An expected-cost comparison of review-only oversight, a targeted shadow of the calculations, and a full duplicate set of books.
Only for the calculations it can actually catch. On illustrative error rates, a manager relying on review alone carries 426,000 a year of expected error cost. A targeted shadow of allocations, fees and carry costs 190,000 and avoids 196,000 of that, roughly paying for itself; a full shadow costs 470,000 more to avoid another 82,000, or 0.17 for every unit spent. Most of what is left is valuation error, which no second ledger catches.
Worked in full in The Private Markets COO by Julian R. Sterling, with every figure reproduced in a free workbook.See the book on Amazon →
Shadow accounting means the manager keeps its own record of what the fund administrator calculates and reconciles the two. Investors and auditors like to hear that it exists; operations budgets dislike what it costs. The way to settle the argument is to treat each design as insurance: what it costs a year, against the expected cost of the errors it stops before they reach investors. That needs three inputs per error type: how often it occurs, what it costs if it gets out, and how likely each design is to catch it.
| Error type | Events a year | Cost if it reaches investors | Expected cost |
|---|---|---|---|
| Capital account allocations | 2.0 | 60,000 | 120,000 |
| Fee and carry calculations | 0.5 | 400,000 | 200,000 |
| Valuation inputs | 0.3 | 900,000 | 270,000 |
| Cash and payments | 0.2 | 300,000 | 60,000 |
| Total | 650,000 |
The cost of an error is the compensation, the reissued statements, the extra audit work and the staff time, not the error's size. How to put a number on one is worked in how to calculate the cost of a NAV error.
| Design | Annual cost | Allocations | Fees and carry | Valuation | Cash |
|---|---|---|---|---|---|
| Review only (analytical review of the NAV pack) | 0 | 30% | 25% | 40% | 50% |
| Targeted shadow (1 FTE + 20,000 of tools) | 190,000 | 85% | 90% | 40% | 50% |
| Full shadow (3 FTE + 150,000 system) | 660,000 | 95% | 95% | 60% | 60% |
A fully loaded FTE costs 170,000. The review-only design is not free in reality, but its cost is the oversight team every outsourced manager already has, so it is the baseline.
Residual loss = Σ events × cost × (1 − detection rate)
Total cost of the design = annual cost + residual loss
In Excel, with events in B2:B5, cost in C2:C5 and detection in D2:D5: =SUMPRODUCT(B2:B5,C2:C5,1-D2:D5)
| Design | Allocations | Fees and carry | Valuation | Cash | Residual | Design cost | Total |
|---|---|---|---|---|---|---|---|
| Review only | 84,000 | 150,000 | 162,000 | 30,000 | 426,000 | 0 | 426,000 |
| Targeted shadow | 18,000 | 20,000 | 162,000 | 30,000 | 230,000 | 190,000 | 420,000 |
| Full shadow | 6,000 | 10,000 | 108,000 | 24,000 | 148,000 | 660,000 | 808,000 |
The targeted shadow avoids 196,000 of expected loss for 190,000, a return of 1.03 per unit spent. It is the cheapest design overall, by 6,000, which is to say it roughly pays for itself on these rates and buys a great deal of comfort for nothing. Its work is concentrated where the administrator's errors cluster: side-letter waivers, excused investors, fee offsets and the carry calculation, the items in how a side-letter fee waiver is applied, where an invoice can be right and an investor's allocation wrong.
The full shadow is a different proposition. Going from targeted to full adds 470,000 of cost to avoid another 82,000 of loss, 0.17 per unit spent. In total it costs 382,000 a year more than doing nothing beyond review.
Under the targeted shadow, 162,000 of the 230,000 residual, 70.4 per cent, is valuation error. A second ledger recomputes the arithmetic from the same inputs, so a wrong multiple or a stale comparable flows through both sets of books. That money belongs in valuation governance, not in more reconciliation.
| Error frequency | Review only | Targeted shadow | Full shadow | Cheapest |
|---|---|---|---|---|
| 0.5x | 213,000 | 305,000 | 734,000 | Review only |
| 1.0x | 426,000 | 420,000 | 808,000 | Targeted shadow |
| 2.0x | 852,000 | 650,000 | 956,000 | Targeted shadow |
| 3.0x | 1,278,000 | 880,000 | 1,104,000 | Targeted shadow |
The targeted shadow breaks even against review at 0.97 times the base error rate and wins by more the worse the administrator is. The full shadow only beats it at 5.73 times the base rate. An administrator making errors that often is a vendor problem, and the right response is the escalation route in the administrator scorecard, not a permanent duplicate of its work.
The control library and incident log that feed rates like these are templates in the free workbooks for this book.
The manager keeps its own independent record of some or all of what the fund administrator calculates, and reconciles the two. A full shadow duplicates the general ledger; a targeted shadow recomputes only the items most likely to be wrong, such as capital account allocations, fees and carried interest. Here the targeted version costs 190,000 a year and the full one 660,000.
Only partly. A shadow ledger recomputes arithmetic from the same inputs, so a wrong multiple or a stale comparable flows through both sets of books. In this case valuation inputs carry 270,000 of the 650,000 of gross expected error cost and remain 70.4 per cent of the residual under a targeted shadow. Those are caught by valuation governance, not by a second ledger.
When errors are much more frequent than the base case. A full shadow adds 470,000 of cost to avoid 82,000 of expected loss over a targeted shadow, so error rates would have to be 5.73 times higher before it pays. That is a reason to change administrator, not to duplicate its work permanently.
This article is one calculation from The Private Markets COO. The book takes the same case from first principles to the decision, chapter by chapter, and every figure it prints is a live formula in the free companion workbooks.
Get the book on Amazon →Free companion files
Also on Amazon UK · Amazon Germany · Amazon France · Amazon Canada
Reading guide: private equity and private markets → · All 1340 articles →
If this book helped, or didn’t, a few lines on Amazon are worth more than they look: they are what the next reader goes on. Write a review. The workbook stays free either way.