Paid media strategy for ecommerce: choosing channels by margin, not by ROAS

Ask about money before you ask about audiences. Five questions cover most of it.

1. What does an order leave behind? Our convention: contribution margin is revenue minus the variable costs that grow with every order, before marketing. That means cost of goods, shipping, payment fees, returns and discounts. Whatever remains after ad spend is contribution after marketing. Other sources draw the line differently, so say which one you mean.
2. What does a new customer cost? Use new-customer CAC (nCAC), not blended CAC. Repeat buyers flatter blended numbers.
3. How long until the margin repays the spend? That is the payback period, counted on margin, not on revenue.
4. What is the ROAS floor? A finance adviser's guide (a vendor) puts it plainly: break-even ROAS is 1 divided by contribution margin, an algebraic identity rather than a benchmark¹. Break-even is the floor. It is not the target.
5. Would these sales have happened anyway? If yes, the channel is collecting credit, not creating demand.

Example (illustrative): An order sells for €100. After goods, shipping, payment processing and returns, €40 is left. Contribution margin is 40% of revenue. Break-even ROAS is 1 / 0.4 = 2.5. A channel reporting 3.0 clears the floor on paper. If half of those orders would have come anyway, it does not.

For the full chain from ROAS to cash, see ROAS looks good but no profit.

Our view: give each channel a job, then judge it on that job. A channel judged on the wrong job always looks bad.

Before you add one, check whether the current channels are saturated. A new channel can be right. It can also be a way of not facing a weak offer. For the channel-by-channel view, see Meta vs Google vs TikTok vs Reddit ads for ecommerce. We cover individual channels in TikTok Shop and GMV Max profitability, reddit ads for ecommerce and ads in ChatGPT.

Both big platforms now push automated campaign types. They optimise towards the goal you set. They do not know your margin.

Performance Max is a goal-based Google Ads campaign type that reaches Google's inventory from one campaign. Google documents the controls you keep: brand exclusions, negative keywords, search themes, final URL expansion and audience signals². Placements and most bidding choices are the platform's. Read more in Performance Max black box.

Advantage+ sales campaigns are Meta's automated sales campaigns, where machine learning handles audience, creative, budget and placement. Meta's own developer blog reports results for them³. Those are Meta's claims about its own product. Do not use them as proof for your account. We cover the trade-offs in Advantage+ vs manual campaigns.

What stays with you:

• The conversion goal and the value you feed it. A wrong goal scales the wrong thing.
• The product feed and the creative.
• Exclusions, such as brand terms and existing customers.
• Measurement. The platform marks its own homework.

Spend can scale faster than the business behind it. The campaign is ready. The warehouse would like a word.

Three limits to check before you raise a budget:

• Margin by product. Best sellers can take most of the spend. If they carry a thin margin, scale pushes you towards the products that pay least.
• Stock. Ads for items you cannot ship are a refund queue in waiting.
• Cash. You pay for ads before the customer pays back. A longer payback period means more cash tied up for longer.

Name the ceiling before you raise spend. Growing sales is lovely. Keeping some of the money is also worth a meeting.

Platform ROAS is a poor common yardstick. Each platform counts conversions its own way. A measurement vendor argues that platform ROAS can overstate results, for example through brand search and retargeting, and understate them, for example for video or in-store sales⁴. That is one vendor's argument, and the vendor sells measurement. Treat it as a reason to check, not as proof.

A fairer comparison uses three layers:

1. Shop-level numbers. Total revenue and contribution margin against total marketing cost. Our convention: MER (marketing efficiency ratio) is total revenue divided by all marketing cost, including agency, creative and tools; blended ROAS is total revenue divided by media spend only. Not everyone draws this line, so say which one your report uses.
2. New-customer cost and payback per channel, from your own order data.
3. Incremental ROAS (iROAS), where you can get it: revenue caused by the ads, divided by spend. A test is one way to get it, not the only one.

Then look at marginal ROAS, the return on the next euro rather than the average. A channel can have a good average and a bad next euro. That is the number that decides whether to scale.

Cut or cap on marginal return, not on habit. Three signals are worth a review:

• The next euro returns less than your break-even, even if the average still looks fine.
• New-customer cost keeps rising while payback gets longer.
• The channel only looks good until you take out brand search, retargeting or existing customers.

Cutting is not always the answer. Sometimes you cap spend, change the goal or fix the feed. Write down what you expect to happen to total sales, then check. If the shop-level numbers do not move after a cut, the channel was probably collecting credit.

For ZFD: Our view is that margin and payback come first and channels second; each channel service can also be bought separately.