Why a strong ROAS does not always mean profitable growth

Return on ad spend divides attributed revenue by advertising spend. It is quick to calculate and useful for comparing campaigns on a like-for-like basis. It leaves out almost everything else that decides whether a sale made money: the cost of the product, delivery and packaging, payment fees, returns and discounts. It also depends on an attribution model, which decides how much of a sale each channel is credited with.

None of that makes ROAS a bad metric. It makes it a partial one. Problems start when it becomes the target that decides budget without reference to the economics underneath.

The figures in this section are hypothetical. Consider £10,000 of revenue attributed to £2,000 of media spend: a ROAS of 5. If product costs, fulfilment, payment fees and returns together come to £7,000, then £1,000 remains after media, before overheads.

Now keep the same ROAS but change the product mix, so that the non-media costs rise to £8,500. The same reported efficiency now produces a £500 loss before overheads. Nothing in the ROAS figure warned that anything had changed.

A more useful reference point is the ROAS at which a sale covers its own variable costs. If a product's contribution margin after those costs is 40% of revenue, advertising breaks even at a ROAS of 2.5, which is 1 divided by 0.4. If the margin is 25%, break-even rises to 4. These illustrative margins assume revenue and costs are measured on the same basis. In practice you need to decide whether revenue includes VAT, discounts and refunds, and treat costs the same way.

This is why a single account-wide ROAS target can hide losses. Products with thin margins can sit below break-even while high-margin products sit comfortably above it, and the blended figure looks acceptable. Grouping products into margin bands, and setting a target for each band, makes those differences visible and gives bidding something better to aim at.

A hypothetical product mix

Imagine two brands in the same account; both are hypothetical. Brand A reports a ROAS of 6 on a contribution margin of 15%, so its break-even is about 6.7 and each sale loses money after advertising. Brand B reports a ROAS of 3.5 on a margin of 45%, comfortably above its break-even of about 2.2. On a ROAS league table, Brand A looks like the one to scale. On contribution, it is the one to review.

Margin analysis is only as reliable as the cost data behind it. A few problems recur:

• Trade prices supplied instead of the actual buy price, which changes the margin.
• Costs recorded at the time of each order, leaving older periods without costs and apparently at full margin.
• Ratios averaged across rows. Margin and ROAS must be recalculated at each level of aggregation, not averaged.
• Revenue from the ad platform, from analytics and from the shop mixed in one comparison. Keep each basis separate and label it.
• Out-of-stock products included in averages without a flag.

If the absolute cost basis is uncertain, the relative ranking of products is often still reliable, and that can be enough to decide where the next pound should go. Say so when you present it.

Even with accurate margins, attributed revenue is not the same as additional revenue. A campaign credited with sales from people who were already going to buy can look efficient without creating much new business; brand search is the usual example. Tests, holdout comparisons where volume allows, and a view of new versus returning customers all help to show whether spend is creating demand or collecting it.

• Which costs belong in the contribution measure, and on what revenue basis?
• What is the break-even ROAS for each product group or margin band?
• Which revenue source is used for decisions, and is it labelled?
• How will you judge whether sales were additional, not just attributed?
• How often will margins be refreshed as prices and costs change?