Bidding in Shopping: target ROAS by margin, real sales and POAS

In the Basic level we said "ROAS without margin tells you nothing". This lesson turns that into a method: how to bid in Shopping to make money, not revenue. Because the trap in Shopping is that everything is measured in euros of sales, and Google optimises towards the euros you show it.

The single-ROAS mistake

A catalogue with products at a 15 % margin and products at a 50 % margin, under one target ROAS of 4:

Same ROAS, one wins and the other loses. And Smart Bidding, which only sees the €100 of revenue, will push both equally. Target ROAS has to be different by margin: break-even is 1 / margin (A: 2; B: 6.7), and the target must sit above it.

Targets by margin group

With the margin label in the feed (custom_label_0) and the structure from lesson 2, each group gets its own target:

Three campaigns (three groups inside one portfolio with different targets is not possible: one portfolio = one target; hence separate campaigns by margin). And an honest decision: very low-margin products are usually better excluded from paid Shopping, unless they pull in orders alongside other products (basket size) or act as a strategic hook.

Real sales, not measured revenue

What Google counts as "conversion value" is what your tag sends it at the moment of the order. Reality arrives afterwards:

• Returns (in fashion, 20-40 %).
• Cancelled or unpaid orders (cash on delivery).
• Fraudulent orders.
• VAT included in the value (if you send it with VAT, ROAS is inflated by 21 %).

Google's ROAS can be 30-50 % better than the real one. Two ways to correct it: send net values (excluding VAT, and adjusting for returns with conversion adjustments or offline uploads) and judge with the ecommerce's real sales matched order by order (module 8, Measurement II).

POAS: return on profit

POAS = gross profit generated / advertising cost. A POAS of 1 is break-even (the profit from sales pays for the ads exactly); above it you win; below it you lose. Compared with ROAS, it incorporates each product's margin and does not mislead across categories.

Per-product calculation: (real revenue − product cost) / ad cost. With the margin label (or a table of margins by product or category) you can work it out in a spreadsheet; without it, you cannot.

And the natural evolution: bidding on profit — sending Google the order's margin as the conversion value instead of the revenue, and setting a target ROAS on that value (which then is a target POAS). Smart Bidding starts optimising towards real money. It requires the website to calculate the margin per order at the moment of conversion; on modern platforms that is feasible.

Signs that you are bidding badly

• A "good" account ROAS and profit that does not grow → low margin subsidised by high margin.
• Products that sell a lot and are the lowest-margin ones → the algorithm has found the easy path.
• High returns on the products Shopping pushes hardest → measured value ≠ real value.
• A budget-limited campaign with a high ROAS → there is room to grow… if the ROAS is real.

💡 Ninja trick: the Shopping Ninja works with real sales (the ones the ecommerce confirms, not just the tag's value), with the per-product or per-category margins from its spreadsheet, and calculates ROAS and POAS per product with a traffic-light Dashboard: green wins, red loses even when the ROAS looks good. Its checkbox actions —exclude loss-making products, negativise, improve titles— come from that reading, not from revenue.

What you should remember

• A single ROAS = a mistake: break-even is 1/margin and it changes per product.
• Campaigns separated by margin with their own target ROAS; very low margin, often exclude.
• Measured value is not the real sale: returns, cancellations, VAT.
• POAS (profit / ad cost) is the indicator in charge; bidding on profit is the evolution.