2 and 20 fee structure

The management fee compensates the GP for running the fund, independent of performance. The dominant convention: a rate (commonly 2% for buyout, sometimes 2.0–2.5% for venture) applied to committed capital during the investment period, when the fee is compensating the GP for sourcing and building the whole portfolio, then a step-down, typically to a somewhat lower rate applied to invested cost of remaining (unrealized) holdings, once the investment period ends and the GP's job shifts from building to harvesting a shrinking book. This step-down matters enormously to Net figures: a fee still charged on the original committed amount long after most of the portfolio has been sold would be charging the LP for capital no longer at work, nearly every modern LPA avoids this by re-basing the fee once the investment period closes. Fees are typically funded by additional capital calls during the investment period (so they count toward PIC, exactly as invested capital does) and, after the investment period, are commonly deducted directly from distributable proceeds before they reach LPs, rather than called separately. Many LPAs also include a fee offset: some or all of any monitoring or transaction fees the GP separately charges portfolio companies are credited back against the LP-paid management fee, a mechanism ILPA has pushed hard toward 100% offset, since without it a GP is effectively double-charged for the same work. Worked Examples 1, 2, 5, and 6 all carry a fully Python-verified fee schedule, on-committed then stepped-down, through to Net IRR and Net TVPI.

Carry, hurdle, catch-up, and the two waterfall architectures

Carried interest ("carry") is the GP's share of profit (commonly 20%) and it is the second half of the classic "2-and-20" shorthand alongside the management fee. It is not paid on every dollar returned; it is paid through a waterfall, a sequence of tiers that determines, dollar by dollar, who gets what:

1. Return of capital: 100% to the LP until contributed capital has been returned in full.
2. Preferred return (the hurdle): 100% to the LP until the LP has earned a minimum compounded return, commonly 8% per annum, on capital that has been called but not yet returned.
3. GP catch-up: the GP receives most or all (a 100% catch-up is standard, though 50/50 or 80/20 catch-ups exist) of the next distributions, until the GP's cumulative carry equals its target share (20%) of the profit distributed so far, i.e., of the preferred return already paid plus the catch-up itself.
4. Carry split: everything beyond that point splits 80/20 (LP/GP) for the remainder of the fund's life.

The subtlety worth burning in: once a 100% catch-up tier fully completes, the preferred return stops mattering to the final split, the GP ends up with exactly 20% of total profit above capital, and the LP with 80%, regardless of the hurdle rate, because the catch-up existed only to get the GP to that 20% share, nothing more. The hurdle only protects the LP's share of the outcome when profit is too thin to complete the catch-up (then the GP earns less than 20%, and the LP keeps more). Treating the pref as a permanent floor above 80/20 is a common and costly misreading. Worked Examples 5 and 6 both verify this arithmetic exactly.

European vs American waterfall: what is the difference?

The two waterfall architectures differ in when the tiers above are tested:

• European (whole-fund). The tiers are tested cumulatively, across the entire fund, at every distribution: the GP receives no carry until the fund, in aggregate, has returned all LP capital plus the full preferred return. This is the more LP-favorable structure, is the majority convention across PE funds worldwide today, and needs no clawback mechanism, because the GP is never overpaid relative to the whole-fund outcome in the first place.
• American (deal-by-deal). Each investment runs its own mini-waterfall (its own capital, its own accrued preferred return, its own catch-up and split) the moment that deal exits, without waiting for the rest of the portfolio. This lets the GP earn carry sooner on an early winner, which is part of why it remains more common in venture capital, where outcomes follow a long-tailed "power law" and an early home run may be the fund's defining event. Its cost is that an early winner can pay the GP carry before a later loss (which the deal-by-deal test never sees) is netted against it, so every American-style LPA carries a clawback provision: at the end of the fund's life, if cumulative GP carry exceeds what a whole-fund test would have allowed, the GP must return the excess to the LPs.

Worked Example 5 runs Konkan's full European waterfall event by event, Python-verified. Worked Example 6 runs Cascadia's full American waterfall deal by deal, computes the exact clawback owed, and is the module's sharpest illustration of why the clawback provision exists at all.

{
  "skill": "Building and debugging a multi-tier waterfall or XIRR model in code",
  "use": "Have a chat assistant draft the Python (or Excel) logic for a tiered waterfall (return of capital -> pref -> catch-up -> split) or an XIRR solver, then use it as a first draft you reconcile line by line.",
  "tools": ["Chat assistants with code execution — Claude, ChatGPT", "A local Python/Jupyter environment (ties to the quant-code region's DA1.02/DA1.03 pandas modules and QD1.04's scipy.optimize usage)"],
  "prompt": "Write a Python function that takes a list of dated capital calls and distributions, a preferred return rate, and a carry rate, and returns the LP/GP split at each distribution date under a European (whole-fund) waterfall with a 100% GP catch-up. Show your work: the running outstanding-capital and accrued-pref balances at each step.",
  "verify": "Reconcile the AI-drafted function against a hand-computed 2-3-event mini-case first (e.g., one capital call, one small distribution that only partially returns capital, one large final distribution) before trusting it on a real schedule — a single date, day-count, or tier-ordering error compounds silently. Recompute the final GP-carry-as-percent-of-total-profit check (it must equal the carry rate exactly once the catch-up fully completes) as an independent sanity test, exactly as this module's own verification script does (Guardrail #3, Recompute).",
  "diy": "You must be able to work a waterfall's four tiers by hand on a small numeric example, unaided, before asking AI to generalize it into code — Worked Examples 5 and 6 are exactly that hand-build. The by-hand version is what tells you whether the AI's code is right; without it you have no way to catch a silent tier-ordering bug.",
  "market": "IN"
}

Modern Data Analysis callout. Every number in this module was produced by a script, not a memory: dated cash flows held as (date, amount) tuples (or, done properly, a pandas DataFrame with a datetime column, the DA1.02/DA1.03 pandas-and-time-series bridge in the quant-code region is exactly this skill), an Actual/365 year-fraction helper, and a root-finder (scipy.optimize.brentq, the code-native cousin of the by-hand bisection in Worked Example 3, the same root-finding idea QD1.04 applies to portfolio-optimization problems) run over an NPV function to solve XIRR. The waterfall itself is nothing more than a small loop carrying forward three running balances (outstanding capital, accrued-unpaid preferred return, and cumulative carry paid) through a sequence of dated events, tier by tier. This is the exact pattern a real fund administrator's software runs, at far larger scale; build the small version yourself once, by hand and in code, and no vendor's black-box waterfall output will ever be un-auditable to you again.