ROAS means return on ad spend: the revenue Meta attributes to the ads, divided by what you spent. A good ROAS is not a universal number. Purchase ROAS in Ads Manager tells you what Meta is willing to claim for a date range. It does not tell you if the business made money.
I publish account views because the number should be readable in context. A 2.05 website purchase ROAS across $24.97M in spend and $51.14M in tracked revenue is not “worse” than an 8.63 ROAS on $446K. They are different jobs. One is scale. One is efficiency on a smaller base. Cutting the first because the second looks prettier is how operators shrink a brand to protect a ratio.
What ROAS means in Ads Manager
In Ads Manager, purchase ROAS is tracked purchase conversion value divided by amount spent, inside the attribution setting on that column. Change the click window or the view window and the number moves without a single customer behaving differently. If you compare two screenshots, check that the attribution setting, the date range, and the purchase event match.
- Website purchase ROAS and a blended “results” ROAS are not the same column.
- View-through conversions can inflate a retargeting ad that did not cause the sale.
- A catalog campaign and a prospecting campaign can share a ROAS target and have completely different incrementality.
Read the spend beside the ratio. An 8 ROAS on a budget that cannot leave learning is a story. A 2 ROAS that holds while spend climbs into the millions is an operation. Both can be real. Only one of them tells you the account can take more budget.
What is a good ROAS
Before you scale or kill anything, line the Ads Manager number up against three checks the platform will not do for you.
- Contribution margin. A 4 ROAS loses money on a product with heavy discounts, high returns, and expensive shipping. A 2 ROAS can fund growth when the margin and the repeat purchase rate are healthy.
- New versus returning customers. Retargeting and existing-customer campaigns often print a beautiful ROAS because the buyer was already coming back. Acquisition can look worse and still be the campaign that keeps the store alive next quarter.
- MER, or marketing efficiency ratio. Total revenue against total media spend, for the same days, tells you whether the business moved. Platform ROAS can rise while MER falls if you shifted spend onto people who would have bought anyway.
Refunds and cancellations never make it back into the ROAS you screenshotted last Tuesday. If your category returns a lot of orders, read net revenue, not the gross conversion value in the ad account.
Attribution is a setting, not a fact
Ads Manager will hand you a purchase if it can draw a line from an ad to an order inside the window you chose. Seven-day click is not one-day click. A view-through window will claim sales from people who saw an ad and bought later for a reason that had nothing to do with that impression. None of this means the column is useless. It means two screenshots with different windows are not a comparison.
Before I argue with a ROAS number I check three settings. Which purchase event is on the column. Which attribution window is on the column. Whether the Conversions API and the pixel are both sending the same order, which quietly doubles conversion value and makes every buyer look brilliant. If those three are messy, the structure conversation can wait.
Retargeting is where view-through does the most flattering work. Someone was already on the site. An ad reminds them. Meta takes the sale. Some of those sales were going to happen. Some were not. I do not kill prospecting because retargeting printed a prettier ratio. I separate the jobs, then ask how many of the orders were first purchases.
MER is the number the business actually feels
Marketing efficiency ratio is simple: total revenue divided by total media spend, for the same days. It does not care which platform claimed the order. When Meta ROAS rises and MER falls, you often shifted budget onto people who would have bought, or you paid more to harvest demand the email list and the returning customers were already bringing in.
I still read Ads Manager every day. I do not let it be the only room in the building. The store total, the refund rate, and the share of orders from new customers sit next to the platform number. A 4.51 purchase ROAS on $269K of spend, or a 6.14 on $10.2M of tracked revenue, only means something after that lineup. Those are published snapshots, not a promise that the next account will print the same ratio.
Contribution margin is the filter people skip because it is less fun than a screenshot. If the product is discounted, returned often, and expensive to ship, a ROAS that looks fine in the ad account can lose money in the bank. If the product is reordered and the margin is healthy, a lower ROAS can fund growth. I would rather hear the margin on the first call than hear a target ratio with no cost behind it.
Do not judge a test on day two
A new ad set with a handful of purchases is noise. Meta’s delivery is still finding who to show it to, and your sample is too small to survive one expensive day. I would rather hold a test until it has enough spend to be wrong for a real reason than “optimize” it every morning.
The questions I want answered in a weekly read are short. What spent. What came back, in platform numbers and in the store. What was new-customer volume. What creative is tiring. What we are changing next, and what we are leaving alone.
If a buyer cannot walk those questions without hiding behind a single ROAS, they are managing the screenshot. The rest of the account structure only matters after this read is honest. That is the work I describe in Facebook ads campaign structure.
When to scale, and when to cut
Scale when the business can take the next dollar, not when the ratio looks proud. I want the purchase event trusted, the margin known, new-customer volume holding as spend rises, and at least one creative that is not exhausted. If those are true, I raise budget in steps I can read, and I leave the working ad alone while the new budget learns. A sudden triple of spend is how a stable account becomes a story about “the algorithm.”
Cut when the spend is large enough to be wrong and the store agrees. A campaign that cannot find a purchase after real budget, a country whose contribution is negative after returns, or a creative whose frequency is up and whose result is down: those get reduced or stopped. A campaign that had two expensive days inside an otherwise profitable month does not. The mistake I see most often is the reverse pair: scaling the pretty ROAS that is mostly returning customers, and cutting the acquisition campaign because its ratio looks worse.
Hold when you do not yet know. That is a decision, not a delay. Write the spend level at which you will be willing to call the test a loss, and do not reopen the argument every morning. Meta’s delivery moves. Your sample does not become wisdom because you refreshed the screen. If the landing page, the offer, or the tracking changed in the middle of the window, throw the comparison out and start the read again.
The published views on this site are there so the ratio has company. $24.97M spent beside $51.14M tracked, at 2.05 purchase ROAS, is a scale read. 8.63 ROAS on $446K is an efficiency read. Using one as the target for the other is how brands shrink themselves to protect a number. Read the column. Then read the business. Then decide.
Give the number a lag, too. Refunds, chargebacks, and cancelled subscriptions do not walk back into the ROAS you screenshotted on Tuesday. A category that returns a large share of orders should be read on net revenue, a week or two later, not on the gross conversion value Meta posted the morning of the sale. If you scale on the gross number and the returns show up after the budget is already higher, you did not find a winner. You rented one. I would rather wait for that read than congratulate a campaign that the warehouse is quietly undoing.
Put the weekly read on one page the founder can scan. Spend. Platform purchase value, with the attribution window written next to it. Store revenue. New orders versus returning orders. Refunds if you have them yet. The one change you are making, and the one thing you are leaving alone. If that page needs a footnote to explain the ROAS, the ROAS was never the point. The page was. Keep it where you will actually open it.