Paid growth is rented. Loops are owned.

A chart that goes up and to the right on the back of ad spend is not growth. It’s a lease. The moment you stop paying, the line stops — that day, not next quarter — and everything you “built” turns out to have been rented by the click.

That’s not an argument against paid acquisition. It’s an argument for knowing which of your two very different growth engines you’re actually running, because they behave nothing alike, and one of them lies to you.

Start with why paid quietly gets worse as it gets bigger. When you first turn on a channel, you buy the cheapest, best-fit users available — the people practically raising their hands. To spend more, you have to reach further: worse inventory, looser targeting, people a step less interested. At the same time the auction you’re bidding in notices your demand and raises your price. So the marginal cost — what it costs to acquire the next cohort — creeps up, even while your blended cost (averaged across all the cheap early users too) still looks fine. Blended CAC is a comfortable number that hides a rising one underneath it. You feel efficient right up until the marginal user costs more than they’re worth.

The deeper problem is that paid demand doesn’t compound. Every user is a fresh purchase. Nothing about acquiring this cohort makes the next one cheaper or more likely — you just buy again. It’s a treadmill: you can run hard and stay in the same place, and the belt only moves while your legs do.

A loop is a different machine. A referral, a shared artifact, a piece of output worth showing a colleague, a network that gets more useful with each person — in a loop, each cohort helps produce the next one, at close to zero marginal cost. That’s what compounding actually means, and it’s why a working loop grows on a rising floor instead of a flat belt. (It only compounds if people stay — a loop poured into a leaky bucket still drains; retention is the floor the loop stacks on.)

None of this makes paid the enemy. Paid is a fine catalyst: use it to seed the hard side of a loop, to bridge to a payback period you understand, or to buy time while the compounding engine warms up. The failure isn’t spending money; it’s spending money instead of building the loop, and mistaking the treadmill’s motion for progress. Because paid keeps the top-line healthy, it’s the perfect place for a dead loop to hide — you keep feeding the machine and never notice there’s nothing underneath.

So measure the things the blended number hides. Watch marginal CAC — the cost of the next cohort, not the average of all cohorts. Watch the payback period, and whether it’s stretching as you scale. And watch the one ratio that tells you which machine you’re really running: the share of new users that arrive without paying for them. If loop-driven share falls as you pour in more spend, you’re not scaling growth, you’re scaling rent. If it holds or rises, the spend is doing its real job — priming something that compounds.