A private credit fund model has to check one thing a cash check can't: that the total out on loan never goes above the money investors put in. Give the model one row that compares the two every month and flags the first month the fund goes over.
The client is a North American credit manager raising its third fund, a $100M vehicle. The fund lends in two ways. About half the money backs warehouse lines, which are credit facilities for other lenders. The rest goes into direct lending, from term loans and revenue-based deals to structured equity and short-term opportunistic credit.
We built the fund's model with the manager's team. It runs month by month for 84 months, the full life of the fund. It has about 15,000 formulas across 17 sheets, and every input lives on one Assumptions sheet.
Idle money was the hard part
Warehouse lines start slowly. A lender might sign for a big line and use only part of it in the first year. The manager didn't want the unused money sitting still, so the model lends it out through the direct lending products and holds it ready for when the warehouse lines grow. The team called this double deployment.
It's useful, and it can go wrong fast if timing slips. The manager asked for every safety rule to be a lever they could move, such as a minimum cash buffer and a cap on how long short-term deals can run.
So the model hands out cash in a strict order. Warehouse lines get first call. Before any cash reaches direct lending, a forward reserve looks at what the warehouse lines will need next month and holds that amount back.
The fund model audit: every check passed but one
Before the manager relied on the model, we audited all of it. Most of it held up. The balance sheet balanced in all 84 months, and cash never went below zero.
Then we added up what the fund had out on loan. In three months of its third year, the total came to more than $100M. The fund was lending money it never raised.
The cause was simple. Each direct lending product had its own cap, and each one stayed inside it. Nothing capped the total. A cash check couldn't catch this, because it only asks whether the bank balance is above zero. It never asks whether the fund lent more than investors put in.
The smaller faults around it
The audit listed ten items in all, and four of them were must-fix. The row meant to total the fund's loan losses was blank, so it showed zero while the product sheets were booking losses. Two products sent the capital engine their balances before losses instead of after, which overstated how much was deployed. And the main warehouse sheet had its start month typed in as month 6, while the Assumptions sheet said month 1. One of those two numbers was wrong, and nobody could tell which from the model.
We wrote each finding up with the exact formula change. In the next version, the warehouse start month reads from the Assumptions sheet, and the capital engine reads the two products' current balances after losses. For the big one, we set out a fund-wide check row and a cap that stops new lending at whatever is left of the $100M.
What we built
- An 84-month fund model, 17 sheets, driven from a single Assumptions sheet
- A capital engine that funds warehouse lines first and reserves next month's warehouse cash
- Loan tracking by monthly batch, with 84 batches per direct lending product
- An audit report with ten ranked findings and a formula fix for each one
- A capabilities report that sets out what the model does and what it doesn't do yet
Why it matters for your fund
Investors in a credit fund read the model to see how their money moves. Every income line and every distribution sits on top of the lending numbers. If the model lends money the fund doesn't have, the figures built on it can't be trusted either.
A balance sheet that balances tells you the sums add up. It doesn't tell you the fund stayed inside its own size.
Give your fund model one row that compares the money out on loan with the money raised, every month, and make it shout the first month it goes over.
Questions founders ask
How do you build a private credit fund model?
Run it month by month for the full life of the fund, with every input on one assumptions sheet. Give each lending product its own schedule, and track loans by the month they're made so repayments and losses follow each batch. A capital engine then sends cash to each product in a set order. Add checks that the balance sheet balances, that cash stays above zero and that total lending stays inside the money raised.
How are private credit funds structured?
Many are closed-end funds. Investors commit money, the manager puts it to work in loans over time, and interest and repayments flow back to investors over a fixed life, often several years. The fund in this case split its money between warehouse lines, which are credit facilities for other lenders, and direct lending such as term loans, revenue-based deals and short-term credit.
Can a fund lend more than its committed capital?
Not unless its terms allow borrowing or reinvesting repayments, and even then there are limits. In a model, the usual risk is a cap on each product and no cap on the total. Each product stays inside its own limit while the fund as a whole goes over, and a cash check won't catch it. A fund-wide row comparing money on loan with money raised will.