Duration Risk and IRR — Why Time, Not Win Rate, Dominates Return

The arithmetic that most memos skip

IRR is a function of both multiple and time: IRR = MOIC^(12/T) − 1, where T is months to resolution. Because time enters the formula as an exponent, its effect on IRR is convex — the marginal cost of an additional six months is not constant, it grows as the base case gets shorter. A case underwritten at an 18-month expected duration and a 2.0x multiple carries a materially different IRR than the same 2.0x multiple realized at 36 months, and the gap between those two outcomes is larger than the gap produced by a full ten-point swing in win probability at a fixed duration. Funders who underwrite win probability to the second decimal and duration to the nearest "12 to 24 months" are optimizing the wrong variable.

This is not a theoretical point. It shows up directly in fund-level IRR reporting: a portfolio with a strong average multiple and a duration tail that runs longer than modeled will report IRR well below what the multiple alone would suggest, and portfolio managers are often unable to explain the gap because duration was never underwritten as a first-class variable with its own distribution.

Duration is not one number — it has a tail, and the tail is where the damage happens

The median duration for a case type understates the risk that matters for fund-level planning. What matters is the tail — the 90th percentile of time to resolution for the relevant cohort — because that tail is what determines whether capital gets trapped, whether a fund needs a bridge facility to meet LP redemption timing, and whether the case survives long enough to be affected by intervening events like a change in the presiding judge or a shift in a circuit's doctrine. Two case types with identical median durations can have very different tails: one clusters tightly, the other has a long right tail driven by appeals, multi-defendant coordination, or docket congestion in a specific venue. A duration model that reports only the median is reporting the number least useful for risk management.

Duration also is not a single clock. Time to settlement, time to adjudication on the merits, time through appeal, and time from judgment to actual cash collection are different processes with different drivers, and a case can resolve legally while still taking a year or more to convert into cash. Underwriting the legal-resolution clock alone and assuming cash follows immediately overstates realized IRR on every case that goes to judgment rather than settlement.

What changes when duration is modeled as a managed variable

Treating duration as a distribution rather than a fixed assumption changes three things in practice. First, position sizing: a case with a wide duration distribution should be sized smaller, at the same win probability, than one with a tight distribution, because the capital-at-risk period is uncertain. Second, portfolio construction: duration distributions correlate within a venue and within a judge's docket, so a portfolio concentrated in cases sharing the same court can carry correlated duration risk that a spreadsheet built around independent case-level assumptions will not catch. Third, monitoring: a case that is tracking toward the tail of its duration distribution six months in is a different risk than one tracking toward the median, and that divergence is detectable well before the case actually resolves — if the monitoring framework is built to watch for it.

Duration risk compounds across a portfolio, not only within one case

A single case's duration distribution matters for that case's expected IRR. A portfolio's aggregate duration exposure matters for something else entirely: whether the fund can meet its own capital-call schedule, its LP redemption terms, and its next vintage's deployment pace. A fund that models duration case by case but never aggregates the resulting distribution across the live book can be individually well-underwritten and collectively exposed to a liquidity mismatch that no single case's file would reveal on its own.

This is the direct link between duration modeling and fund-level treasury management: a book whose duration distribution skews toward the tail in the same period a fund faces a redemption window or a follow-on capital call is a structural risk that shows up on the fund's own balance sheet before it shows up in any single case's underwriting file — and it is only visible if duration is tracked as a portfolio-level, time-varying aggregate rather than a static per-case input frozen at intake and never revisited.

The same logic extends to how a fund reports performance to its own LPs. A fund-level IRR figure that blends resolved and unresolved positions, without disclosing the duration distribution still open in the unresolved book, can look stronger than it will ultimately prove to be once the long-tail cases in the open book actually resolve. LPs evaluating a fund's track record benefit from seeing the duration distribution of both the resolved and the still-open portion of the book side by side, not a single blended number that quietly assumes the open positions will behave like the ones that have already closed.

What to ask for from an intelligence provider

• 01The full duration distribution for the relevant case type and venue — median, tail (90th percentile), and dispersion — not a single expected-duration estimate.
• 02Separate duration models for time-to-legal-resolution and time-to-cash, not one clock treated as a proxy for both.
• 03Evidence that duration and outcome probability are modeled jointly, since the two are not independent — cases that run long often do so for reasons correlated with how they eventually resolve.
• 04A methodology for updating the duration estimate as the case develops, so drift from the original underwriting assumption is visible before resolution, not discovered at it.