Marginal ROAS vs CRO: The Math on the Next $5,500

WRITTEN BY

Gary Culaev, Founder and CEO, Weblics Consulting

Marginal ROAS vs CRO: The Actual Math on Your Next $5,500 a Month

The mistake in most CRO vs ads math is comparing your average ROAS with CRO, when the money you're deciding on is the next dollar, and the next dollar usually earns less. Here are both budget requests, run through the same four questions, with every formula written in one line you can paste into your own spreadsheet.

$1.10 in, $0.25 back.
‍
That's what a dollar of Meta media leaves you at a 3:1 ROAS and a 45% gross margin, once you add ten cents of agency fees and creative for every dollar of media. Three dollars of revenue, $1.35 of gross profit, minus the $1.10 it cost.
‍
Now two requests land on the same desk. Marketing wants 10% more for Meta. The Meta line is $55,000 a month including fees and creative, so that's $5,500, all in. A CRO proposal asks for the same $5,500 a month.
‍
Same money, same month. The rest of this piece puts both requests through the same four questions.

In this guide

Why most CRO vs ad spend math compares the wrong numbers
Question 1: what does the next dollar return?
Meta, Google and CRO at the published medians
Question 2: where does each one break even?
Question 3: what happens when the money stops?
Question 4: how sure can you be?
Both requests in one table
When to turn down both requests
FAQ

Why Most CRO vs Ad Spend Math Compares the Wrong Numbers

The ROAS in your ads dashboard is an average of everything you've spent, and a 10% budget increase buys the next dollars, which usually return less than the average.
‍
Meta says as much in the documentation for Robyn, its own open-source marketing mix model: "each additional unit of advertising investment increases the response at a declining rate." Robyn calls the return on that extra unit the "next dollar response". It's the only return that matters when you're deciding whether to spend more.
‍
Yes, ads are easier to see. Ads Manager shows a ROAS every morning, while a CRO test takes a month to read. But the number on that dashboard describes the money you've already spent. The next $5,500 has its own return, and no dashboard shows it until you've spent it.
‍
So compare the next dollar of ads with the next dollar of CRO. The people who ask what the next dollar returns tend to make better budget calls than the ones who add 10% because the average looks good.

Question 1: What Does the Next Dollar Return?

At a 3:1 marginal ROAS and a 45% margin, the next ad dollar returns about $0.23 in profit. On the model below, the CRO request returns about $1.76 per dollar over its first year and $9.12 over two.
‍
Here's the ad side in one line:
‍
Profit per ad dollar = marginal ROAS × gross margin ÷ (1 + fees) − 1
‍
At 3:1, a 45% margin and 10% fees, that's 3 × 0.45 ÷ 1.10 − 1, or about $0.23 for every dollar, the same $0.25 on every $1.10 you started with. On $5,500 a month, it's about $1,250 of profit.
‍
Now suppose the extra 10% comes back at 2.5:1 instead. The profit on the same $5,500 falls to about $125 a month.
‍
To find your own marginal ROAS, you need two comparable periods with different budgets:
‍
Marginal ROAS = change in ad revenue ÷ change in ad spend
‍
Pick two months with similar seasonality and no big launches, where the Meta budget moved. If spend went up $8,000 and Meta-attributed revenue went up $18,000, the extra money came back at 2.25:1, whatever the dashboard average says.
‍
On the CRO side, the same one-line logic applies:
‍
Profit per CRO dollar = gross margin × added revenue over the period ÷ program cost − 1
‍
To put a number on it, here's the store we'll use for the rest of this piece. It's the same one as in our breakdown of what one A/B test costs: $300,000 a month on 150,000 sessions, a 45% gross margin, a $5,500-a-month program for twelve months, $66,000 in all. The program runs one test a month on the templates every order passes through, the product page and the cart. One test in three wins, which is the share Analytics-Toolkit found across 1,001 real tests, and each winner lifts revenue by 7.5%, the median winning lift in the same data. Winners go live the month after they're read, so the first is live in month four, the second in month seven, the third in month ten and the fourth in month thirteen. Lifts are added, not compounded, and we count first orders only.
‍
By month twelve, the winners have added about $405,000 of revenue. At a 45% margin that's $182,250 of gross profit on $66,000 spent, or about $1.76 for every dollar.

What $5,500 a month returns per dollar: 23 cents from extra ad spend, $1.76 from CRO at twelve months, $9.12 at twenty-four

$0.23 against $1.76 is the headline. The next three questions decide whether you believe it.

Meta, Google and CRO at the Published Medians

Run the same line on what DTC accounts actually report and the ad side gets its most generous reading, because a published median is the average dollar in the account, not the next one.
‍
In Top Growth Marketing's DTC benchmark panel, the median purchase ROAS on Meta is 2.96, across 15 brands and $2.46 million of managed spend. On Google Ads it's 4.28, across 28 brands and $4.36 million. Both are account-level and platform-attributed, which means they count branded search, retargeting and every other cheap click in the account. The next dollar buys none of those, so treat the two numbers below as the ad side's best case rather than its likely one.
‍
Through the same formula, at a 45% margin and 10% fees: Meta's median returns about $0.21 of profit per dollar out, Google's about $0.75. The CRO program on the same store returns $1.76 over twelve months.

Two panels comparing CRO, Google Ads and Meta Ads: monthly spend for the same profit, and net profit per dollar

Ad medians from Top Growth Marketing's DTC panel, checked October 2026. CRO figures are the twelve-month model above.
‍
Turn it around and the gap is easier to feel. To add the same first-year profit the program adds, you would have to put about $12,900 a month into Google Ads at its median, or about $45,900 a month into Meta at its, against $5,500 a month for the program. That's before the next-dollar discount, which applies to both ad columns and to neither CRO column.
‍
None of this says stop buying ads. The two aren't alternatives: a winning test on the page your ads land on raises the return on every ad dollar you were already spending, which is the next section but one.

Question 2: Where Does Each One Break Even?

With a 45% margin and 10% fees, Meta has to return 2.44:1 on the next dollar just to break even on the first order, and CRO breaks even when its winners add enough revenue to cover the program.
‍
Break-even ROAS = (1 + fees) ÷ gross margin

Break-even ROAS on the next ad dollar by gross margin, from 3.14 to 1 at a 35% margin down to 1.69 to 1 at 65%

Margin moves this more than anything else. A store at 35% margin needs more than 3:1 on the next dollar to make any money on the first order. A store at 65% can live with much less.
‍
For CRO, the break-even is a lift:
‍
Break-even lift = program cost ÷ (gross margin × revenue through the tested pages over the months the winners are live)
‍
On the same store, suppose the whole year produced a single winner, live from month four. To pay for the program by month twelve, that one winner has to lift revenue by $66,000 ÷ (0.45 × $300,000 × 9 months), or about 5.4%. If it stays live through month 24, the break-even lift falls to about 2.3%. Either is below the 7.5% a typical winner delivers, which is why a program can pay back on one winner.
‍
Both formulas count first orders only. If your customers reorder, count the second order in both columns, because a better converting page also sells more to people who come back.

Question 3: What Happens When the Money Stops?

Ad profit ends the month the ad budget ends, while a winning test keeps paying on every visitor after the program is over.
‍
Run both requests for twelve months and then stop.
‍
Month 1. The extra ad spend earns its $1,250 right away. The CRO program is still running its first test, so it has earned nothing yet.
‍
Month 12. The ads have earned about $15,000 on $66,000 of spend. The CRO program has three winners live and a fourth about to ship, and has earned about $182,250 of gross profit on the same $66,000.
‍
Month 24. The ad column hasn't moved since month 12. The four winning tests are still live on your pages and earned another $486,000 of gross profit between months 13 and 24, for about $668,250 in all, or $9.12 per dollar. That column assumes the winners hold for the full two years, which is the assumption doing the most work in this piece. Some lifts fade as the page around them changes, so rebuild the winners into your theme and re-check them.
‍
There's one more effect that doesn't show up in either request. A winning test on the pages your ads land on lifts the return on every ad dollar you already spend. A 7.5% lift turns a 3:1 ROAS into roughly 3.2:1 on that traffic, and you don't pay a cent more for it.
‍
Effective ROAS after a lift = ROAS × (1 + lift)
‍
That's the median winning lift, by the way. In Georgi Georgiev's analysis of 1,001 A/B tests at Analytics-Toolkit, the tests that produced a significant winner had a median lift of 7.5%.

Question 4: How Sure Can You Be?

Neither number is certain: ad returns are attributed and the next dollar's return isn't visible in advance, while CRO results come from controlled tests, but most tests don't win.
‍
On the ad side, the figure you see is what the platform attributes to itself. The honest way to find the next dollar's return is to spend it on purpose, for a set period, and measure what changed.
‍
On the CRO side, each test compares a variant against the current page on split traffic, which is as close to a clean measurement as marketing gets. The catch is the win rate. In the Analytics-Toolkit data, 33.5% of tests produced a statistically significant winner, about one in three, and that's among experienced practitioners. A CRO request that doesn't say how many tests it expects to run for each winner is missing its most important line. Our model above assumes exactly that rate, and the table below shows what happens in a year that does worse.
‍
In a budget meeting, the person who asks "what did the last 10% return?" is usually the one who turns out to be right.

Both Requests in One Table

Depending on the marginal ROAS, extra ad spend returns anything from an 18-cent loss to a 64-cent profit per dollar, and that return stops when the spend does. The CRO program ranges from a small loss in a bad year to several dollars per dollar once the winners have had time to run.
‍
Each row assumes $5,500 a month for twelve months, a 45% margin and first orders only. The spend stops at month 12, so for the ad rows the 24-month column matches the 12-month one. The CRO rows use the model from question 1, on a $300,000-a-month store.

Eight scenarios for the same $5,500 a month: ad spend from 2 to 1 up to 4 to 1 and four CRO outcomes, at 12 and 24 months

The 25% row is one winner of the size in our CRO vs ad spend case study, where conversion goes from 2.0% to 2.5%. The median row is what a typical year of testing delivers. The bad-year row is the one to keep in view: a program that finds one winner in twelve tries loses a little money in year one and only comes out ahead because the winner keeps working in year two.
‍
At 2:1, every extra dollar of Meta spend loses you 18 cents. At 4:1, it makes you 64. The whole decision often comes down to which of those your next dollar really is.

When to Turn Down Both Requests

If each order loses money once returns, shipping and discounts are counted, both requests just help you lose it faster, so fix the unit economics before you buy traffic or conversion.
‍
A better conversion rate sells more orders. So does more ad spend. If the order itself is underwater, both requests multiply a loss. Our unit economics checklist covers what to count per order.
‍
Two more cases call for a no on both. If your store is under about 50,000 sessions a month, a four-week test can only confirm lifts of around 18% or more, well above the 7.5% a typical winner delivers, and when to start CRO covers what to do with the budget instead. If your marginal ROAS is also under break-even, the money does more elsewhere. And if you can't fulfill more orders next quarter, neither request has anywhere to go.
‍
Here's the one thing to do before the budget meeting. Find the last month you raised the Meta budget, divide the change in Meta revenue by the change in spend, and compare the result with the break-even table above. If it's under your line, the 10% request has answered itself.
‍
For the CRO column, our CRO ROI calculation guide walks through the formula we use at Weblics, line by line. And since the whole CRO case rests on the win rate, the cheapest way to improve it is to test ideas with better odds from the start. Our product page optimization checklist is the list we start from at Weblics, and it's the same list we'd want to see behind any $5,500 proposal. If you'd rather have both columns run on your own figures, send us your store.
‍
The question isn't whether ads or CRO is better. It's what the next dollar buys.