ROAS is a snapshot. It tells you what happened on the first purchase. It doesn't tell you whether that customer will ever buy again, refer someone, or cost you a fortune in support. If you're optimising for ROAS alone, you're often optimising for the wrong customers — the one-and-done buyers who look good on paper but never come back.

What goes wrong when ROAS is the only lever

Brands scale spend on channels and creatives that deliver the best first-purchase ROAS. Then they notice retention is weak, refunds are high, or LTV is nowhere near what they assumed. The creative that crushed for "add to cart" might be attracting price-sensitive one-time buyers. The audience that converted cheap might have a terrible repeat rate. You only see that when you look at LTV by cohort and by source.

What LTV-first actually means

You're willing to pay more for a customer if their lifetime value is higher. So the metric that matters is LTV ÷ CAC (or some variant like payback period). That means you need a working LTV model — even a simple one. For ecommerce, we often use 12-month or 24-month revenue per customer by cohort. For subscription, it's LTV from your normal churn and ARPU maths. The point is to have a number you can attach to each acquisition source and creative segment.

How we build the model

We pull cohorts by first-touch source (and optionally by campaign or creative bucket). We track revenue from those cohorts for 6–24 months depending on the business. That gives us LTV by source. We divide by CAC for that same source and get LTV:CAC. Channels or segments with a higher ratio get more budget; the rest get cut or tested. We've seen brands shift 30–40% of spend from "best ROAS" to "best LTV:CAC" and end up with higher total profit and more repeat revenue.

Getting the data in shape

You need first-touch (or consistent) attribution so that a customer's LTV is tied to where they came from. If your analytics only do last-click, you'll have to approximate — e.g. use UTM at first visit and persist it. It's not perfect but it's way better than ignoring LTV. Then slice by device, creative theme, or audience so you can see which type of acquisition actually pays back. That's when acquisition stops being a cost centre and starts being an investment with a return.