The Goldfinch case study - PRISM

Goldfinch is the case study. Its dashboard showed roughly 20% losses while one investor found the real figure was near 70%. One borrower moved $1.9M of loan money to a struggling parent company. Total documented losses across the protocol reached about $18M, and it wound down in June 2026.

What went wrong, in order

What happened	What it tells you
Reported losses ~20%; actual near 70%	The reporting layer, not the credit, was the failure.
$1.9M of loan money moved to a parent company	Cash left through a door no contract was watching.
~$18M documented losses, wound down June 2026	The structure did not contain it.

Why tranching would not have saved it

No tranche structure prevents any of that. Tranches decide who loses first. They cannot stop money leaving through the wrong door.

A waterfall allocates losses it can see. Goldfinch’s problem was that the losses were not visible, and that cash was moveable. Both of those are upstream of the waterfall.

What PRISM takes from it

Each of the four controls maps to a specific part of this failure:

Money moved to a parent company → collections land where the originator cannot reach them, the single most important protection.
Losses under-reported → weekly collection data with an automatic stop.
Creditors reaching the receivables → the invoices legally belong to the pool company.
No skin in the game → the originator’s own money deposited first.

The uncomfortable part

The boundary where Goldfinch lost control is a boundary PRISM also has. Once cash touches an ordinary bank account, protection depends on a trustee, a contract and a legal system. See the honest limit.