Five findings in one quarter-end close, and not one of them was a modelling error. Every one was a control that did not exist.
Because a close that agrees with itself only one way has not been checked. It has been computed. Those are different things, and the difference is usually discovered by an auditor rather than by the controller.
Net asset value can be arrived at from two directions. A fund's books permit both, and a well-built close computes both independently and then compares them.
| Route | What it is |
|---|---|
| Partners' capital roll-forward | Opening capital, plus contributions, less distributions, plus the period's allocated income and unrealised movement, allocated across the investor register. |
| Net assets from the balance sheet | Investments at fair value, plus cash and receivables, less payables, accrued expenses, accrued carried interest and any borrowings. |
The two must produce the same figure. They are built from different records, by different processes, and they touch different controls. When they agree, the close has been verified. When they disagree, the difference is the finding, and the size of it tells you where to look.
It is 30 September. The administrator's draft trial balance has arrived. Five things are wrong with it. Working the corrections through, the bridge runs from a draft partners' capital of $125m to a corrected net asset value of $128.5m — and the balance sheet route, computed independently, agrees at $128.5m.
Three and a half million on a $125m fund is 2.8 percent of net asset value. It is well inside the range that would not be visible in a variance review, and well outside the range that anyone would accept in a statement sent to investors.
The finding that matters is not the $3.5m. It is that none of the five errors was a modelling mistake. Every one of them came from a control that did not exist. And every one was found the same way: by connecting two records that should have agreed, and noticing that they did not.
Three ties do most of the work in a private fund close. Each one connects two independently maintained records, which is exactly why each one catches errors that a single-route computation cannot.
The sum of every investor's capital account must equal the fund's net assets. This is the tie that stops a bad quarter becoming a restatement, because it is the one that catches allocation errors — income allocated on the wrong participation percentages, a late-closing investor not equalised, a transfer recorded on one side only. The fund total can be perfectly correct while the split across investors is wrong, and only this tie sees it.
Cost and fair value per the investment schedule must agree to the ledger balances. This catches valuations approved by the valuation committee but never posted, and posted movements that no longer match an approved valuation.
Bank balances reconcile to the ledger, and every call and distribution ties to what actually moved. This catches capital called at the wrong participation, expenses paid out of the fund that belong to the manager, and recallable distributions treated as ordinary ones.
The balance sheet route depends on accruals, and accruals are estimates that someone has to remember to make. Three recur:
Build the close so that both routes are computed from their own sources, in the same file, and have the file say out loud whether they agree. Not a cell you have to interpret — a stated result. Then make every correction an input you can set to zero, so that the effect of any single adjustment on the final number can be seen by turning it off.
Two routes to the same number is not redundancy. It is the control. It is also the cheapest one available, because both figures already exist in the books; the only work is refusing to accept the first one on its own.
Every figure above is a live formula in the companion files for The Private Equity Fund Controller Playbook. Change one input and the rest of the sheet answers. They are free, and they need no account and no email address.