ROI matters more, because it is the only one of the two that includes the cost of the goods. ROAS tells you what the advertising did. ROI tells you whether the business made money. Use ROAS to steer campaigns week to week and ROI to decide what to sell at all.
The short version
- ROAS is a campaign metric. Revenue divided by ad spend, and it knows nothing about your landed cost or Amazon's fees.
- ROI is a business metric. Profit divided by everything you put in, including inventory, freight, and the agency fee.
- A 5x ROAS can be a loss on a product with thin margin and a high return rate.
- A 2x ROAS can be excellent during a launch, where you are buying rank and reviews rather than this month's profit.
- Score an agency on which one they volunteer first, before you ask.
The mechanism, which is where the confusion starts
ROAS exists because Amazon reports it. It sits in the ads console, it updates daily, and it requires no information the platform does not already hold. That is its entire appeal, and also its limit. Amazon does not know what you paid your factory, what freight cost, what your storage bill looks like in Q4, or what you pay anyone to manage the account. So it cannot tell you whether you made money. It can only tell you the ratio between attributed revenue and ad spend.
ROI needs the numbers Amazon cannot see. Take the selling price, subtract landed cost of goods, referral fee, fulfillment fee, storage, returns, promotions, ad spend, and management cost. What is left is contribution profit. Divide it by what you invested to get it. That is the number a business is run on.
Both are useful. They answer different questions, and treating them as substitutes is how sellers end up scaling a campaign that is quietly funding their own losses.
| Question | Use ROAS | Use ROI |
|---|---|---|
| Which campaign gets tomorrow's budget? | Yes | No |
| Which keyword should be paused? | Yes | No |
| Should this product exist at all? | No | Yes |
| Is the agency fee earning itself back? | No | Yes |
| Are we ready to raise the ad budget? | Partly | Yes |
| Should we enter this category? | No | Yes |
Score your own reporting
Give your current setup a mark out of ten. Weighting matters more than the total, so I have written the weights in.
| Criterion | Weight | Full marks means |
|---|---|---|
| Landed cost per unit is in the model | 3 | Factory price, freight, duty, and inspection, per SKU |
| Returns are netted out | 2 | Return rate applied to revenue and to unit cost |
| Promotions are counted as cost | 2 | Coupons, deals, and discounts reduce the profit line |
| Management fee is included | 1 | The agency invoice appears in the arithmetic |
| ROAS is reported by campaign type | 1 | Launch campaigns separated from harvest campaigns |
| One person can reproduce the number | 1 | From your own exports, in under an hour |
Anything below seven and you are steering on ROAS whether or not you believe you are. Most sellers score their first two points and stop, because the remaining points require touching the supply chain rather than the ads console.
The number that comes before either of them
Neither metric will save a product in a market too small to support it. Before quoting anyone, I want to know the size of the market, and my floor is $2 million a year. Below that there is not enough revenue to capture profitably once cost of customer acquisition is paid. A product in a $400,000 category can post a beautiful ROAS on a hundred units a month and still never pay back its tooling.
That is the buyer side test worth applying to any agency: ask them to size the market before they quote a fee. If a proposal arrives before anyone has looked at category revenue, growth trajectory, return rate, and the rating gap between the incumbents, you are being sold hours rather than an assessment. Our research runs to more than 90 data points before a single number is quoted, and I would hold any competitor to a version of that standard.
What a ROAS report will not tell you
A ROAS report will not tell you that your best performing campaign is your branded search campaign, which largely harvests demand you already earned. Strip branded terms out and look again. Many accounts discover that half of the reported return is a receipt for traffic that was arriving anyway.
It will also not tell you what a return costs. On Amazon a returned unit takes back the revenue, keeps most of the fee, and often cannot be resold at full price. A category with a 20 percent return rate needs a materially higher ROAS to break even than one at 4 percent, and no ads dashboard will mention that.
The last omission is time. ROAS is measured in an attribution window. Rank, reviews, and repeat purchase are measured in quarters. Judging a launch on a 7 day window is judging a marathon on the first mile.
Related answers
- KPIs to track for an Amazon agency engagement
- How does Amazon ads performance pricing work
- KPIs an Amazon agency should report weekly
- What services drive quickest ROI on Amazon
- Hiring an Amazon agency: the complete guide
If you want the profit view of your account rather than the ads view, start with the free audit at Flapen.
