Ask a principal four years into their career at a mid-market fund to explain how their carry will be calculated. Most can describe the headline: twenty per cent over an eight per cent hurdle, subject to vesting. Very few can explain how the catch-up works, what happens to their allocation if a deal in another vintage goes badly, when the first distribution is realistically expected, or what an interim clawback provision would do to a payment they have already banked mentally.
This is not a failure of intelligence. It is a design outcome. In most firms the waterfall lives in a spreadsheet maintained by the CFO, validated once a year by the auditors, and understood in full by perhaps three people. Everyone else holds a rough mental model and a hope.
That arrangement has survived because it has never obviously broken. It is breaking now, slowly, for a reason that has little to do with accounting and a lot to do with the fact that the people affected by it have more options than they used to.
Compensation you cannot model is compensation you discount
Carried interest is meant to do a specific job. It aligns the investment team with the fund's outcome over a horizon longer than any annual bonus can reach, and it retains people through the awkward middle years when the work is done and the value has not yet been realised. It does that job only to the extent that the recipient can form a credible expectation of what they will receive and when.
When the calculation is opaque, people apply a discount. They cannot price the asset, so they undervalue it, and they undervalue it by more than the genuine uncertainty warrants. A vice president who cannot explain their own carry treats it as a lottery ticket rather than a holding, and lottery tickets do not retain anyone. Meanwhile a competing firm makes an offer with a number attached, in cash, that the candidate can verify immediately. The comparison is not close, and it is not close because one side of it is legible and the other is not.
We see the consequence most clearly in exit conversations. Departing mid-level professionals frequently discover, at the point of leaving, that their entitlement differs materially from what they believed. Sometimes it is larger. The direction is not the point. The point is that a compensation instrument whose value is revealed only on departure has failed at the thing it exists to do.
The complexity is genuine, and pretending otherwise helps nobody
The honest defence of the status quo is that this is hard. It is. A whole-of-fund European waterfall and a deal-by-deal American one behave differently enough that intuition trained on one is actively misleading about the other. Catch-up rates of eighty or a hundred per cent change the shape of the curve. Recycling provisions alter the capital base. Interim clawbacks, escrow arrangements and the treatment of write-offs against realised gains all interact.
Then there are the side letters. A fund with sixty LPs might carry a dozen negotiated variations affecting fee offsets, excluded investments, most favoured nation elections and, in some cases, distinct economics for a cornerstone investor or a co-investment vehicle. Layer on multi-currency positions, blocker entities, GP commitment funded partly through management fee waiver, and vesting schedules that differ by hire cohort, and the result is genuinely intricate. The CFO who says this cannot be reduced to a one-page explanation is not being obstructive.
But intricacy is an argument for better explanation, not for less of it. Derivatives are complicated and are nonetheless priced, decomposed and explained daily. Nobody accepts "the model is complex" as a reason for a trader not to understand their own book. The reason the waterfall escapes that standard is historical: it was built in a spreadsheet by one person for the purpose of instructing a bank transfer, not for the purpose of being understood by anybody else.
The same opacity slows every LP query you receive
The internal cost is the more interesting one, but the external cost is easier to measure. When an LP queries a distribution, asks how the preferred return was compounded on a late-drawn commitment, or wants to reconcile their capital account against a management fee offset, the request lands on the same small group who maintain the spreadsheet. The answer requires reconstructing a calculation rather than retrieving an explanation.
To put rough numbers on it, and these are illustrative rather than measured, a non-trivial capital account query at a mid-market fund might take a finance team four to eight hours spread across five to ten elapsed days, most of it spent rebuilding the reasoning behind a figure that was correct when it was produced. Multiply that across an LP base during a distribution quarter and a meaningful portion of the finance function's year is spent explaining numbers that were never recorded with their explanation attached.
Operational due diligence teams have started probing this directly. They ask to see the calculation, not the output, and they notice when the answer depends on the availability of one individual. Key person risk in the investment team is disclosed and understood. Key person risk in the carry model is neither, and it is arguably more acute.
Reproducible is a stronger standard than correct
The target most firms should set is not accuracy, which they largely have, but reproducibility with reasoning attached. Given the same fund documents, the same cash flows and the same side letters, the calculation should produce the same result twice, and each figure should carry a trace back to the clause and the transaction that generated it. A distribution line should be able to answer the question "why this number" without a human reconstructing it from memory.
This is where the technology genuinely earns its place, and it is worth being precise about why. The hard part is not arithmetic. The hard part is that the rules governing the arithmetic live in prose: a limited partnership agreement, an amended and restated version of it, twelve side letters and a subscription document, running to several hundred pages of drafting that qualifies itself in three places. Reconciling a computed number against that prose has always required a lawyer's reading and an accountant's care at the same time.
Long-document reasoning is the specific capability we rely on Claude for in this work. It will read a partnership agreement alongside the side letter set, identify where a specific LP's economics depart from the default, and show the clause that drives the departure, with its reasoning exposed rather than buried in a score. That last property is what makes it usable in a finance function. A model that produces a distribution figure is a liability. A model that produces a distribution figure, the clauses it relied on and an account of where it was uncertain is a working paper, and working papers are something an audit committee knows how to handle.
What an explained waterfall looks like in practice
The practical form is undramatic. Every carry participant can see a current statement: their allocation, its vested portion, the fund's position against the hurdle, and a modelled projection under two or three exit scenarios that the firm states plainly are modelled rather than expected. Every LP capital account line can be expanded to show the inputs, the applicable clause and the date. Every recalculation is versioned, so the difference between this quarter's model and last quarter's is itself a readable artefact rather than a mystery.
Firms hesitate at the first of those, and the hesitation is worth naming honestly. Making allocations legible internally invites comparison, and comparison invites difficult conversations about relative contribution. That is a real cost and some partnerships will judge it too high. Our observation is that the conversations happen anyway, later, with worse information and usually at the point of resignation.
Legibility is the cheapest retention spend available
The forward-looking version of this is quietly attractive. A firm that can show a principal exactly what they hold, how it accrues and what has to happen for it to pay, has converted an abstraction into something closer to an asset the person can reason about. That changes how it is weighed against a competing offer, and it changes it without a single additional pound of economics being granted.
The same structure that makes carry legible to the team makes capital accounts legible to LPs and makes the model legible to a diligence team. Three problems that firms currently treat separately turn out to be one problem: the reasoning behind the numbers was never written down where anybody could find it. Writing it down is now a tractable piece of work rather than a heroic one, which is a better position than the sector has been in for a long time.
