Not Every SKU Should Be Advertised

The short version

Automated shopping and catalog campaigns make it trivial to advertise everything you sell. That convenience hides a real decision, because products differ in ways that determine whether advertising them makes money:

Margin varies, often by a lot.
Return rate varies, and returns come out of margin.
Repeat behavior varies — some products lead to a second purchase and some do not.
Competitive pressure varies, so acquisition cost differs by product.

A single account-level ROAS target applied across all of them will overspend on some and underfund others by construction.

The four categories

Sorting the catalog is more useful than optimizing the average.

Advertise aggressively. Good margin, low returns, leads to repeat purchase. These deserve more than an average target allows.

Advertise as an entry point. Thin margin but strong at producing a first customer who comes back. Judged on first-order ROAS these look bad; judged on what the customer becomes, they are your best acquisition tool.

Advertise carefully. Good margin, high return rate. Profitable if the creative sets accurate expectations, unprofitable if it oversells.

Do not advertise. Thin margin, high returns, no repeat. Every ad loses money. These are frequently in the feed because nobody excluded them.

Why they end up advertised anyway

The feed includes everything by default. Exclusion is an action someone has to take.

Automated campaigns allocate on predicted conversion, not on your margin. A cheap, easy-to-sell, high-return product is attractive to the algorithm and unattractive to you.

Reporting is at campaign level. If you never look at product-level profit, the losers are invisible inside an acceptable average.

Bestsellers absorb the blame-free budget. The top sellers look fine, and nobody checks whether the tail is subsidised by them.

Making it visible

The work is joining data the ad platform does not have.

Pull spend and revenue by product.
Apply true margin per product, not category averages.
Apply return rate per product.
Add repeat contribution where you can measure it.
Sort by contribution, not revenue.

The output is usually uncomfortable: a meaningful share of spend sits on products that do not clear their own costs.

What to do about it

Exclude the clear losers from the feed. The single highest-return action, and it is a filter, not a strategy.

Split campaigns by product economics, not by category. Products with similar margin and return profiles can share a target. Products with different ones cannot.

Set targets per group. An entry product and a high-margin product should not answer to the same ROAS number.

Feed the right signal. Where you can send margin or value rather than revenue, automated bidding optimizes toward profit rather than sales.

Practical rule

Sort the catalog by contribution after margin and returns before you tune any campaign. Excluding the products that cannot pay for their own advertising outperforms almost any bid optimization you could run on top of them.