Decision rule: how much to risk on an unproven partner

Decision rule: how much to risk on an unproven partner
The mistake is not working with unknown partners. It is committing before one has completed a full payment cycle.
The staging I use:
— Stage one: minimum viable volume. Enough to generate a real invoice, small enough that losing it is annoying rather than damaging. The goal is not profit. The goal is watching one payment cycle complete.
— Stage two: once that payment arrives on time and matches your numbers, scale to a level you could still absorb losing entirely.
— Stage three: only after two clean cycles does a partner get to be a meaningful share of your volume.
— Never let one partner hold the majority of your receivables, regardless of how good they have been. Good partners get acquired, run out of funding, and change policy.
And at every stage, keep your own tracking. If you cannot reconcile independently, you are not testing them. You are trusting them.
Trust is built one settled invoice at a time.