Capital Stack Diagnostics: A Framework for Asset Allocators
- Gaurav Shah
- Aug 8, 2025
- 9 min read
Updated: 3 days ago
A GP once walked us through a 41% IRR.
The arithmetic held and was also largely a function of a NAV credit line (once you traced it) that had deferred capital calls long enough to squeeze the denominator, sitting on top of exits that were paper mark-ups with no cash behind them. Nothing in the deck was false and perhaps that was the problem.
The fund was being measured by an instrument its own capital structure had been built to move.
This is about fund-level capital architecture, not the debt and equity layering of a property deal. How a private fund forms, sequences, deploys and returns capital, and what that sequence tells an allocator that the performance summary won't.
Why it matters more now than five years ago comes down to arithmetic. Bain's Global Private Equity Report 2026, running on data through Q3 2025, counts 32,000 unsold portfolio companies worth $3.8 trillion. The industry's distribution rate on a NAV basis has sat below 15% for four consecutive years, which is a record and not one anyone wanted. When real liquidity gets that scarce, manufacturing the appearance of it stops being occasional and becomes structural.
I've written more on where these structures break down in Why Most Capital Stack Designs Fail, and How Operators Fix Them for Connectively.
Why Traditional Allocation Models Are No Longer Enough
The New Complexity in Fund Structures
Continuation vehicles, NAV loans, staged closings, sidecar co-investments, deferred capital calls, back-levered exits. All of it is now normal.
The scale is easy to underestimate. Jefferies put full-year 2025 global secondary volume at $240 billion, up 48% and the biggest year on record, with GP-led transactions accounting for $115 billion. Single-asset continuation vehicles crossed half of all CV volume for the first time.
None of this is bad in itself. It reflects a maturing market, and some of it genuinely helps LPs who want liquidity. What it does is make fund comparison harder in a way most allocation models haven't caught up with. Two funds showing the same TVPI can have arrived there through completely different capital pathways carrying completely different risks. Ranking them on that number alone is intellectually dishonest.
Growing Misalignment Between Fund Metrics and Real Risk Exposure
In theory, DPI is cash returned to LPs but in practice we've watched it inflated through sales to affiliated vehicles, recycled distributions, and back-levered realizations. Each one produces a number that's technically accurate and economically misleading.
Where this goes next is worth watching. Rede Partners' NAV Financing Market Report 2026 found that most NAV loans currently fund additional investment rather than liquidity, but that lenders expect use of these facilities to generate DPI to expand over the coming year. Borrowing against a portfolio to pay a distribution shifts timing and adds a cost. Neither of those is a return, and the cost lands on the LP.
So the question stopped being what your DPI is. It's how you built it.
The Capital Stack as a Diagnostic Lens
Reconstructing Where and How Capital Is Deployed
Start by rebuilding the anatomy. Where did GP capital actually go? Was LP capital warehoused before the first call? Were early closes staged to backsolve IRR?
Tracking the timing of cash flows, particularly around calls, distributions, recycling and reserve allocation, gives you a granular view of real capital efficiency. It lets you separate returns that came from operating value creation from returns that came from financial choreography. Those two look identical on a summary page and behave nothing alike in a downturn.
Reading the Hidden Signals in the Stack
A fund's capital pattern tells you about the people running it.
A GP who consistently delays capital calls until after value inflections is usually running a disciplined book. A manager who back-levers assets to hit a DPI target right before the next raise is optimizing for a fundraising narrative. Neither shows up in a pitch deck. Both are visible in the cash flow record, if you ask for it at the right granularity and are willing to read it.
Red Flags and Underwriting Traps
Structurally Engineered IRRs
IRR is the most gamed metric in private markets and subscription lines are the standard instrument. Defer the calls, compress the denominator, lift the reported number. Underlying performance doesn't move at all.
There's a second version of this that gets less attention: staging the final close after a significant markup event. It creates a real asymmetry between early and late LPs while improving net IRR optics for the manager. Legal, disclosed somewhere in the fine print, and almost never raised in an IC meeting.
Optical DPI vs. Economic DPI
We looked at a secondary transaction where a partial asset sale produced a DPI of 1.1x while roughly 80% of capital was still tied up in long-dated positions. On the cover page, a fund returning capital. Underneath, a thin realized multiple and remaining assets facing valuation compression.
The distinction is worth making explicit:
Optical DPI | Economic DPI | |
Source of cash | Affiliated vehicle, continuation fund, or borrowing | Arms-length sale to an unaffiliated buyer |
What remains | Same assets, different wrapper | Exposure genuinely reduced |
Cost to LP | Fees and interest, often both | Transaction costs only |
Repeatable? | No. It borrows from future distributions | Yes, if the strategy works |
What it proves | The GP can access liquidity markets | The GP can sell an asset to someone with no reason to buy it |
That last row is the one to sit with. Re-underwrite DPI on permanence, liquidity, and whether the exit was arms-length or engineered.
How Should Allocators Evaluate a Fund's Return Construction?
Frameworks are easy to produce and hard to use. What actually changes an IC discussion is a short list of questions where you already know what a good answer sounds like. These are the five we bring.
"What was your subscription line balance at each quarter end for the last three years, and what's your IRR with and without it?"
A good answer arrives with the number already run, because the GP has done it themselves. A bad answer is a promise to come back to you. Managers who've never computed their own unlevered IRR are telling you something they didn't intend to.
"Of your realized DPI, how much came from sales to unaffiliated third parties?"
Good answer: a percentage, immediately. Bad answer: a pivot toward total value.
"Which companies have you held longer than five years, and what specific event gets each one out?"
Bain puts 39% of portfolio companies past the five-year mark, up from 29% in 2019, with average hold at exit sitting near seven years. A good answer names the event and roughly when. "Market conditions" isn't an answer.
"On your continuation vehicle, who set the price and who else bid?"
With single-asset CVs now more than half of GP-led volume and average deal size near $900 million, this is standard equipment rather than an exception. A good answer describes a competitive process with named participants. A fairness opinion on its own documents a process. It doesn't create one.
"How much of your reserve capital went into bottom-quartile companies?"
Good answer: they know, because they track it. Bad answer: they don't track it that way. Reserve discipline is where judgment and sunk-cost bias part company, and it rarely surfaces in reporting. The same failure mode turns up in frontier tech diligence, where science teams keep funding the experiment that nearly worked.
None of this needs a proprietary model. It needs someone in the room who has sat on the other side of the table and can hear when an answer is evasive.
Operator-Led Diagnostics for Investment Committees
Moving Beyond Scorecards
Conventional diligence stops at benchmarking TVPI, IRR and vintage peer sets. That gives you a ranking, not an understanding.
Our diagnostic work maps three things. Capital cadence, meaning how capital moves relative to when value actually gets created. Portfolio construction discipline, covering sector stacking, stage drift and check sizing. And execution deltas, the gaps between what the strategy says and what operationally follows.
Here's the difference between a consultant's view and an operator's. A benchmark tells an IC where a fund sits. A diagnostic tells them why it sits there and what would have to change. We went into that distinction properly in operator-led versus consultant-led fund strategy.
Institutional Capital’s Edge in a Compressed Exit Market
Exit optionality has narrowed. Funds are taking longer to distribute, and Bain's data shows every vintage from 2017 to 2021 underperforming historical DPI benchmarks.
Allocators who can read capital construction in that environment can spot managers with durable operating capability, value creation that's realized rather than marked, and honest GP-LP alignment on exit timing. Done properly this is real-time underwriting, and the window for it is before the commitment, not at the re-up.
From Diligence to Deployment: Building the Right Playbook
Embedding Capital Analysis into the Allocation Process
Sophisticated LPs run this at three points. Pre-commitment, to vet capital strategy, pacing and return construction. At re-up, to work out whether performance came from luck, leverage or execution. And in secondaries, reverse-engineering cash flows to establish true remaining value against a market pricing LP portfolios at 87% of NAV.
The Future: Allocator-Led Pressure on Fund Transparency
Quarterly PDFs are quietly losing their status as adequate reporting. LPs increasingly want real-time cap table visibility, tranche-level cash flow data, and stress tests across scenarios.
Dissecting a fund's capital construction is moving from premium capability toward baseline expectation. GPs who resist that shift are making a selection statement about which LPs they want, whether they mean to or not.
Case Study: When Optics Masked the Truth
In 2024 a mid-sized family office asked us to assess a GP's flagship fund. The headline numbers were strong: 2.3x TVPI, 1.6x DPI, IRR north of 30%.
A capital construction review turned up three things. Around 40% of DPI came from a GP-led secondary into a continuation vehicle the manager also controlled. NAV was supported in part by inside-round markups. Reserves had gone heavily into underperforming companies in sectors adjacent to, rather than inside, the fund's stated thesis.
The family office passed. Over the following year the portfolio took markdowns and the next fund came to market considerably later than the GP had signalled.
The analysis avoided a bad outcome rather than producing a good one. That's harder to put in a track record and worth more than most of what goes in one.
Where diagnostic work needs to turn into structural change, our Private Equity & Venture Capital Advisory works with GPs and LPs on fund architecture, alignment models, and the mechanics of getting capital back.
Key Terms, Defined
DPI (distributions to paid-in). Cash actually returned to LPs divided by capital called. The closest private markets get to a cash-on-cash measure, which is exactly why it's worth checking how it was produced.
TVPI (total value to paid-in). DPI plus the current marked value of whatever remains. Because that second half is an estimate made by the manager being measured, TVPI is a claim rather than a result.
Capital call. The notice requiring LPs to send committed money. Call timing feeds directly into reported IRR, which makes call patterns a diagnostic in their own right.
Subscription line. Short-term borrowing that lets a fund invest before calling LP capital. Useful operationally. It also raises reported IRR without improving anything about the investment.
NAV loan. Borrowing secured against the fund's portfolio rather than against LP commitments. When the proceeds fund a distribution, the LP is receiving their own future returns early, minus interest.
Continuation vehicle. A new fund, usually raised by the same GP, that buys assets from the existing fund. Creates liquidity for LPs who want out, extends the hold for those who stay. Price discovery is the entire question.
Frequently Asked Questions
How should allocators evaluate a fund's reported IRR?
Ask for IRR calculated with and without subscription line usage, plus quarter-end facility balances going back three years. The gap between those two numbers tells you how much of the reported return is financing rather than performance. Managers who can't produce it quickly haven't looked.
How can an LP tell whether DPI is real or engineered?
Trace each distribution back to its source. Cash from an arms-length sale to an unaffiliated buyer is economic DPI. Cash from an affiliated vehicle, a continuation fund the same GP controls, or a NAV facility is optical. Ask what percentage of realized DPI came from unaffiliated third-party buyers.
How should allocators evaluate a continuation vehicle offer?
Find out who set the price and whether anyone independent competed to buy the asset. A fairness opinion documents a process without creating one. With single-asset continuation vehicles now over half of GP-led volume, treat these as primary transactions needing full diligence, not administrative rollovers.
How do you evaluate an operating partner's track record across market cycles?
Separate returns earned between 2009 and 2021 from returns earned since. Cheap capital flattered a lot of judgment. Look for value creation that came from operating change rather than multiple expansion, and ask for a specific case where the partner was personally accountable for something that went badly.
What should an investment committee ask before a re-up?
Three things. What share of returns came from execution rather than leverage or market timing. Which companies have been held past five years and what gets them out. How reserve capital was allocated across performance quartiles. The answers separate skill from vintage luck.
