Most Amazon advertising reports are built to make the account look busy rather than to show whether the money came back. Impressions climb, clicks climb, a screenshot is shared, and nobody can say what the quarter earned. Amazon ads ROI is a different question from advertised sales, and it is the one a brand owner is really asking when they open a report. Answering it means connecting spend to margin, not to gross revenue, and accepting that some numbers will be estimates.
We run this work for brands selling into several marketplaces from Istanbul, and the pattern is consistent: accounts do not fail because the bidding was clumsy. They fail because nobody defined what a profitable order looks like before the campaigns were built. Once a landed cost per unit exists, and the fee drag on it is understood, the conversation changes. Instead of arguing about whether an ACOS number is good, you can say which campaigns fund the business and which buy volume you cannot afford.
Start with the unit, not the campaign
Before touching a bid, we rebuild the product economics from the seller’s own numbers: cost of goods, inbound freight, the charges Amazon applies to that category and size band, returns at the rate the product actually experiences, and storage cost where the item moves slowly. Amazon revises its fee schedules regularly, so we pull current figures from the account rather than quoting them from memory. What comes out is a contribution margin per unit.
That percentage sets the break even advertising cost of sale. If a unit contributes roughly a third of its selling price after everything else, then advertising that consumes a third of revenue leaves nothing. Everything above that line is loss making in the strict sense, which can still be a deliberate choice during a launch, but it should be a choice and not a surprise. In the accounts we manage, this single calculation usually reclassifies a quarter to a third of campaigns from acceptable to clearly unprofitable.
What Amazon ads ROI actually measures
Attributed sales in the advertising console are not the same as sales caused by advertising. A branded search campaign catches buyers who typed the brand name and would very likely have bought anyway. A broad discovery campaign catches buyers who had never heard of the product. Both report a return, and both numbers are accurate, but only one describes incremental revenue. Treating them as equivalent is the most common reason a profitable looking account produces disappointing bank balances.
The practical fix is segmentation, not a clever formula. We split campaigns into brand defence, category competition and discovery, and hold each to a different target. Brand defence is a cheap insurance policy and should run at a very low cost of sale. Category competition is priced against the margin calculation. Discovery is judged over a longer window, because the first order from a new customer is rarely where the money is.
- Contribution margin per unit, refreshed whenever cost of goods or fees move
- Break even ACOS derived from that margin, set per product and not per account
- Total advertising cost of sale, so organic revenue enters the picture
- New to brand share of orders, as a proxy for whether spend is buying reach
- A repeat purchase window long enough to suit the product’s replenishment cycle
Reporting that survives a finance review
A useful report has three columns nobody can argue with: what was spent, what came back, and what was left over. We build ours from the bulk sheets and the business reports together, so spend sits beside total ordered product sales for the same period. That produces total advertising cost of sale, a blunter and more honest figure than campaign level return. When it drifts upward while campaign metrics stay flat, advertising is cannibalising organic demand.
Where the money usually is
Improving Amazon ads ROI is rarely about discovering an untapped keyword. In the accounts we take over, the recoverable spend sits in four familiar places: search terms converting far below the account average and never negated, automatic campaigns left running long after their harvesting job was done, placement multipliers applied uniformly across products that convert very differently, and dayparting that ignores when the category’s buyers are active. None of it is exciting. All of it is measurable.
The other reliable gain is on the listing rather than in the console. Advertising pays for the click; the detail page decides whether that click becomes an order. When conversion on a page sits well below its category peers, every campaign attached to it is overpaying, and bid tuning will not correct that. We pause aggressive spend, fix images, title structure and bullets, then restart. The same bids produce a different return.
Setting expectations before the first month
Amazon’s attribution windows mean the picture keeps changing after the fact, so a report pulled too early will understate performance. We do not make decisions on data less than a fortnight old, and we do not judge a restructure inside a single month. Ranking, review velocity and organic pull all move slowly, and pulling spend in week two because a chart dipped simply resets the clock.
What we commit to instead is clarity. You should be able to open one sheet and see spend, contribution, and the direction of travel for each product line, with the assumptions written down beside the numbers. When Amazon ads ROI is defined that way, the monthly conversation stops being about whether advertising is working and becomes a straightforward decision about where the next unit of budget goes.
Keep reading: Amazon Ads · Amazon