What is iROAS?

By Harry Burman · Updated October 20, 2026

iROAS, or incremental ROAS, is return on ad spend counted on caused sales only: the orders a holdout test shows would not have happened without the ads, divided by spend. $12,000 of caused sales on $5,000 of spend is an iROAS of 2.4. Ads Manager might show 5.6 on the same spend because it counts credit, not cause.

iROAS stands for incremental ROAS. It equals incremental sales divided by ad spend, where incremental sales are the extra orders a holdout test attributes to the ads. Reported ROAS uses credited revenue from Meta, Google or Amazon Ads; iROAS uses caused revenue from an experiment. The two diverge when retargeting reaches buyers who would have purchased anyway, or when Meta ads send shoppers to Amazon without a tagged path either platform sees. For Meta to Amazon brands, a geo holdout that pauses ads in matched states and compares Amazon ship-to orders is the usual way to estimate iROAS. Compare iROAS to break-even ROAS, 1 divided by margin before ads, to see whether ads pay. Every estimate needs a range, because a few weeks of orders are noisy. The incremental ROAS calculator turns a test result into iROAS and profit after fees.

The iROAS formula

iROAS = incremental sales ÷ ad spend
incremental sales = sales with ads − sales without ads (estimated)

Incremental sales are the topic of incremental sales; the dollar form is incremental revenue. You never observe both numbers for the same people at once, so you estimate "sales without ads" with a holdout group or a model calibrated by tests.

Example: a holdout finds the ads caused $12,000 of Amazon revenue on $5,000 of Meta spend. iROAS = 2.4. The same month, Ads Manager credits $28,000, a reported ROAS of 5.6.

iROAS vs. reported ROAS

Reported ROAS divides credited revenue by spend. Each platform defines credit with its own windows and rules. iROAS divides caused revenue by spend, where "caused" comes from a test design, not a dashboard setting.

Reported can be too high. Retargeting and catalog ads reach people already close to buying; credit counts them as ad-driven.
Reported can be too low. Someone sees a Meta ad, does not click, and searches your brand on Amazon. Neither Meta nor Amazon Attribution sees the full path.

Gordon and others (2019) compared large Facebook experiments with common non-experimental methods and found the latter often misstated ad effects. iROAS from a holdout is the direct estimate; reported ROAS is a fast, biased shortcut.

iROAS when you advertise on Meta and sell on Amazon

Meta's terms bar placing your pixel on Amazon, so Ads Manager cannot count Amazon checkouts. Amazon Attribution counts tagged clicks within 14 days, not view-through paths or untagged brand search. iROAS for this stack usually means a geo holdout test: pause Meta in matched US states, keep ads on elsewhere, compare Amazon orders by ship-to state using difference-in-differences.

Facebook ads for Amazon walks through setup; what is incrementality explains why cause differs from credit.

What is a good iROAS?

One above your break-even ROAS after product cost, Amazon fees and other variable costs. There is no industry benchmark that overrides your margin. A holdout might find iROAS of 2.5 while break-even is 3.2: the ads lose money even if reported ROAS looked strong.

Use the break-even ROAS calculator for the threshold and the incremental ROAS calculator to convert test lift into iROAS and profit.

How to measure iROAS

Export Amazon orders by ship-to state (or ZIP3) for a baseline period and the test window.
Pick test and control regions with matched market testing.
Pause Meta ads in the holdout regions with location targeting; keep spend and creative stable elsewhere.
Compare regions with difference-in-differences; report incremental orders, sales and iROAS with an uncertainty range.

Size the test with the holdout test calculator before you pause spend. Read how lift is calculated.

Where Spillmark fits

Spillmark runs geo holdout tests for brands that spend on Meta and sell on Amazon. It measures what your Meta ads cause on Amazon, takes off Amazon fees and your product costs, and reports the range around lift so you can compare iROAS to break-even ROAS.

Frequently asked questions

What does iROAS stand for?

Incremental return on ad spend: caused sales from a holdout test divided by ad spend. It is also called incremental ROAS.

How is iROAS different from ROAS?

ROAS in a platform report uses credited revenue. iROAS uses sales the ads caused, estimated with a holdout. They often disagree, especially for retargeting and for Meta ads that drive Amazon sales.

Can Ads Manager show iROAS?

No. Conversion lift studies inside Meta measure tracked outcomes on Meta's side. They do not replace Amazon order data for a brand whose sales happen on Amazon.

What is a good iROAS for Facebook ads?

Any iROAS above your break-even ROAS for the product and channel you are testing. That break-even depends on margin before ads, not on a generic benchmark.

How do you calculate iROAS from a geo test?

Estimate incremental Amazon sales in the ads-on regions versus the holdout, divide by Meta spend in those regions during the test, and report a range around the point estimate.