Direct deals without a deal team: a due diligence standard for family offices buying a single asset

Family offices are buying directly, with small teams

Direct investing is now mainstream among family offices. Citi found 70% of 346 family offices engaged in direct investments (Citi, September 2025). Four in ten of those had increased that activity in the past year. J.P. Morgan found 64% holding direct minority stakes and 49% holding direct majority stakes (J.P. Morgan, 2026). Nearly half, 47%, put more than half of their private allocation into direct positions.

The teams doing this work are small. J.P. Morgan put the average annual operating cost of a family office at just over US$3m, and 40% spend under US$1m. A chief investment officer is present in only 24–31% of offices, depending on the office’s age. Over 90% of North American family offices reported difficulty recruiting (RBC and Campden Wealth, 2025).

Family offices know where the gaps are. BlackRock surveyed 175 single family offices in 2025 (BlackRock, 16 June 2025). Of these, 75% reported internal gaps in private-market analytics, 63% in deal sourcing and 57% in reporting. J.P. Morgan found 52% outsourcing direct investments wholly or in part.

What diligence is worth

The academic evidence supports spending time on diligence. Cumming and Zambelli (2017) studied 178 investments by 27 Italian private-equity firms between 1999 and 2006. More weeks of due diligence was associated with higher three-year return on assets and better operating margins. Diligence carried out mainly by the investor’s own team showed the stronger link. Diligence carried out mainly by outside agents gave what the authors call puzzling results, which they attribute to agency problems.

Wangerin (2019) reached a similar conclusion for corporate acquisitions. Buyers who spent less time between agreement and closing showed lower profitability after the deal and more goodwill write-downs. Competitive pressure was one reason for shorter diligence.

The lesson for a family office is twofold. Time spent testing a deal is not wasted. And outside advisers help most when their interests are tied to the buyer’s, not to the deal closing.

Four tests that put numbers on diligence

Diligence is often treated as a checklist. Four published methods turn it into numbers a family office can act on. The first values diligence itself. The second corrects forecasts. The third measures model risk. The fourth tests the debt.

Test 1. Is the diligence worth its cost?

Howard (1966) showed that a numerical value can be put on reducing any uncertainty before a decision. For a family office deciding whether to invest, the decision is to buy or to walk away. In its standard form, the value of information is:

$$\mathrm{VOI}={\mathbb{E}}_x[max(\mathbb{E}[V|x],0)]-max(\mathbb{E}[V],0)$$

Here V is the net value of the deal to the family, and x is what diligence reveals. The second term is the value of deciding now. The first is the value of deciding after learning x, when the family can still walk away. The value of information is never negative. It is highest when the deal looks close to break-even and the uncertainty is wide. Diligence is worth buying when its expected value exceeds its cost.

Test 2. How wrong are the forecasts likely to be?

Flyvbjerg (2006) reported average construction cost overruns of 44.7% for rail, 33.8% for bridges and tunnels, and 20.4% for roads. Rail passenger forecasts were on average 51.4% too high. Bain (2009) found large errors and optimism bias in the traffic forecasts of more than 100 privately financed toll roads.

Flyvbjerg’s remedy is reference class forecasting. Choose a class of similar past projects, build the distribution of their outcomes, and place the new project within it. For cost, the result is an uplift on the sponsor’s estimate:

$$C_{\mathrm{adj}}=C_{\mathrm{base}}\times (1+u_p),u_p=F^{-1}(p)$$

Here F is the distribution of cost overruns in the reference class, and p is the confidence the investor wants of staying within the adjusted budget. Flyvbjerg (2006) reports the UK guidance that results.

Project type	Uplift for 50% confidence	Uplift for 80% confidence
Roads	15%	32%
Rail	40%	57%
Bridges and tunnels	23%	55%

The 50% level suits an investor with many projects, where overruns and savings offset. A family office buying one asset has no such offset, so the higher confidence level is the prudent choice.

Test 3. How likely is the model to contain an error?

Powell, Baker and Lawson (2009) audited 50 spreadsheets in operational use, with 270,722 formulas between them. Errors that changed a result affected 0.87% of formula cells on average. If each of n formulas were wrong independently with probability e, the chance of at least one error would be:

$$P(\text{at least one error})=1-{(1-e)}^n$$

The average workbook in the study had about 5,400 formulas. With e at 0.87%, this formula gives a probability close to 1. The audit itself found result-changing errors in 86% of workbooks. The gap shows that errors cluster rather than occur independently. Either way, the conclusion is the same: a model that has not been independently reviewed should be assumed to contain errors.

Test 4. How much room does the debt leave?

Project lenders size debt on the debt service coverage ratio (DSCR), the cash flow available for debt service (CFADS) in a period divided by the debt service due. The question for the equity is how often the DSCR will fall below the level at which lenders stop cash reaching shareholders. The model below is our own illustration, not taken from a single published source. If CFADS in period t is lognormal with median F_t and volatility σ, the probability that the DSCR falls below a covenant level k is:

$$P({\mathrm{DSCR}}_t<k)=N\left(\frac{\ln (k{\mathrm{DS}}_t/F_t)}{\sigma }\right)$$

Here DS_t is debt service and N is the standard normal distribution. For example, take a median DSCR of 1.30, a distribution lock-up at 1.10 and σ of 15%. The chance of a lock-up in any one period is then about 13%. In a lock-up, cash stays in the company instead of reaching the family. The equity case depends as much on this probability as on the base-case return.

Framework: the five-gate diligence standard

A single asset should pass five gates before the family signs. Each gate has a question it must answer, and each can stop the deal.

Gate 1. Why is this asset for sale?

The first question is about the seller, not the asset. Fang, Ivashina and Lerner (2015) found that investors’ own deals outperformed co-investments offered by sponsors. The advantage shrank where information was hard to obtain. A deal offered to a family office may be one that others passed over. The family needs to know who else has seen it, and why the seller is leaving.

Gate 2. Are the forecasts credible?

Revenue and cost forecasts drive the price. The test is to compare them with a reference class of similar assets, not with the sponsor’s own history. Traffic, charter rates, power prices and occupancy all need an outside view. Where no reference class exists, the forecast should be stress-tested against the weakest comparable outcome.

Gate 3. Do the contracts deliver the cash?

For a real asset, the contracts are the asset. Charters, offtake agreements, concessions, leases, construction and operating contracts set who pays, for how long and under what conditions. Lenders take security over these contracts, the asset and its insurance. The investor needs the same clarity: termination rights, indexation, counterparty strength and what happens if a contract ends early.

Gate 4. Is the model right?

The financial model turns forecasts and contracts into value. Models are error-prone. Powell, Baker and Lawson (2009) audited 50 spreadsheets in operational use and found errors in 94% of them. In 86%, at least one error changed a result. A family office should not rely on a model it cannot question line by line. An independent review of the model should come before any valuation is accepted.

Gate 5. Does the financing hold?

Most single assets carry project debt. The debt terms set how much cash reaches the equity, and when. Lenders run their own diligence, which gives the family a second view. Their covenants also show where the lenders see risk. The family should read the term sheet as closely as the asset.

What’s the bottom line?

Family offices are already direct investors, but most do not have deal teams. The evidence suggests that diligence effort pays, and that diligence led by the buyer’s side works better than diligence the buyer merely receives. Forecasts and models deserve the closest scrutiny.

A written standard turns good intentions into a repeatable process. Five gates cover the seller’s motive, the forecasts, the contracts, the model and the financing. A family office that cannot staff all five can buy the missing skills, as long as the advisers work for the family.

Summary

Surveys from 2025 and 2026 show most family offices investing directly, often with small teams and acknowledged gaps in analytics and deal sourcing. Research links more diligence effort to better results after the deal, and finds the link strongest when the buyer’s own side leads the work. For a single real asset, two risks stand out. Infrastructure cost and demand forecasts are often wrong, and most operational spreadsheets contain errors. A five-gate standard tests the seller’s motive, the forecasts, the contracts, the model and the financing. Four published tests put numbers on the process: the value of the information, the forecast uplift, the chance of a model error and the headroom in the debt.

This Insight is for general information only and does not constitute investment, legal or tax advice. Figures from public sources are as reported at the dates shown.