A note on reading a limited-partnership agreement before building the deal model: the document is not a legal artefact to be passed to counsel and then ignored. It is the operating manual for the cash flows the GP can credibly run. Six clauses, in particular, change the model materially. The modeller who reads them first, and reads them carefully, builds something the IC will accept. The modeller who waits for legal review to surface them builds twice.
Why the LPA sits upstream of the model, not alongside it
A fund model has two layers. The deal layer captures the underlying asset: revenue, cost, capital expenditure, exit. The fund layer captures the legal and economic structure imposed on those flows: when capital is called, what return tiers apply, who gets paid in what order, what fees are charged, what events can suspend or accelerate the structure. The LPA defines almost every input to the fund layer. Read the deal first if you must; read the LPA before you wire anything to it.
The clauses below are not an exhaustive list. They are the clauses that, in our experience advising funds and family offices, change the output the most when they are misread.
The six clauses that drive the model
1. The waterfall
The waterfall determines the order in which distributions move between LPs and the GP. Two structures dominate institutional practice: the European whole-of-fund waterfall, where LPs receive their committed capital plus a preferred return across the entire fund before the GP earns carry, and the American deal-by-deal waterfall, where carry can be paid on each realisation, subject to claw-back at fund end.
What to model: the precise sequence of tiers, the basis on which each tier is calculated (contributed capital, invested capital, or fund NAV), and the events that trigger each distribution. A waterfall miscoded by one tier can move the GP's economics by several hundred basis points across the fund life. Build the waterfall as a separate worksheet with explicit cash buckets per tier; do not nest it inside the deal cash flow.
2. The hurdle (preferred return)
The hurdle is the minimum return LPs receive before the GP participates in profits. Eight per cent compounded annually is the institutional default; lower rates appear in lower-risk strategies and higher rates appear in higher-risk or emerging-market vehicles.
What to model: the rate, the compounding basis (typically annual, occasionally quarterly), the day-count convention, and the capital base on which it accrues (committed, drawn, or invested). The choice of base matters: a hurdle on committed capital pays the GP later than a hurdle on drawn capital, and the difference compounds across an eight to 10 year fund life. Flag the day-count to legal if the LPA does not specify it.
3. The catch-up
The catch-up is the band after the hurdle where the GP receives a disproportionate share of distributions (commonly 100 per cent or 80 per cent of profits) until the GP has caught up to its target carry split on aggregate fund profits. The catch-up converts a tiered waterfall into something closer to a flat split once the hurdle is cleared.
What to model: the catch-up rate, whether it is full (100 per cent to the GP) or partial (commonly 80:20 or 50:50), and the exact arithmetic by which the structure transitions from catch-up to the standard carry split. Catch-up clauses are written ambiguously more often than any other section of the waterfall. If the formula is not unambiguous in the LPA, the model output is not defensible; flag it.
4. GP commitment
The GP commitment is the capital the manager itself contributes to the fund alongside LPs. One per cent of fund size is the conventional minimum; two to five per cent appears in funds where the manager wants to signal alignment, and 10 per cent or more is occasionally seen in family-office sponsored vehicles.
What to model: the size of the GP commitment, the funding mechanism (cash, management-fee offset, deferred), and whether the GP commitment participates in carry or is excluded. The funding mechanism in particular is overlooked: a GP that funds its commitment through a management-fee offset has different cash dynamics from one that contributes external capital, and the difference shows up in the GP's economics, not the LP's. The LP returns may be identical; the GP's reported track record will not be.
5. Key-person and removal provisions
Key-person clauses suspend investment activity if named individuals cease to be actively involved with the fund. Removal provisions allow LPs (typically by supermajority) to dismiss the GP for cause or, in some funds, without cause.
What to model: not a base-case impact, but a scenario. The key-person scenario suspends the investment period, freezes new deals, and converts the fund into a wind-down vehicle running existing assets only. The downside model should include this case explicitly, with assumptions about the cessation date, the assets in flight at that point, and the management-fee adjustment that typically accompanies suspension. Most fund models ignore key-person risk; the IC will ask about it.
6. NAV facility and subscription-line provisions
Subscription credit lines (drawn against unfunded LP commitments) and NAV facilities (drawn against fund-level asset value) are now ubiquitous in private capital. The LPA defines what is permitted: facility size as a percentage of commitments or NAV, allowed uses, maximum tenor before LP capital is called, and any LP consent thresholds.
What to model: the effect of the facility on reported IRR and on the timing of LP cash calls. A subscription line that defers LP calls by 12 months can lift fund-level IRR by 200 to 400 basis points without changing the underlying asset cash flow at all. That gap between fund-level IRR and asset-level IRR is the single most contested number in current LP-GP conversations. Model both, and present both. Hiding the facility effect inside a single headline IRR is the sort of thing the IC will, eventually, find.
What to flag for legal before finalising
The modeller is not the lawyer, and the model is not the LPA. The following categories of ambiguity should be flagged in writing to the legal reviewer before the model is signed off:
- Any waterfall tier whose calculation basis (committed, drawn, invested, NAV) is not stated unambiguously in the LPA.
- The day-count and compounding convention of the hurdle, where the LPA defers to "market practice" or is silent.
- The catch-up arithmetic, where the transition between bands is described in prose rather than formula.
- Any management-fee offset, transaction-fee sharing, or expense reimbursement provision that affects the GP's net economics.
- The interaction of subscription-line interest with the hurdle calculation, which a surprising number of LPAs leave open.
The right output of this review is a short memorandum: the assumption the model has taken on each point, the LPA reference, and the question for legal. The exchange usually closes in a day. The alternative (rebuilding the waterfall after IC) costs a week and erodes confidence in the model.
A closing observation
Modellers tend to treat the LPA as a finished input: a set of percentages handed down from the legal team to be entered into a sheet. It is more useful, in our experience, to treat the LPA as a draft specification that the model interrogates. Ambiguities surface in the modelling, not in the reading. A waterfall that produces an odd carry distribution in year seven is, in most cases, pointing at a clause that the modeller and the lawyer have read differently. Catching that gap before the IC pack is printed is what the work is for.