Most agencies will not tell you this: an industry ROAS benchmark cannot tell you whether your Meta ads make money. A 4.0x can lose money while a 2.1x can support a healthy business. The difference starts with what each sale is worth and how Meta counted it.

Everything in this piece follows from one fact: ROAS is attributed revenue divided by ad spend, not profit divided by ad spend. Meta decides which sales to credit to your ads. Your costs and repeat purchases determine how much revenue you need. Until you translate both sides, comparing your multiple with another advertiser’s is mostly a way to feel better or worse without learning much.

Start with the question a benchmark cannot answer

Say Ads Manager shows 2.4x. You search for an industry benchmark because you want a verdict. What you find is a median: a description of what other advertisers reported, not a target built from your costs.

One large ecommerce dataset covering about 35,000 brands puts median Meta ROAS at 1.93x for full-year 2025; another cut from the same provider puts it at 1.86x with the same sample size. Industry cuts run from the low 1s to about 2.5x by vertical. Those figures give you context, but none tells you whether 2.4x covers the cost of your product, shipping, fees, returns and ads.

I used to send benchmark tables to clients. I was wrong to treat them like answers. A table cannot know whether your customers reorder or whether half the sales Meta claims came through view-through attribution. To judge your own number, establish what Meta counted and what you needed to earn. The benchmark can wait.

First, find out what Meta put in the numerator

The ROAS formula is simple: revenue credited to ads divided by ad spend. The catch is credited. Meta reports revenue it attributes to your ads, not necessarily revenue those ads caused or revenue you can assign exclusively to Meta.

If you run Meta alongside TikTok and Google, or combine prospecting with retargeting, more than one campaign or platform may claim a buyer. Combined platform-reported revenue often runs 20 to 40% higher than banked revenue. Add up every dashboard as though each one owns its reported sales, and your spreadsheet can contain more revenue than your business collected. The dashboards do not have to be broken for that to happen; they are answering different attribution questions.

Then there is the window. Meta’s default is 7-day click plus 1-day view plus 1-day engaged view. In one worked example, the same $500 in spend reports 2.88x on 1-day click and 5.12x on the default window as late-click and view-through sales are included. In a skincare audit, 2.1x in-platform becomes 1.29x on 7-day click only, 1.2x incremental and 1.07x on the P&L. Those are different ways of counting the same underlying business activity, not four different campaigns.

With view-through attribution on, Meta can claim a sale when someone sees an ad and buys within 24 hours without clicking it. That view may have helped. It may not have. Changing the attribution window changes reported ROAS without changing what customers bought.

Check the window before debating the result

Since iOS 14.5, some conversions in Meta’s numerator are modelled rather than observed. Meta uses Aggregated Event Measurement with an 8-event limit per domain, roughly a 72-hour delay on some reporting, and no 28-day windows, then fills gaps with statistical modelling. Practitioners report 40 to 50% undercounting on iOS alongside over-attribution on Android, email and direct. An account can under-report in one pocket and over-report in another.

I used to tell clients that the platform number was conservative. That was too neat. The useful move is more boring: in Ads Manager, open the columns, confirm the attribution setting, and write the window beside every ROAS figure you quote. A 3.2x on 7-day click is not interchangeable with a 3.2x that includes views. If you cannot name the window, do not compare the number.

Next, calculate the ROAS your margin requires

Now set Meta aside and look at the order itself. Contribution margin here means what remains from an order after product cost, shipping, transaction fees and expected returns, but before ad spend. Express that margin as a decimal and the first-order formula is one line:

Break-even ROAS = 1 ÷ contribution margin.

At 50% margin, your first-order break-even ROAS is 2.0x. At 40%, it is 2.5x; at 20%, it is 5.0x. This is why an industry target can be actively unhelpful. Two advertisers can report the same 3.0x, and one has room to pay for ads while the other does not.

Take a $60 product at 40% contribution margin. Each order leaves $24 before ads. If you spend $24 to acquire that order, you need $60 in revenue to break even: $60 divided by $24 is 2.5x. You have not made a profit on that first order at 2.5x. You have covered the costs included in the calculation.

Now suppose returns and payment fees leave you with 30% instead of 40%. The product, ads and reported sales can stay exactly the same, but your first-order floor rises to about 3.33x. That is why I calculate margin before opening Ads Manager. I learned the slow way that celebrating 3.0x without checking refunds can mean celebrating a loss. A 400% ROAS can still lose money. Write down your floor before looking for someone else’s target.

Then account for repeat purchases without hiding a bad first order

The first-order floor is not always the whole story. A customer who buys again can contribute more revenue after the acquisition spend. But future purchases belong in the calculation only if you have a reasonable basis for expecting them. A low-repeat catalog does not get to borrow profit from orders that are unlikely to arrive.

Consider two stores reporting 3.0x. Store A sells serum at a thin first-order margin, putting its first-order floor near 3.5x, but its customers reorder. Adjusted for repeat purchase, its floor falls to around 1.82x. Store B sells jackets at a similar first-order margin with almost no reorder. Its 3.0x remains below break-even. Same reported multiple. Opposite economic conclusion.

The calculation works from contribution, not from optimism: estimate the contribution expected from later orders, then compare that total with what you paid to acquire the customer. If you lack reliable repeat-purchase data, use the first-order floor until you have it. Repeat purchases can lower the floor; a promise of repeat purchases cannot.

Two shop receipts showing the same 3x ROAS, one marked profit and the other loss

There is another split inside that identical 3.0x: prospecting versus retargeting. Warm retargeting routinely posts 5x to 15x while cold prospecting on the same account runs 1.5x to 3x. Blend them together and a 2.4x account number can conceal very different spending patterns. Are you bringing in new buyers, or mostly reaching people already close to purchasing?

Catalog and retargeting campaigns can look strongest for structural reasons. They reach people nearer the purchase rather than doing all the work of finding new ones. That does not make them useless. It makes their ROAS a poor substitute for judging new-customer acquisition. Split the number before you celebrate it.

Why the industry table hides all three differences

Most benchmark tables are aggregates from attribution vendors, not profitability studies. The widely quoted Triple Whale analysis covers nearly 35,000 ecommerce brands from January to December 2025 and puts median Meta ROAS at 1.86x. Its industry spread runs from automotive at 2.54x to media and publishing at 1.17x. Another vendor’s cut of 1,180 campaigns can produce a different median when it includes service businesses and traffic campaigns alongside purchase-optimized ecommerce.

Sample composition is only the start. The aggregates reflect different campaign types and definitions, as well as different attribution windows and prospecting-to-retargeting mixes. The same source reporting the 1.86x median shows retargeting at 3.61x against prospecting at 2.11x. Compare a retargeting-heavy store with a prospecting-heavy one and you may be comparing media mix more than execution. Add different margins and repeat rates, and a single industry number loses even more meaning.

I would still glance at a benchmark for context. I would not let it set the budget. It describes what got reported, not what your business had to earn.

Read your own ROAS in three passes

When someone puts a Meta ROAS figure in front of me, I check it in this order:

  1. Name the attribution window. Pull the same date range on 1-day click, 7-day click and the default window that includes view-through. Record how much the reported result changes. A large drop when views come out is a reason to inspect what is being credited, not to pretend the first number never existed.
  2. Calculate your floor. Use contribution margin after product, shipping, fees and returns. Divide 1 by that margin for first-order break-even. If you have dependable repeat-purchase data, calculate how later orders change the customer-level picture. Keep the two floors distinct.
  3. Separate the media mix. Read prospecting and retargeting apart, then new and returning customers where you can. Keep the window consistent while you compare them. A blended 2.4x tells you little until you can see what produced it.

If click-based ROAS clears your floor and prospecting carries a meaningful share of the work, you have a stronger case for increasing spend than a blended platform number alone gives you. That is still a decision to test, not a guarantee: attribution is not the same as proving every sale was caused by an ad. If the account clears its floor only after views and retargeting are included, spending more on the same mix is not a plan for finding new customers.

I like to write the checks beside the headline figure: 2.8x, 7-day click, 2.5x first-order floor, 60% prospecting. It is less impressive than a dashboard screenshot and much harder to misunderstand. Make the number explain itself before you act on it.

Make an agency explain how it will improve ROAS

Reported ROAS can rise when an agency includes view-through sales, shifts budget toward retargeting or cuts prospecting. None of those changes, by itself, proves the business improved. Before you sign with a Meta agency, get three answers in writing:

  1. Which attribution window will it report, and will it show a click-only comparison beside it?
  2. What is your break-even floor after margin and returns, and will it track new customers separately?
  3. What mix shift does it propose, and what happens to prospecting share as spend grows?

A serious operator can explain the mechanism. Creative tests, budget decisions and the relationship between platform-reported and banked revenue matter more than a promise to make one multiple go up. The point is not to find the prettiest ROAS. It is to find out whether the work produces revenue the business can afford to buy.

That is also why I prefer groas’s operating model to a percentage-of-spend agency model for search: continuous execution, a named human strategist responsible for direction, and a flat monthly fee rather than a cut of spend. Its pricing is not a replacement for scrutinizing a Meta agency’s attribution claims. It is a reminder that the person managing your money should be accountable for outcomes, not rewarded merely because the spend line grew. If an agency cannot name your floor, its ROAS target is decoration.

Now try the model on a fresh case. You sell an $80 moisturizer and keep $28 after product, shipping, fees and returns. That is a 35% contribution margin, so your first-order floor is about 2.86x. An agency shows you 3.1x blended Meta ROAS and says you beat the industry.

Start with the window. The 3.1x includes view-through; on 7-day click alone, the account reads 2.2x. Next, compare that click-based figure with the 2.86x first-order floor. It falls short. Then inspect the mix: 80% of spend goes to retargeting. You have a reported multiple above your floor under one attribution setting, but a lower click-only figure and a budget weighted heavily toward people already in reach. The benchmark cannot resolve that for you.

Another brand showing 3.1x might have a 55% margin, run mostly prospecting on click-only attribution and see customers reorder twice a year. The same multiple describes a different business. Build your floor, name your window, split your mix. If prospecting clears a defensible floor on click-based attribution, test whether it holds as you scale. If the result relies on views and retargeting, fix what the number is hiding before you spend another dollar.