ROAS vs LTV: How I Decide What to Scale (and When to Kill It)

Every scaling decision in paid UA comes down to one question: is this cohort worth more than it cost? ROAS and LTV both try to answer it from opposite ends. Lean on either one alone and you’ll either scale too slowly or light money on fire.

What each one actually tells you

• ROAS (return on ad spend) is revenue over spend across a window like D7 or D30. Fast, but early and incomplete.
• LTV (lifetime value) is what a user is ultimately worth, net of fees, refunds and churn. True, but slow to confirm.

ROAS is the speedometer: it tells you how things look right now. LTV is the fuel gauge: it tells you whether you’ll actually make it.

Why ROAS alone gets you in trouble

D7 ROAS rewards whatever monetizes fast, which isn’t always what retains. Scale purely on early ROAS and you can pile budget into a cohort that looks great at day 7 and churns out by day 60. You felt profitable; the P&L disagreed two months later.

Why LTV alone is too slow

Real LTV needs cohorts to mature, which can take months. If you wait for certainty before scaling, your competitors have already bought the market. You can’t run a media account on a metric that arrives a quarter late.

How I actually combine them

• Use early ROAS (D3 to D7) as a fast proxy for daily decisions.
• Build an LTV model from your cohorts and back-test how well early ROAS predicts it.
• Translate both into payback: the months to recover CAC. That’s the number I scale on.

Payback is where ROAS and LTV meet. It’s fast enough to act on and honest enough to trust, because it’s grounded in margin, not gross revenue.

A simple scaling rule

If a cohort clears its payback target and early ROAS is trending to model, scale it in steps the data can confirm. If payback is slipping, hold. The goal isn’t the highest ROAS or the highest LTV; it’s the most profit you can deploy at target payback.

When to kill it

Kill when a channel or creative consistently misses payback after a fair test and a real creative refresh, not after one bad week. Most things that look dead are just under-fed creative. The few that are genuinely underwater, cut without sentiment.