut the pin codes and creatives where RTO is destroying contribution, and rebuild your target around delivered revenue rather than placed orders.

One cultural fix belongs on the list too. Stop treating a ROAS target as a performance grade for the media buyer, because the moment it becomes one, the account gets quietly optimised towards the number rather than towards profit, and the easiest way to hit a target is to shrink spend until only the warmest audiences remain. Judge the team on contribution rupees. The multiple is a diagnostic, not a scoreboard.

Then set the cadence. Monthly for blended and break-even, weekly for campaign decisions, quarterly for the cost inputs since shipping rates and gateway charges change without asking you. The brands that get this right are rarely the ones with better media buyers. They are the ones who know what a sale actually earns them.