Judge Amazon Ads by Total Profit, Not by ROAS

ROAS is the number most Amazon ad reports put first, and it is easy to like: a 4.0 means every ad dollar was followed by four dollars of sales. The trouble is that sales are not profit. ROAS ignores what the product costs, what Amazon charges, what comes back as returns and whether the shopper would have bought anyway. Two products can share the same ROAS while one makes money and the other loses it. This article replaces ROAS with questions that end in profit: what each ad order earns, what a new customer costs, what that customer is worth over time and what your ads do to organic sales.

What ROAS leaves out

Return on ad spend is ad-attributed sales divided by ad spend. It is the mirror image of ACoS: ROAS equals 1 divided by ACoS, so a 4.0 ROAS is the same thing as a 25% ACoS. Both describe one ratio from opposite ends, and neither includes any cost except the ad itself.

Three things are missing. The first is margin. You keep only what is left of each sale after product cost, Amazon fees and returns, so a ratio of sales to ad spend cannot tell you whether an order paid for itself. The second is timing. Returns usually arrive after the sale, so a recent report can look better than the final result. The third is causation. An ad gets credit when a shopper clicks and then buys within the reporting window, even if that shopper was already on the way to buying. A search for your own brand name is the clearest example.

A small hypothetical shows the gap. Two products sell for $40 and both run at a 4.0 ROAS, which is a 25% ACoS, so each ad order costs $10 in advertising. Product A earns $14 per unit before ads. Product B earns $8. Product A keeps $4 after ads, and product B loses $2 on every ad order.

The fix is to compare ACoS with each product's break-even ACoS, which is profit before ads divided by price. Product A breaks even at 35% ($14 ÷ $40), so 25% is comfortable. Product B breaks even at 20% ($8 ÷ $40), so 25% is a loss. The same campaign result is good news on one product and bad news on the other.

Make profit per ad order your weekly number

Work out profit before ads for each product: the price, minus product cost, Amazon's fees and an allowance for returns. Then work out ad cost per order, which is ad spend divided by ad orders. The difference is profit per ad order, and it is the number to track by product every week.

A negative number is not automatically a problem. It is a problem when you cannot say what you are buying with it. Spending ahead of profit is reasonable for a launch with an end date, for a first order from a customer who is likely to buy again, or for holding a top position against a competitor. Spending ahead of profit with no reason attached is just a leak.

Roll the numbers up by product, not only by campaign. Accounts often have a few products that carry the profit and a few that quietly drain it, and an account-wide ROAS averages the two together. Keep brand-name campaigns apart from non-brand campaigns as well, because they do different jobs and deserve different targets.

Count the sales your ads did not cause

Ad reports cannot tell you which attributed sales were incremental, meaning sales that would not have happened without the ad. Brand-name campaigns are the obvious suspect, since shoppers who type your brand are likely to find you anyway. Defending those terms can still pay, because a competitor could otherwise take the top spot, but you should know that part of the credit is borrowed.

The total-business view fills in what the ad console cannot. In your sales reports, take total units for a product and subtract ad-attributed units to estimate organic units. Then watch TACoS, which is total ad spend divided by total sales, counting all sales and not only ad sales. If ad spend rises, TACoS falls over several months and total sales grow, the ads are probably feeding organic demand. If total sales stay flat while ad spend climbs, much of the attributed revenue was probably business you would have won anyway.

Treat organic lift as a hypothesis, not a rule. Amazon does not publish how ad-driven sales affect organic ranking, and price changes, promotions, stock levels and season all move organic sales on their own. A cleaner check is to change one thing at a time: lower the budget on one product, keep its price and stock steady for a few weeks and watch total units rather than ad units.

What a new customer costs, and what they are worth

Customer acquisition cost is ad spend divided by the number of new customers it brought in. Amazon does not hand you customer lists, so the practical stand-in is new-to-brand data: orders from shoppers who have not bought from your brand within a lookback period. Some ad types report it, including Sponsored Brands and Sponsored Display. Use only the spend and the new customers from the same ad types. Dividing all of your ad spend by a partial customer count overstates the cost.

In a hypothetical month, $1,800 of Sponsored Brands spend that brings in 150 new-to-brand customers gives an acquisition cost of $12 per customer.

Lifetime value should be counted in profit, not revenue. Take the profit before ads on the first order and add the expected profit from repeat orders. Suppose a new customer's first order earns $9 before ads, and customers like this place 0.5 more orders on average over the period you care about. Repeat orders add $4.50, for $13.50 in total. A $12 acquisition cost leaves $1.50 per customer, even though the first order alone lost $3 after ads.

Be conservative with the repeat estimate. Brand Analytics includes repeat purchase reporting for brands that have access, and your own order history is another source. A consumable that shoppers reorder can justify a higher acquisition cost than a one-time purchase, where the first order is the ceiling. Watch payback time as well. Profit that arrives over a year is little help if you need the cash this quarter.

A review rhythm that ends in profit

These numbers only help if you look at them on a schedule. A one-page sheet is enough: for each product, list the price, profit before ads, break-even ACoS, current ACoS, profit per ad order and TACoS. Update it from the same reports each time so the changes are comparable.

Write down why you changed a bid, a budget or a price, and what you expect to happen. A month later the note tells you whether the change worked, instead of leaving you to guess from memory.

The short version

• Convert ROAS to ACoS and compare it with each product's break-even ACoS (profit before ads divided by price).
• Track profit per ad order by product every week.
• Judge brand-name and non-brand campaigns separately.
• Watch total sales and TACoS over months, not only ad-attributed sales.
• Calculate acquisition cost only from ad types that report new-to-brand data.
• Estimate lifetime profit conservatively and spend only part of it on acquisition.
• Recalculate margins every quarter, since fees, costs and prices move.