ROI frameworks that survive a CFO review

The CFO isn't arguing with your attribution model. They're arguing with your math. Here's the three-line framework that survives every finance review we've put it through.

Line 1: Incrementality-adjusted ROAS, not platform-reported ROAS

Platforms report attributed conversions that overlap. Run a 4-week holdout test once a quarter on 10–20% of your traffic. Measure the difference between "got the ad" and "didn't get the ad." That incremental conversion delta is your real ROAS numerator.

Typical finding: incrementality-adjusted ROAS runs 60–75% of platform-reported ROAS across paid social; 75–85% on paid search brand; 95%+ on paid search non-brand for high-intent categories.

The number shrinks. The defensibility multiplies. The CFO is thrilled.

Line 2: Payback period — the number they actually care about

ROAS is a ratio. Payback period is a time. CFOs live in time.

Payback period = CAC / (ACV × gross margin × monthly retention). In clean terms: "how many months before this new customer has paid back what we spent acquiring them?"

• Under 6 months: aggressive reinvestment territory. Grow into it.
• 6–12 months: healthy for most SaaS / consumer-subscription.
• 12–18 months: worth running; watch it quarterly.
• Over 18 months: requires venture capital or a very rich balance sheet. Not a volume-growth engine.

Separate the payback math by channel. It turns out some channels have 3-month payback and some have 22-month payback. The channels aren't all made equal, even if their ROAS looks similar.

Line 3: The kill-when-you-can rule

Every ad, campaign, and channel has a half-life of relevance. The discipline is killing them when their incremental contribution drops below the portfolio average, not when they personally go negative.

Most accounts miss this because "it's still profitable on its own" feels like a reason to keep spending. It isn't — it's a cannibalization flag. Kill when you can, not when you must.

What to stop presenting

• "Attributed revenue" without qualifying which attribution model. Every platform's ROAS is overcounted by 20%+. Saying "$4.2M revenue attributed" without the adjustment is a credibility-killer in finance meetings.
• "Reach" and "impressions." These are inputs, not outputs. Unless you're running brand-awareness campaigns with a separate KPI, cut them from the slide.
• "Projected LTV" on a customer base younger than your reported LTV horizon. If the average LTV is 24 months, don't project it on customers who've been around 3 months.

The one slide that replaces the deck

| Channel | Incremental ROAS | Payback (months) | Decision | |---|---|---|---| | Google Search brand | 7.4× | 3.1 | Hold / grow 10% | | Google Search non-brand | 3.8× | 6.8 | Grow 25% | | Meta Prospecting | 2.1× | 11.2 | Hold | | Meta Retargeting | 5.6× | 4.3 | Grow 15% | | LinkedIn | 1.4× | 18.0 | Review / kill |

One slide. Defensible numbers. An actual decision next to each row. Every CFO says yes.