Arbitrage looks transparent: a payout per lead, some costs, the difference is yours. In practice the difference is smaller than expected almost every time, and the reasons repeat.
Three deductions from the payout
- Approval rate. You are paid for accepted leads, not delivered ones. Find that number before the campaign, not after.
- Hold period. Money arrives later, spending happens now. Not a loss, but working capital you need to have.
- Source quality. A lead from cold commenting and a lead from intent search are different goods. If your source approves below average, the headline payout and the banked payout diverge.
The cost side
Accounts, proxies, subscription, model calls — and the time spent finding placements. That last one gets underestimated: a working setup does not last forever, and sourcing new channels is continuous work rather than a one-time task.
Where margin actually lives
In niches where the contact is cheap and the lead is expensive. Cheap contact means audiences sitting in open discussions, where one comment costs one model call. Expensive lead means verticals that pay tens of dollars, not cents.
The inverse — mass-market low-payout offers on cold traffic — rarely clears: the contact cost has a floor and the payout is below it.
Test before you commit
One niche, one or two accounts, two weeks. That is enough to get your own conversion and approval numbers. Scale what already has numbers; ten accounts on a hunch is the expensive way to test one. Sizing the fleet afterwards is covered in this piece.