Validation
How to estimate Amazon product profitability accurately
Most profitability models are optimistic in four specific, predictable places. Fix those four and your spreadsheet starts telling you the truth — which is occasionally unwelcome and always cheaper than finding out later.
Updated September 11, 2026 · Written by the SellRadar team
The four optimistic assumptions
Almost every model that turns out to be wrong is wrong in the same four places. Before anything else, check whether yours contains them.
- Using an aspirational price. You modelled at what you hope to charge, or at the top listing's price. Use the median price of the page — that is where the market has settled, and it is where you will end up.
- Using the factory quote as your cost. Your cost is landed: unit price plus freight, duties, customs handling and inspection. The quote is typically well under the real number.
- Assuming organic traffic. A new listing with no reviews and no rank is invisible. Advertising is not a growth lever in month one, it is the cost of being seen at all, and it is frequently the largest line after goods.
- Ignoring returns. In categories where fit, size or expectation mismatch is common, returns are a structural cost, not an exception.
The full cost stack
Work down this list. Every line is a real cost that people leave out.
| Line | What it is | Commonly forgotten? |
|---|---|---|
| Unit cost | Factory price at your actual order quantity | Quoted at their quantity, not yours |
| Freight | Sea or air, per unit | Often |
| Duties and customs | Varies by product classification and destination | Very often |
| Inspection | Third-party QC before shipment | Almost always |
| Amazon referral fee | Percentage of sale price, category-dependent | Rarely |
| FBA fulfilment fee | Based on size tier and weight | Rarely |
| Storage | Monthly, higher in Q4, worse for slow movers | Often |
| Long-term storage | Charged on aged inventory | Almost always |
| PPC | Cost per acquired sale, highest at launch | Underestimated rather than omitted |
| Returns and disposals | Refunds, unsellable units, removal fees | Often |
| Packaging and inserts | Per unit | Often |
| Photography and listing | One-off, amortised over the first order | Usually |
| Samples | Sunk before you sell anything | Usually |
Amazon's own FBA revenue calculator is authoritative for the referral and fulfilment lines because it uses Amazon's actual fee schedule rather than an approximation. It is free and public. Use it rather than any third-party estimate for those two lines specifically.
Building the model
- Start from the page's median price. Open the top 20 organic listings, take the median. Not the mean, which one premium outlier distorts.
- Run that price through Amazon's FBA revenue calculator with realistic dimensions and weight — including packaging, since that is what ships.
- Subtract landed unit cost. Everything from the factory to Amazon's warehouse.
- Subtract a PPC allowance. Model it as cost per acquired sale rather than as a percentage of revenue. At launch a meaningful share of your sales will be advertised.
- Subtract a returns allowance appropriate to the category.
- Amortise one-off costs — photography, samples, tooling — across the first order quantity, not across an imagined lifetime.
What remains is contribution per unit. Multiply by a realistic monthly volume — not the market's total, but the share you could plausibly hold from the position you would actually enter at.
Three stress tests
A single-point model tells you almost nothing. Run these three and the answer becomes decision-grade.
1. The price war test
Drop your price ten percent and hold costs. Does the product still contribute? Someone in your category will discount, and if a ten percent move erases your margin you cannot sit through it — you will be forced to either follow them down into a loss or lose position.
2. The PPC doubling test
Double your assumed advertising cost per sale. Launch periods routinely run above steady-state, and competitive categories run above that. If the product only works at a favourable advertising cost, you are betting on a variable you do not control.
3. The slow-mover test
Halve your assumed monthly volume and add the extra storage that implies. Inventory sitting longer costs more per unit and eventually attracts long-term storage fees. Products that only work at high velocity are fragile in a way the base model hides.
A product that survives all three is genuinely robust. A product that fails one is a maybe with a specific known risk. A product that fails two should not receive an inventory order.
Putting the number in context
There is no universal target margin, and any article giving you one is guessing at your circumstances. What matters is the relationship between three things: margin, velocity, and capital.
- A thin margin on a fast-moving product can work — capital cycles quickly and volume compensates.
- A thin margin on a slow-moving product does not, because storage accrues while your capital sits still.
- A healthy margin on a slow mover can be perfectly good, particularly if the category is stable and defensible.
- The dangerous quadrant is thin margin plus uncertain velocity, which is exactly what an optimistic model produces.
Also weigh margin against how contested the position is. Comfortable margin on a page you can hold is a business. The same margin on a page where three better-capitalised sellers will notice you is a temporary situation, because they can afford to compete on price for longer than you can.
Which is why profitability and competitive analysis are one question rather than two. SellRadar runs the fee-stack math against the page's actual price band inside the same analysis that reads the top 20 listings — so "is there money in this" and "can I hold a position here" resolve together rather than in separate tools.
Unit margin is not the same as a business working
Everything above computes contribution per unit. That is necessary and it is not sufficient, because a product can be profitable per unit and still not be worth doing.
The question the unit model does not answer is: what does this product earn against the capital and attention it consumes?
- Capital efficiency. A product with modest margin that turns over quickly recycles your capital several times a year. One with better margin that sits for months does not. The annual return on the same money can favour the thinner-margin product substantially.
- Attention cost. Every product needs monitoring, reordering, review management and occasional problem-solving. A product earning a little is still consuming a share of a finite resource, and a catalogue of marginal products is a common way sellers end up busy and not much better off.
- Downside exposure. A product needing a large minimum order to work at all carries a different risk profile from one that works at a small order, even at identical unit margin.
- Concentration. A product that would be most of your revenue makes the business fragile in a way the unit economics do not show.
A useful discipline is to compute contribution per unit, then annualise it against the capital tied up and ask whether that return justifies the attention. Products that pass the unit test and fail this one are the ones sellers keep for years out of sunk-cost attachment rather than because they earn their place.
Frequently asked questions
What profit margin should I target on an Amazon product?
There is no universal figure — it depends on velocity and capital. A thinner margin works on a fast mover where capital recycles quickly; the same margin on a slow mover loses to storage costs. What matters more is whether the margin survives a ten percent price drop and a doubling of advertising cost, because both will happen.
What costs do sellers most often forget?
Duties and customs, third-party inspection, long-term storage on aged inventory, and returns. Beyond outright omissions, the most damaging error is understating PPC — a new listing has no organic visibility, so advertising is the cost of being seen rather than an optional growth expense.
Should I use the top listing's price in my model?
No. Use the median price across the top 20 organic listings. The top listing usually has review depth and brand recognition you will not have on day one, so its price is not available to you. The median is where the market has settled and is a realistic estimate of where you will end up.
How do I estimate PPC costs before launching?
Model it as cost per acquired sale rather than as a percentage of revenue, and assume launch costs run well above steady state. Then double it as a stress test. If the product only works at a favourable advertising cost, you are betting on a variable that competitors influence and you do not control.
Is Amazon's FBA calculator accurate?
For referral and fulfilment fees, yes — it uses Amazon's own fee schedule rather than an approximation, which makes it the authoritative source for those specific lines. It does not include your product cost, freight, duties, advertising, returns or storage, so it gives you two lines of a much longer stack rather than a profitability answer.
How much volume should I assume when modelling revenue?
Not the market total, and not the top listing's volume. Estimate the share available at the position you would realistically enter at, which is rarely near the top of page one in the first months. Then halve it as a stress test and check whether the product still makes sense with the additional storage that implies.