Return on advertising spend is a useful performance metric, but only with accurate attribution.
ROAS connects advertising investment to revenue. The calculation is simple:
ROAS = (Sales Attributed to Ads) / (Cost of Ads)
The result can help marketers compare campaigns and allocate budgets.
ROAS’s usefulness depends on the accuracy of its attribution.
Trackable Sales
For all of its potential, ROAS has an Achilles heel, according to Mike Murphy, vice president of marketing at Incremental, an attribution firm.
ROAS is misleading if the process gives an ad too much or too little value or ignores purchase intent and additional sales.
Imagine an ecommerce company that spends $10,000 on retail media. The store attributes $50,000 in sales to the campaign using a last-touch attribution model. The resulting ROAS ratio is 5:1, meaning that for every dollar spent, the store gains $5 in revenue.
The seemingly obvious decision would be to invest more in retail media advertising. But this example has a problem. Here, “ROAS credits the last ad touchpoint before a sale, whether or not it caused anything,” wrote Murphy in response to my email questions.
“That leaves a very big gap. ROAS can take credit for organic sales that would have happened anyway, or for a sale an earlier touchpoint drove. It also only sees what can be tracked directly,” Murphy wrote.
This is especially true in a retail media environment where buying intent is already very high.
Incrementality
Murphy is highlighting the difference between attribution and incremental sales, or “incrementality.”
- Attribution assigns credit to direct touchpoints.
- Incrementality asks whether the advertising actually created any additional revenue.
In the retail media example, a shopper may have intended to buy the very product advertised. She was, after all, shopping on a retailer’s site. If so, ROAS could differ significantly.
Thus ROAS becomes more useful when marketers can compare attributed and incremental sales, measuring how much business the campaign actually added.
Consider again the company buying $10,000 worth of retail media, perhaps on the Amazon marketplace.
The company’s marketing team finds that about half the time the retail media ad appears, its products are not otherwise visible on the search results page. In those cases, the ad gets full credit for the resulting sale.
But when it appears alongside the products’ organic result, the ad gets credit only for the incremental revenue it produced, lowering the ROAS to 3:1 rather than 5:1. Presumably the organic results would have driven sales without the ad.
Too Little Credit
The attribution problem can also work in reverse, Murphy notes.
An ad may produce revenue that ROAS doesn’t track. Retail media is again a good example.
A shopper might see an ad on the Amazon marketplace and later visit the merchant’s own ecommerce site to consummate the purchase. Or she might encounter the ad on one device and complete the purchase on another.
In either case, the ad may have produced the sale, but the attribution system may not detect it.
Google Ads, for example, uses conversion modeling to attribute sales across devices and privacy restrictions. Google notes that, without modeling, its reported conversions would represent only the observable portion of campaign performance.
ROAS is too low if an ad produced a sale, but attribution failed to recognize it.
Testing
Marketers can test whether attributed sales are actually incremental.
For smaller ad budgets, Murphy suggested a simple holdout test. A company could stop advertising a group of products for several weeks while continuing normal spend for another group, then compare the sales.
The results are directional, not precise, and show the broad impact.
Advertisers with larger budgets and close relationships with retail media networks may have access to randomized or geographic tests unavailable in self-service platforms.
Regardless, the aim is to assess ROAS accuracy.
Source of Truth
If advertising investment rises, sales and profit should too. If ad spending falls, so should sales.
Metrics beyond ROAS can provide context.
Yet bottom-line profit is the ultimate arbiter.
“Your P&L should be your first source of truth — it doesn’t lie,” Murphy wrote.
