ROAS vs ROI: which marketing metric to use, and when

ROAS and ROI both look like answers to “is our marketing working?” but they answer different questions, and choosing the wrong one is how marketing teams end up reporting healthy numbers while the business loses money. This guide explains both, gives you the formulas, and tells you when each one lies.

ROAS — Return on Ad Spend

ROAS measures the revenue generated per dollar of ad spend.

ROAS = Revenue from ads / Ad spend

A ROAS of 4× means every €1 spent on advertising returned €4 of revenue (gross, not profit).

What it’s good for:

Comparing campaigns, channels and creatives in real time
Optimising bids and budgets at platform level (Meta, Google, TikTok)
Detecting underperforming ads fast

What it doesn’t tell you:

Whether the revenue is profitable (no cost of goods, no overhead)
Whether the customer will churn next month
Whether you’d have got that sale anyway (organic, branded search)

A 5× ROAS sounds great, but if your gross margin is 20% and the product needs €1 of customer support per €10 of revenue, you’re losing money on every “winning” ad.

ROI — Return on Investment

ROI is broader. It measures the net profit per dollar invested.

ROI = (Net profit / Investment) × 100

For marketing, the investment includes ad spend + agency fees + tooling + the share of headcount working on the campaign. Net profit subtracts COGS, overhead and (in B2B SaaS) the cost of servicing the customer.

An ROI of 50% means each €1 of total marketing investment returned €0.50 of profit.

What it’s good for:

Comparing marketing against other capital allocations (R&D, hiring)
Justifying budgets to finance / the board
Long-horizon decisions where churn and lifetime matter

What it doesn’t do:

Move fast enough to optimise daily ad campaigns
Capture brand-building or attribution-resistant effects (PR, organic)
Work without a credible attribution model

ROAS vs ROI side-by-side

Question	Use ROAS	Use ROI
”Should we keep this Meta campaign live?”	✓
“Should we hire two more marketers?”		✓
“Which keyword wins?”	✓
“Is paid marketing as a channel sustainable?”		✓
“Is this a profitable customer cohort?”		✓
“Should I shift budget from TikTok to Google?”	✓

The cleanest mental model: ROAS at the campaign level, ROI at the channel and program level.

When ROAS lies

Three failure modes to watch for:

Branded search. People who already know your brand and would have bought anyway are double-counted as paid wins. Strip branded search out of ROAS to see the real incremental performance.
Last-click attribution. If you only attribute to the last click, you’ll over-credit lower-funnel ads and under-credit awareness. ROAS reported under last-click is systematically biased toward retargeting.
Returns and refunds. ROAS based on order revenue is a forecast, not a fact. Subtract returns and refunds; in e-commerce verticals with >10% return rate this can flip a “winner” into a loser.

When ROI lies

Two failure modes:

Attribution windows. “Marketing ROI of 40%” — over what window? 30 days? 12 months? The longer the window, the more LTV gets counted, the higher the apparent ROI. Quote the window explicitly.
Allocated overhead. Including or excluding the cost of headcount, tools, agency fees and brand campaigns swings the answer wildly. Pick one method and stick to it.

Incremental ROAS — the version that actually predicts profit

The single best metric for performance marketing teams who want to deploy capital well: incremental ROAS (iROAS).

iROAS = Incremental revenue from ads / Ad spend

Where “incremental” means the revenue you would NOT have got without that ad. Measured via:

Geo-holdout tests — turn ads off in matched regions, compare
Conversion lift studies — Meta and Google offer these for paid accounts
Marketing-mix modelling — heavier-weight, statistical approach

Incremental ROAS routinely shows that 30-60% of “ROAS-winning” campaigns drive zero or negative incremental revenue (you were paying for sales that would have happened anyway). We covered this in detail in Incremental ROAS vs ROAS.

How to use both metrics together

A pragmatic operating cadence:

Daily: track ROAS at campaign / ad-set level. Pause anything below your floor (typically 2.0× for e-commerce, 3.5× for B2B lead gen).
Weekly: review ROAS by channel. Shift budget toward the winners but don’t yet retire losers — give them 7-14 days.
Monthly: compute incremental ROAS via your test plan (always have one running). Adjust channel budgets based on iROAS, not headline ROAS.
Quarterly: compute channel-level ROI (with allocated overhead, LTV attribution). This is what goes to the board.

Benchmarks

Vertical	Healthy ROAS	Healthy marketing ROI
E-commerce DTC	3.5-5×	30-50%
Subscription / SaaS	4-6×	50-100% (over 12 months)
Marketplace	2.5-3.5×	40-80%
Local services	3-5×	100%+

These move with your gross margin: a 70% gross-margin product can be profitable at 2× ROAS; a 25% gross-margin product needs 5×+ just to break even on the marginal sale.

FAQ

Which is better, ROAS or ROI?

Neither — they answer different questions. ROAS optimises ad tactics; ROI evaluates the marketing program. Use both.

What’s a good ROAS for Meta / Google ads?

It depends on your gross margin. As a rule of thumb, ROAS needs to be at least 1/gross-margin to break even. A 25% gross-margin business needs 4× ROAS just to not lose money.

How is ROAS different from CAC?

ROAS is revenue-based and short-term. CAC is per-customer, focused on acquisition cost. CAC pairs with LTV (LTV/CAC ratio); ROAS pairs with gross margin to determine break-even point.

Can ROAS be negative?

Not directly — ROAS is a ratio, so it’s zero or positive. But it can be below 1× (you spent more than the revenue), and combined with low margins that means the campaign is losing money.

NextScenario unifies ROAS, iROAS, CAC and gross margin per channel in a single dashboard — connecting to Meta, Google, TikTok, Stripe and your bank. Book a 30-min demo — we’ll show you your real iROAS by channel with your real data.

For the Spanish version of this guide, see ROAS vs ROI: ¿Cuál es la mejor métrica?.