ROAS in marketing:
reading channel results and the blended number

Same ratio, different story

When a marketing report says ROAS 5.0×, the figure may combine several campaigns that behave nothing alike. A branded search campaign can show 10× because it captures people who were already looking for you, while a prospecting social campaign may show 2× and bring in customers who would otherwise never have heard of you. The blended number hides that.

Working out ROAS at channel level, and then blending with the right weights, lets you decide where the next rupee should go, and where an apparently weak number is actually doing a different job.

Channel ROAS and blended ROAS

Because the denominator is total spend, a big channel dominates the blend. A small, very efficient channel moves the overall number only a little.

The same logic extends to any grouping: by campaign, by device, by audience, by product line. Always sum the revenue and sum the spend first, and divide once at the end. Averaging ratios is the commonest arithmetic error in marketing reports, and it nearly always makes weak, low-spend segments look more important than they are.

A two-channel example

Weights matter. Move ₹5,000 from search to social and the blend drops toward 4.0 if each channel keeps its own ratio, even though nothing else changed.

ROAS, CPA and ROI side by side

ROAS and CPA are two views of the same thing when order values are similar. If the average order is ₹1,000 and ROAS is 5, CPA is about ₹200. When order values vary, they diverge, and checking both avoids optimising for the wrong one.

Reading ROAS over time

A single week's ROAS is noisy. Daily numbers swing with promotions, payday cycles and the odd large order, so compare like with like: the same days of the week and the same length of period. Plot a rolling four-week figure beside the weekly one and you will see the direction without the noise.

Seasonality also bends the number. A festival-season campaign can show a ROAS far above its off-season level because buyers are already looking, and the efficiency does not carry over to a quiet month. Judge the campaign against its own season's history and against break-even, not against last month's peak.

Finally, keep a note of changes. When ROAS moves, the cause is often something you did: a new creative, a price change, a different audience or a landing page edit. A short change log turns a mystery in the report into an answer you can act on.

Attribution: who gets the credit

Attributed revenue depends on a rule. Last-click gives all credit to the final touch before purchase, which favours branded search and retargeting. First-click favours discovery channels. Multi-touch models split the credit, and platforms usually report their own view inside their own walled garden.

• Platforms overclaim: each reports conversions it influenced, and totals can exceed actual sales.
• View-through credit: an impression someone scrolled past can be credited for a later sale.
• Offline and cross-device sales are often missing or misassigned.
• Organic and direct sales that would have happened anyway get claimed by whichever ad was seen last.

Be explicit about what you are reporting at the top of every report: the attribution model, the window, the date range and whether revenue is before or after returns. Two people comparing ROAS figures without agreeing on those four items are usually arguing about definitions, not performance.

What to do with a ROAS report

1. 1Rebuild ROAS from your own revenue data wherever you can, with one agreed attribution rule.
2. 2Split brand and non-brand campaigns, and prospecting and retargeting, so the easy wins do not mask the hard ones.
3. 3Compare each channel with its own break-even, which depends on margin and not on a single company-wide target.
4. 4Look at trend and volume; a ratio on 12 orders is not evidence.
5. 5Run a holdout or geo test occasionally to learn how many sales an ad truly adds.

Setting targets by channel

A single target for every channel is convenient and usually wrong. Retargeting and branded search sit close to purchase and should clear a high bar. Prospecting on social or video sits further away and is judged on how many new customers it brings at a cost you can afford, over a longer window.

• Branded search: expect the highest ratio, and treat it as defence of demand you already have.
• Retargeting: strong ratio, limited scale, and prone to claiming sales that would have come anyway.
• Prospecting: lower first-order ratio, but the source of new customers and later repeat sales.
• Email and messaging: very high ratio because the audience is yours; the cost is mostly the list and the content.

When a lower ROAS is acceptable

Top-of-funnel campaigns feed the bottom. If you cut a weak-looking awareness campaign, branded search volume and retargeting audiences often shrink a few weeks later, and the overall ROAS falls with them. That is why teams look at the mix and not each line in isolation.

A defensible target is a blended ROAS above break-even with a margin of safety, plus a reasonable share of spend directed at reaching new customers even if their first-order ratio is lower.