Shopping profit by product: the losers campaign ROAS hides.

Most Shopping accounts are judged at the campaign level, and most campaigns contain dozens or thousands of products. The campaign ROAS is a weighted average of all of them. An average can be comfortably above target while several of its members lose money on every sale.

The reason is simple. Every product has its own margin, so every product has its own breakeven ROAS. A single campaign target treats them as if they were the same.

Five products, one campaign

Here is a month in one Shopping campaign, as arithmetic rather than any client’s result. Margin here is revenue less product cost.

The campaign is profitable at 3.7x, comfortably above its blended breakeven. It is also carrying three products that lose money. B and D have identical ROAS, 3.3x, and both are underwater, because their margins are thin enough that 3.3x does not cover them. E has the worst ROAS in the campaign and loses the least, because its margin is healthy. Sorting this campaign by ROAS would point you at the wrong product first.

Stop the losses on B, D, and E without losing their sales, and the campaign’s profit after ads rises from $15,900 to $20,000. The average never showed that $4,100 was available.

What margin should include

The table uses product cost alone to keep the arithmetic readable. A real calculation should subtract everything that scales with the order, because each of those costs varies by product too.

• →Product cost, from the store or the feed.
• →Shipping you pay for, which is often a much larger share of a cheap product’s price than an expensive one’s.
• →Payment and marketplace fees.
• →Returns and refunds, by product. A dress with a high return rate has a lower real margin than its price tag suggests.
• →Discounts, when a product is usually sold on promotion.

What is left is contribution margin per product. Breakeven ROAS is one divided by it. The breakeven ROAS calculator does the arithmetic for a single product.

The same two products, with every cost

Now take products A and B from the table and apply the rest of the costs. Suppose A ships cheaply and rarely comes back, while B is bulky and often returned. As arithmetic:

Product A is still comfortably profitable. Product B went from a modest loss to a serious one, because a 9% contribution margin needs more than eleven dollars of revenue for every dollar of spend. That is a target no bidding strategy will reliably hit, which means the question for B is not really about bids. It is about price, shipping, or whether B should be advertised on its own at all.

Getting the numbers

The ad platforms know spend and attributed revenue by product. They do not know your costs unless you tell them. The store usually does: Shopify records product cost per variant, and order data gives refunds by line item. Put the two together by product ID and the table above falls out.

This is part of how we read Shopping accounts for clients. Orders, line items, product cost, and refunds come from the store, spend and claimed revenue come from the platforms, and the two are joined product by product, so the conversation is about margin rather than about a campaign average.

Restructure around margin bands

1. Group products into three or four margin bands, for example under 30%, 30% to 50%, and over 50%.
2. Write the band into the feed as a custom label, so it can drive structure in Shopping and Performance Max.
3. Split campaigns or asset groups by band, and give each one a target ROAS set from that band’s breakeven, not from the account average.
4. Recalculate the bands when costs, prices, or return rates change, and at least once a quarter.

Applied to the five products above, the split is simple. A, C, and E have margins of 50% or more, so they share one campaign with a target of about 2.2x, above all three breakevens. B and D, at 25% and 20%, go into a second campaign with a target of at least 5.0x, which covers D and gives B headroom. The thin-margin campaign will probably spend less. That is the point: it stops buying revenue that costs more than it earns, while the high-margin campaign keeps a target it can actually hit.

Expect the thin-margin campaign’s volume to fall and its profit to rise, and judge both campaigns together on profit after ads over a month rather than on either one’s ROAS in the first week. If the thin-margin products stop selling entirely, that is information too: at their current prices and costs, paid search was not a profitable way to sell them.

Bidding to profit directly is the next step, and it is covered in bidding to margin with POAS. Margin bands are the practical halfway house. They work with the bidding strategies you already use, and they stop a single target from quietly subsidizing thin-margin products with fat-margin ones.

Performance Max and Shopping specifics

• →Targets are set per campaign, not per product. If products need different targets, they need to be in different campaigns. Asset groups and listing groups control which products are shown, not what return each one must earn.
• →Custom labels are the practical bridge. The feed allows five of them, so a margin band can sit beside labels you already use for season, price, or bestsellers.
• →Attributed revenue is usually credited to the product that was clicked. Google’s conversions with cart data can report the products actually sold in the order, which matters when people click one item and buy another.
• →A campaign with too few conversions per band will struggle to learn. When bands are small, merge the two nearest and accept a slightly blunter target rather than splitting volume the system needs.

A review cadence that keeps it honest

Margins drift. Suppliers change prices, shipping rates rise, and a product’s return rate can change with one bad batch. A monthly review of profit after ads by product catches most of it: sort by profit, not ROAS, and look first at the products losing the most money in absolute terms. A quarterly review resets the bands themselves. And every stock change deserves a glance, because advertising a product that is about to sell out spends money on demand you cannot serve.

Before you cut a loser

A product that loses money on its own sales may still be doing a job. Check three things before switching it off.

• →Is it an entry product? If it brings in first-time customers who come back for higher-margin products, judge it on the value of those customers, not the first order. The split between new and returning customers answers this.
• →Does it sell other products in the same basket? Attributed revenue sometimes credits only the clicked item. The order may tell a better story.
• →Is the loss temporary? A product on clearance, or one whose cost just rose, may need a price change rather than an ad change.

If the answer to all three is no, lower its target or move it to a band with a stricter one. Exclusion is the last step, not the first, because a product with no ads also has no data.