Articles

Why refining an operational due diligence rubric makes it weaker

Averaging cannot express “this issue is disqualifying”. Adding domains halves the weight of the one that could kill the fund, and re-weighting changes nothing at all.

Julian R. Sterling · from Operational Due Diligence in Private Equity

Twelve managers are scored across eight operational risk domains and the scores are averaged. Three of the twelve carry a flaw that is terminal on its own — no independent custody, an unmanageable conflict, self-marked valuations with no oversight. The averaging model gives all three its top rating.

Worse than that. The manager with no independent custody of client assets scores 1.500. A manager with no fatal flaw at all scores 2.250. The model does not merely fail to catch the disqualifying issue — it ranks it 0.75 of a point higher.

Every operational due diligence framework carries a warning against averaging your way to comfort. The warning is correct and almost always unquantified, which is a strange thing for a warning about arithmetic. Here is what it is worth in decisions.

The model

Eight domains, scored 1 (low concern) to 5 (high). Approve at 2.0 or below, approve with conditions at 3.0, not yet at 3.5, reject above.

Risk domain	Can it be terminal on its own?
Governance	—
Structure & terms	—
Finance & controls	—
Admin, custody, cash	yes
Service providers	—
Compliance & conflicts	yes
Technology & resilience	—
Valuation	yes

Three of the eight domains can be terminal on their own. That is not a controversial claim — most frameworks name the same three. The question is whether the scoring model does anything with the fact, and an average does not: it has no way to express “this one is disqualifying regardless of the rest”.

Twelve managers, two decision rules

Twelve managers, one scoring sheet, two decision rules.

Manager	Worst veto domain	Average	Decision on the average	Decision with veto rules
Alderman Credit	5	1.500	Approve	Reject (veto)
Brightmoor Partners	1	1.250	Approve	Approve
Caldwell Special Sits	2	2.250	Approve with conditions	Approve with conditions
Draycott Infrastructure	5	1.625	Approve	Reject (veto)
Ellerby Growth	3	3.000	Approve with conditions	Approve with conditions
Fenwick Real Assets	5	1.500	Approve	Reject (veto)
Garrow Mid-Market	3	2.250	Approve with conditions	Approve with conditions
Halstead Ventures	3	3.500	Not yet	Not yet
Inverleith Secondaries	1	1.000	Approve	Approve
Jarrow Opportunistic	4	3.625	Reject	Not yet (veto)
Kelsall Private Credit	2	2.125	Approve with conditions	Approve with conditions
Larkhall Buyout	4	2.250	Approve with conditions	Not yet (veto)

Five decisions out of twelve change when the same scores are read with veto rules instead of an average. Three of those five are managers the average approves and the veto rules reject outright. Nothing about the underlying diligence changed; only the arithmetic did.

The comparison that should end the argument

One has no independent custody of client assets. It scores three quarters of a point better.

	Alderman Credit	Caldwell Special Sits
Admin, custody, cash	5	2
Every other domain	1	2 or 3
Average	1.500	2.250
Decision on the average	Approve	Approve
Which one has a terminal flaw?	this one	neither

How good does the rest have to be?

Solve it directly. With n domains, one scored 5 and an approval threshold of 2.0, the others may average up to (2n − 5)/(n − 1) and the fatal flaw still disappears.

A single domain scored 5, and the overall score still lands on Approve.

Domains in the rubric	Average the others may carry	Weight of the fatal domain
5	1.250	20.0%
8	1.571	12.5%
10	1.667	10.0%
12	1.727	8.3%
16	1.800	6.2%
20	1.842	5.0%

Read that table twice. At eight domains the other seven need to average 1.571 — not perfection, just good. At sixteen domains they need only 1.800. Refining the rubric made it easier to hide a fatal flaw, by 0.229 of a point, because splitting eight domains into sixteen halves the weight of the one domain that could kill the fund.

Nobody running that exercise would describe it that way. A team that adds granularity is trying to capture nuance and believes it is tightening the model. It is doing the opposite, and the direction of the error is the dangerous one: the more careful the framework looks, the more comfortably a disqualifying issue rides through it.

And the debate committees actually have does not change anything

Weightings are where the argument goes. Should custody carry three times governance? Should valuation be doubled? Two plausible schemes, applied to the same twelve managers.

Zero from re-weighting. Five from restructuring.

Change made to the model	Decisions it changes, out of 12
Re-weight: controls-heavy	0
Re-weight: governance-heavy	0
Replace averaging with veto rules	5

Zero decisions out of twelve change under either re-weighting. Five change when averaging is replaced by veto rules. Committee time spent debating weights is time that cannot change an outcome — while the design decision that determines every outcome, whether the model may say “this issue is disqualifying”, is usually taken in one sentence and never revisited.

How close is each answer to a different one?

One more diagnostic worth running on your own rubric. Measure each manager’s distance to the nearest threshold in analyst-points — one notch of disagreement on one domain. It tells you which conclusions survive a reasonable difference of opinion between two people who did the same work.

One analyst-point is one notch of disagreement on one domain.

Manager	Average	Nearest threshold	Analyst-points to a different answer
Alderman Credit	1.500	2.0	4
Brightmoor Partners	1.250	2.0	6
Caldwell Special Sits	2.250	2.0	2
Draycott Infrastructure	1.625	2.0	3
Ellerby Growth	3.000	3.0	1
Fenwick Real Assets	1.500	2.0	4
Garrow Mid-Market	2.250	2.0	2
Halstead Ventures	3.500	3.5	1

A rating one analyst-point from a different answer is not a rating. It is a coin flip with a decimal place, and it should be reported as such to whoever relies on it.

What to change on Monday

Three things, none of which requires rebuilding anything. First, name the domains that are terminal on their own and give the model a rule that says so — a veto, not a weight. Second, before adding domains to a rubric, compute what the addition does to the weight of the terminal ones; if the answer is that it dilutes them, add the veto instead of the nuance. Third, publish the analyst-point distance next to every rating, so a committee can see which of its conclusions are robust and which are arithmetic.

The warning against averaging your way to comfort is right. It just needs a number attached before anyone acts on it, and the number is five decisions in twelve.

Also on this site

• Can a fund call more than the commitment I signed?151,970,000 called against a 150,000,000 commitment: the recycling clause, and the 1,970,000 nobody signed for.
• Should a retained 1250 per cent position be deducted or risk-weighted?Risk-weighting a retained 1250 per cent position costs 168.75 per cent of face at a 13.50 per cent target; deduction costs 100.
• How do you recompute a private fund's management fee after offsets?Gross fee 4,312.50, offsets 2,175, net 2,137.50. Recomputing a private fund's management fee quarter by quarter.
• Can a 30 per cent marketability discount be evidenced?The protective-put method peaks at 28.6 per cent and cannot reach a conventional 30 per cent discount at any holding period.
• What does a half-price move-in offer actually cost a self-storage store?The £141 discount buys a customer worth £2,188 — and is only 52.7 per cent of what a move-in really costs.
• Why do gross flows keep rising while the growth rate falls?The growth rate peaks at 48.1 per cent in month 29; gross flows are still climbing at month 84. Both numbers are right.
• Can a REIT trading at a discount to NAV still acquire accretively?REIT acquisitions turn dilutive at a 6.2 per cent discount to NAV, not at zero — 11.6 per cent funded at the target mix.
• Does a sale and leaseback pay more than the tenant's bonds?The over-rent implies 269 basis points of credit spread, against 385 on the tenant's own bonds.
• How much of a credit fund's diversification benefit survives a crisis?A four-sleeve credit fund's diversification is worth 0.767 of a point; only 27.6 per cent of it survives a crisis.
• Is a combined downside worse than the sum of the individual moves?Two downside moves together cost 5,906,035, not the 6,220,557 they sum to: the cross term is 314,523.
• What multiple is an investor really paying at 1.80x cover and a 6.0 per cent yield?At 1.80x cover and a 6.0 per cent yield the buyer pays 9.26x EBITDARM, not 11.00x — £2,661,227 apart.
• What does an 18 per cent discount to net asset value actually buy?An 18 per cent discount to reported net asset value is worth 9.18 per cent once three computable deductions are made.
• What does losing the anchor actually cost a shopping centre?An anchor paying €561 600 exposes €1 985 894 of rent when it leaves — 3.5 times its own rent.
• Why does the cheaper refurbishment package lose money at the exit?The plant package saves 1,400,000 on the contract and loses 501,000 at the exit, where the buyer prices its remaining life.
• Does a first close cover the management company's costs?A 100,000,000 first close at 1.5 per cent pays 125,000 a month against a 200,000 burn: the house loses 75,000 a month.
• Does higher leverage dilute the operating partner's contribution?Organic operating improvement contributes 59.85 million whether the deal is levered 3.5 times or 6.5 times.
• Does hedging monthly instead of daily lose money?Monthly rebalancing leaves a standard deviation of 2.3031, 20.1 per cent of premium; daily leaves 0.5265.
• How do you compute a capital-adjusted yield on an office?Quoted at 6.00 per cent, this office returns 3.20 per cent once the re-letting cycle is deducted: €160.04/m² a year.
• What does a university nomination agreement actually cost?A nomination quoted at 8.8 per cent off the rate is really 21.2 per cent: 6 unpaid weeks, £153,568 a year.
• What does leaving a credit score cut-off unchanged actually cost?The optimum cut-off moved from 550 points to 580; the policy did not. One year of that inertia cost 2,713,298.
• Can you average fund IRRs to get a portfolio return?Three funds at 11.52, 10.12 and 19.51 per cent average to 13.72 per cent. Pooled, they earned 11.78 per cent.
• Why is our money-weighted return lower than the manager's time-weighted return?Time-weighted 4.8012, money-weighted 4.1272: flow timing cost the scheme 6,357,745.18 and appears in no report.
• Does a constant WACC overstate value when the debt amortises?A constant WACC against amortising debt values this concession at 263.05, not 242.18 — an overstatement of 8.6 per cent.
• How much more margin does a five-year hedge call than a rolling one?A locked five-year hedge calls 4.48 times the margin of a rolling one and exhausts 36,000,000 of liquidity at 7.09 per cent.
• Does the over-rent in a sale-and-leaseback repay the over-price?Twelve years of 520,000 over-rent comes to 6,240,000 against an over-price of 7,619,048. The buyer is never repaid.
• How much more rent will an occupier pay to cut its labour cost by five per cent?A five per cent labour saving is worth 32 per square metre of rent, a 36 per cent premium. Property is 10.2 per cent of cost.
• Does a 20 per cent promote actually pay the GP 20 per cent?A 20 per cent promote paid $17,200,000 on $70,000,000 of profit — 24.6 per cent. The definition, not the rate, decides.
• What would a buyer actually pay for a bank reporting 11.40 per cent capital?A bank showing 639,600 of CET1 capital is worth 278,809 to a buyer, once the loan book is marked.
• What does a missed capital call actually cost a sponsor?The same 1.5 times dilution leaves a defaulting sponsor on 7.04 per cent or 2.99 per cent, depending on the formula.
• What does it cost when an insurer cannot look through a fund?No look-through pack means 49.00 per cent, not 18.90 — €45.2 million of capital on a €150.0 million allocation.
• Do uncapitalised leases count as debt in a leverage covenant?Treating 4,150,000 of operating leases as debt moves year-one accretion 4.1351 points. Both legs of that treatment are wrong.
• Does a twelve-month energisation delay cost more than an eight per cent rent miss?A twelve-month energisation slip costs 399 basis points of levered return; an eight per cent rent miss costs 257.
• Does more gearing always increase the manager's promote?The promote peaks at 2,993,873 at 65 per cent gearing and falls to 2,876,129 at 70, while the equity return still rises.
• How many expense allocations do you have to test?59 clean allocations prove an error rate below five per cent; twenty prove only fourteen.
• What does a pre-let rent discount actually cost?16,000 a year of rent conceded shows as 145,727 in the appraisal and costs 278,261 in the price — 1.91 times.
• What does an ESG value creation plan actually add to enterprise value?€1.1 million of delivered EBITDA at 8.5× is €9.4 million of enterprise value, against €830,000 of cost.
• What is a 50 basis point MFN actually worth to an existing lender?A 50 basis point MFN steps the margin from E + 550 to E + 700 and pays existing lenders €2.06 million a year.
• Why is realised loss severity higher than appraisal severity?Marlowe Tower sold at a 25.90 per cent shortfall; the trust wrote off 37.24 per cent. Where the 11.34 points go.
• Does an additional tier 1 shortfall reduce headroom to the distribution trigger?A 40.8 million additional tier 1 shortfall takes 13.6 basis points out of the buffer: headroom is 2.8432 per cent, not 2.9796.
• How much of an accelerated placement's cost never appears on an invoice?The commission on a EUR 385,225,000 placement was 13.50 per cent of what it cost. The other 86.50 per cent was uninvoiced.
• How much of a private fund's reported IRR reaches the investor?A net IRR of 13.2935 per cent, a sleeve return of 6.2571, and a 2.9307-point treasury decision nobody minuted.
• How do you compute the reverse stress level for an open-ended credit fund?A semi-liquid credit fund's reverse stress level, computed: 17.2 per cent of net asset value against a 4 per cent gate.
• How do you set an advance rate from a loss curve?Base loss 2.4 per cent, downside 6.0 and transition 1.5: the advance rate lands at 88 per cent and 12 per cent enhancement.
• Why do a DCF and a transaction multiple disagree by 26.8 per cent?Three methods, 18,480,000 apart on one company. Back each into the model and the entire gap is 1.96 points of growth.
• How much of an earn-out can the buyer's accounting adjustments remove?€1.30 million a year of ordinary accounting entries removes €6.60 million of a €12.0 million earn-out.
• What does it cost to roll on worse terms than the incoming buyers?Blind-pool terms cost a rolling investor about €400,000 on a €5 million position — eight points of net multiple.
• What is an EBITDA add-back bridge actually worth in leverage turns?The add-back bridge is worth 1.6866 turns of leverage; the 4.25x covenant leaves 0.0336 of a turn. A ratio of 50.3.
• What should a monthly treasury report contain?Nineteen lines, each with the covenant or limit it is measured against — and 27,403,718.84 without the revolver on every one of them.
• Where should a company park its surplus cash?361,824.00 between the best and worst yield, and 93,204,281.16 of the surplus already spoken for.
• Did my currency hedge fail if it lost money?A loss of -8,130,844.04 against spot, and the hedge did exactly what it was for — the break-even is the forward, not the spot.
• Should corporate debt be fixed or floating?78.1818 per cent of gross debt fixed, and a net floating exposure after cash of -608,000.00 — the proportion, not the instrument.
• Does an undrawn revolver count as liquidity?67.4567 per cent of the headroom, and it leaves at a fall in annual revenue of 8.4127 per cent — not the 13.3333 per cent the board deck implies.
• Does supply chain finance count as debt?66,049,315 released, and leverage of 2.7622× or 3.1362× depending on which balance the test is asked of — the second is a breach at today's EBITDA.
• How much of a company's cash is actually available?One balance sheet, five correct answers, 214,000,000 down to 68,400,000 — and the definition alone is worth 0.3690 turns of leverage.
• How do you set a minimum cash balance?Three components, 124,596,281 in total — and the double count that turns a margin of 27,403,719 into an apparent shortfall of -3,988,281.
• Is a 2/10 net 60 discount worth taking?An annualised 14.8980 per cent against a revolver at 4.50 per cent — worth 5,471,890 a year, and 0.3132 turns of reported leverage.
• At what wealth does a family office pay?A five-person office costs the same $2,442,000 at $100m as at $2bn. Against a tiered multi-family schedule the two curves cross exactly once.
• Does the fulcrum move with enterprise value?A worked capital structure at three enterprise values. The break migrates from the first lien to the subordinated notes without a single document changing.
• European vs American waterfall, and the clawbackThe same deals, the same total profit, two calculation orders — and a $9M promise to repay that is only as good as the entity that made it.
• How many companies does a reserve pool defend?Holding pro rata to the Series D costs $14.76m per company. A $40m reserve defends 2.71 of forty — and 61 per cent of the cost is the last round.
• How much room a 1.15x downside actually leaves79 basis points of margin. The leverage covenant fails first, five basis points before the coverage covenant the chapter reports on, and 4.6 per cent of EBITDA is the whole cushion.
• How the GP catch-up is actually solvedMost write-ups state the catch-up formula then hardcode the answer. Set it as a solve and it moves on its own when the carry rate does.
• Margin says one thing, return on capital says anotherA 350 basis point margin advantage becomes 3.4 points of return on capital — and vanishes if the default estimate is 1.45 times out.
• Six ways to state the same fund's return7.12 per cent, 18.04 per cent, or 1.323 times the public market — one fund, one set of flows. The eleven-point gap is the valuation, and it is not cash.
• Why prepaying to cure a covenant costs eleven times a deposit233,192 against 2,602,535 on the same breach. The ratio is 1/(k·c) — it ignores the size of the decline and turns on two terms the borrower negotiates.
• The IRR at which a closed-ended fund merely tiesA drawdown fund advertising 13.25 per cent ties an evergreen netting 8.64. On committed rather than called capital, 78.7 per cent of the edge vanishes.
• The month a remediation trigger has to be armed byMonth 8, on a one-year window and a 14-week hiring lead time. The date is 12W − L/4.33 and holds for any portfolio size — the book only changes the euros.
• Two routes to the same NAVFive findings in one quarter-end close, and not one of them was a modelling error. Every one was a control that did not exist.
• What a CBAM certificate buffer costs to holdCarry and protection are both linear in the buffer, so the break-even probability is identical at 5 per cent and at 30 — and 2027 costs exactly twice 2026.
• What a clawback actually collateralises24 million recovers 13.20 net of tax and 7.20 is escrowed. The rule is e ≥ 1 − t, and every euro won on the tax clause lands uncollateralised.
• What a co-investment programme actually saves37 basis points, and a 22.8 per cent total-loss rate erases it. At the usual per-deal cap a single failure is 37 per cent of the annual cohort.
• What a diligence exercise actually recoversEighteen basis points is only the part that reaches the yield. Three of the four instruments never do, and the exercise is worth thirty-one.
• What a follow-on reserve is actually worth0.019 turns under plain dilution, 0.285 under pay-to-play. The optimum is a property of the shareholders’ agreement, not of the portfolio.
• What a lease mark-to-market is actually worth5,788,000 becomes 807,365 once the over-market stream, the expiry dates and the cost of re-letting are priced — and nothing at all below a 43.7 per cent renewal rate.
• What a minimum payment actually does£81.53 leaves the account and £20.46 reaches the debt. On minimums alone the set clears in 258 months and repays 1.78 times what was borrowed.
• What a month of delay costs in an enforcement463,000 euros of present value a month. No single error about enforcement, and no pair of errors, moves the indifference point from 58.7 cents to the 72 on the table.
• What a pro rata cheque actually costsDefending costs 1.25 times the initial cheque across two rounds, not more in each — and below a 521 exit the follow-on dollars come back worth less than they cost.
• What a subscription line does to the IRR20 per cent becomes 38.41 on the same asset while the multiple falls. Two dollars of interest buys 18.4 points, and the longer the line runs the cheaper each point gets.
• What an 80-cent recovery is actually worthComposition and time do more damage to a recovery than the headline number suggests. Two arithmetic steps most screens skip.
• What one point of fees actually costsOne point on the annual charge turns 738,846 into 567,961 over forty years — 23 per cent of the pot, and 2.43 of wealth lost per 1 of fee.
• When the promote stops clearingCarried interest is not proportional to performance near the hurdle. It is a step function, and $2m of headroom is what stands between a promote and none.
• Why a secondaries bid-ask gap has no zoneThe seller’s floor and the buyer’s ceiling are one formula at two rates. At equal required returns the zone is exactly zero — here it is 11.02 points apart.

This note is drawn from Operational Due Diligence in Private Equity. The book is on Amazon.

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.