Set your advertising budget from contribution margin per unit, not from a target ACOS you inherited from a forum post. The method has four steps: calculate what a unit actually contributes after every variable cost, convert that into a break-even advertising cost of sale, set your target below break-even by whatever margin the business needs, then budget each SKU separately. Most sellers skip the first step, which makes the other three arithmetic performed on a number that is wrong.
Here is the whole calculation, worked on a real set of figures.
Step 1: Calculate true contribution per unit
Contribution margin is what one incremental sale adds to the business after every cost that varies with that sale. Start from selling price and subtract, in order:
- Landed unit cost, including freight, duty, brokerage, and inbound shipping, not the factory invoice alone
- Marketplace referral fee
- Fulfillment fee
- Monthly storage attributable to the unit
- Expected return cost, expressed per unit sold
- Transaction and prep costs
Fixed costs stay out. Rent, salaries, software, and your own draw do not change because one more unit sold, so including them here produces a number that answers a different question.
Take a product at $34.99. Landed cost $11.20 after freight and duty. Referral fee at 15 percent is $5.25. Fulfillment $5.80. Storage allocated at $0.35. Returns at a 9 percent rate, where a third of returns are unsellable, cost roughly $1.40 per unit sold once you account for the lost unit and the fulfillment already spent.
Contribution per unit: $34.99 minus $11.20, $5.25, $5.80, $0.35, and $1.40. That leaves $10.99.
Step 2: Convert contribution into break-even ACOS
Break-even advertising cost of sale is contribution per unit divided by selling price.
$10.99 divided by $34.99 is 31.4 percent. Spend exactly that on advertising per unit sold and the product contributes nothing. Every point below it is contribution you keep.
This single number is what most advertising decisions should be anchored to, and it is specific to the SKU. A seller running a blanket 25 percent ACOS target across a catalog is simultaneously overspending on the products with thin contribution and underspending on the ones that could absorb far more.
Step 3: Set the target below break-even
Break-even is a ceiling. Subtract from it the margin the business needs out of each sale to cover fixed costs and produce profit.
If the product needs to deliver $6 of contribution after advertising, allowable ad spend is $10.99 minus $6, or $4.99 per unit. That is a target ACOS of 14.3 percent. At 400 units a month, the budget is $1,996.
Notice what this does. It converts an advertising budget from a guess into a derived figure that moves when costs move. When freight rises or the fulfillment band changes, the allowable spend drops automatically, and you find out in the month it happens rather than at year end.
Step 4: Budget per SKU, and split by intent
Account-level budgets hide everything that matters. Two products with identical revenue can have contribution margins of $11 and $3, and the second one cannot support a tenth of the spend the first can.
Within a SKU, separate defensive spend from growth spend. Defending your own branded search terms usually converts at a low ACOS and protects revenue you would mostly keep anyway, which makes its true incremental value lower than its reported efficiency. Discovery and competitor targeting run at a higher ACOS and buy genuinely new customers. Judging them against the same target penalizes the campaigns doing the harder work.
A practical split: hold branded and defensive spend well under target, allow discovery campaigns to run up to break-even on products where a repeat purchase is likely, and cap anything above break-even to a defined test budget with an end date.
Step 5: Re-run monthly
Every input in step one moves. Freight rates change, return rates drift with seasonality, storage spikes in the fourth quarter, and a packaging revision can push a product into a higher fulfillment band without warning. A contribution figure calculated in March and used in November is fiction.
The work is getting the data assembled reliably, which is where sellers stall. Doing it by hand across a few hundred SKUs is not sustainable, and most sellers eventually push the allocation into a system that carries landed cost, fees, returns, and advertising down to the product level. Sellerboard, A2X, and ConnectBooks all handle parts of this, and some sellers build it in a warehouse against raw marketplace reports instead. The choice matters less than picking one and letting it produce the same number twice.
Where this method breaks
Three honest limitations.
It ignores customer lifetime value. For consumables and products with genuine repeat purchase, first-order break-even understates what you can afford to pay, sometimes badly. If you have reliable repeat data, run the calculation on expected lifetime contribution instead. If you do not have reliable repeat data, do not assume it into existence.
It treats advertising as fully incremental, which it is not. Some portion of ad-attributed sales would have happened organically. This makes reported ACOS look better than reality, and the effect is strongest on branded terms.
It says nothing about rank. Spending above break-even to establish a new product can be correct, because the position may hold organic sales afterward. That is a financing decision with a defined budget and an exit date, not an ongoing target, and it should be recorded as such so it does not quietly become the new normal.
The habit worth building
Compute contribution per unit for your top twenty SKUs this week. Derive break-even ACOS for each. Then compare those against what you spend today. In most catalogs the exercise finds at least one product being advertised past break-even and at least one starved of budget it could comfortably support.
That comparison, repeated monthly, is most of what advertising budget discipline consists of. Background guidance on tracking and managing business costs is available through the Small Business Administration, though the per-unit allocation work is specific to your catalog.