Every catalogue has a product that everyone in the business believes is the winner. It tops the sales report, it is the one mentioned in meetings, and it is usually where the next inventory order goes. Often it is also somewhere near the bottom of the list when you rank by money kept rather than money collected.
Revenue rank and profit rank are different orderings, and only one of them should drive decisions about stock and ad budget. This guide works through calculating profitability at the SKU level — through COGS, returns and advertising — and is deliberately careful about the part most analyses get wrong, which is attributing ad spend to products.
Why the best seller is rarely the most profitable
Three forces push volume and profit apart, and they compound.
- Price sensitivity. The products that sell fastest are often the ones priced most aggressively, which means the thinnest margin per unit.
- Advertising dependence. High volume frequently means high paid volume. A product that needs $18 of advertising per sale is a different business from one that sells organically at the same price.
- Returns. Categories that sell well — apparel above all — often come back at rates that quietly remove a large share of the gross profit, along with shipping in both directions.
None of these appear on a sales-by-product report. All of them appear in the bank account.
The SKU-level calculation
Product realized revenue (delivered units × price)
− Product COGS (delivered units × landed unit cost)
= Product gross profit
− Linked ad spend (only spend genuinely attributable to it)
= Contribution after ads
Note what is and is not in there. Revenue is from delivered units, so returned and cancelled orders are already excluded rather than subtracted afterwards. And the ad spend line is linked spend, not a share of the total — which is the distinction the rest of this guide is mostly about.
Getting unit cost right
Landed cost, not invoice cost: manufacturing or purchase price plus inbound freight, duties and per-unit packaging. On imported goods the difference is routinely 10–20%, and it lands entirely in your reported margin.
If your supplier price has changed over time, the cost applied to a sale should be the cost of the units that were actually sold. Applying today's price retroactively rewrites the history of every product whose cost has moved, which makes trend analysis meaningless in exactly the categories where cost volatility matters most.
Two products, worked through
Same store, same month. Product A is the quiet one; Product B is the best seller everyone talks about. Illustrative figures.
| Product A | Product B | |
|---|---|---|
| Selling price | $85 | $45 |
| Units delivered | 420 | 1,150 |
| Realized revenue | $35,700 | $51,750 |
| Landed unit cost | $31 | $27 |
| Product COGS | −$13,020 | −$31,050 |
| Gross profit | $22,680 | $20,700 |
| Gross margin | 63.5% | 40.0% |
| Linked ad spend | −$5,400 | −$16,800 |
| Contribution after ads | $17,280 | $3,900 |
| Contribution % of revenue | 48.4% | 7.5% |
| Units returned | 17 (3.9%) | 246 (17.6%) |
| Return handling cost | −$204 | −$2,952 |
| Contribution after return costs | $17,076 | $948 |
Product B produced 45% more revenue and roughly one eighteenth of the contribution. On the sales report it is the star. On the money it is close to breaking even, and a modest rise in ad costs or return rate would take it below zero.
The return handling cost above is the operational cost of processing a return — return shipping and labor, at $12 per returned unit here. The lost margin on those units is already absent from revenue, because returned orders never became realized revenue. The two are separate costs and both are real.
What to do with that
The naive conclusion is "stop selling B". Usually wrong. The useful questions are narrower:
- Is B's return rate a product problem — sizing, quality, a misleading photograph — that can be fixed?
- Is B's ad cost per sale structural, or is it being advertised to an audience that converts badly?
- Is B priced below where the market would tolerate? A $4 increase on 1,150 units is $4,600, which more than quadruples its contribution.
- Does B bring in customers who later buy A? That is a real effect, but it needs evidence from repeat purchase data rather than assumption.
The attribution problem
Everything above depends on the "linked ad spend" line, and that line is where product profitability analysis usually stops being honest.
Some ad spend clearly belongs to a product: a campaign advertising one SKU, a product ad set, a shopping campaign for a specific item. Much of it does not. Brand campaigns sell the store. Broad prospecting sells a category. A campaign can promote six products and report spend at campaign level only. And a customer can see a product ad and buy something else entirely.
The alternative is less satisfying and much safer. Report linked spend against products, report unlinked spend as its own visible total, and let the reader see how much of the advertising the product-level view actually covers. A contribution figure covering 60% of ad spend is useful as long as you know it covers 60%. The same figure presented as complete is misleading.
Reading contribution when attribution is partial
Practical approach:
- Rank, don't judge absolutely. If unlinked spend is distributed roughly evenly, relative ranking between products survives even when the absolute contribution is overstated.
- Treat the figure as a ceiling. Contribution calculated on linked spend alone is the best case; the true figure is lower by whatever share of the unlinked spend belongs to that product.
- Be most careful with products near zero. A product at $900 contribution with substantial unlinked spend in the business is plausibly negative. A product at $17,000 is not.
- Improve linkage before improving the model. Structuring campaigns so spend maps to products is worth more than any allocation formula.
Where product profitability fits
Product-level contribution does not carry the business. Fulfillment, payment processing, salaries, software and warehousing sit below it, funded out of the total contribution of everything you sell. A catalogue where every product shows positive contribution can still belong to a business losing money — see how to calculate ecommerce profit for the full waterfall, and contribution margin for where this number sits relative to gross and net profit.
Common mistakes
Ranking products by revenue
The default sort order on almost every sales report, and the reason low-contribution products get inventory and ad budget they have not earned.
Using a blended margin for every SKU
Applying the store's overall gross margin to each product makes every product look identical by construction. The whole point of SKU analysis is the variance.
Allocating all ad spend across products
Covered above. It converts an unknown into a number, and the number is then treated as a measurement.
Ignoring the operational cost of returns
Excluding returned orders from revenue captures the lost sale. It does not capture outbound shipping, return shipping, inspection and restocking — which are incurred whether or not the unit can be resold.
Deciding on a month with eleven orders
Small samples produce dramatic percentages. A product with a handful of delivered orders can swing from best to worst in the catalogue on two returns. Look at it; do not restructure the range around it.
How ORVX reports this
ORVX calculates profitability per product from delivered units and per-product costs, and reports gross profit, margin and contribution after ads at SKU level.
Ad spend is linked to products where the link genuinely exists. Where it does not, the spend stays visible as unlinked rather than being distributed by a formula — which means a product-level contribution figure in ORVX reflects spend that actually belongs to that product, and you can see how much of the total advertising the view accounts for. Ads Intelligence reports the linked and unlinked split directly.
Frequently asked questions
How do I calculate profit per product?
Start with the revenue from delivered units of that product, subtract the landed cost of those units to get gross profit, then subtract the ad spend genuinely linked to the product to get contribution after ads. Operating costs such as salaries and software sit below product level and are not normally allocated to individual SKUs.
Why is my best-selling product not my most profitable?
Usually some combination of a thinner margin, higher advertising cost per sale, and a higher return rate — the three things that tend to travel with volume. A sales report shows none of them, which is why the ranking changes when you sort by contribution instead.
How should I handle ad spend that isn't linked to a product?
Keep it visible and unallocated. Spreading it across products by revenue or unit share produces contribution figures built partly on an assumption, and they are indistinguishable from figures built on real data. Better to know that your product view covers a given share of ad spend than to have a complete-looking number you cannot defend.
Should returns be subtracted from product revenue?
It depends on your revenue convention. ORVX recognizes revenue from delivered items only, so a returned order was never revenue and is not subtracted again later. Whichever convention you use, the operational cost of handling the return is a separate cost and should be counted on its own.
How many orders do I need before product profitability is meaningful?
There is no fixed threshold, but the smaller the sample the more a single return or refund moves the result. Treat a product with very few delivered orders as an early indication rather than a conclusion, and give it another period before making an inventory or budget decision on it.