When does a waterfall error surface?

A waterfall gets modeled once, at fund formation, and then isn't touched again until money actually needs to move through it. The governing rules sit in the LPA, while the model that implements them often goes untouched until the fund's first meaningful realization.

The assumption that matters: hurdle basis

The specific failure point is the hurdle basis — how the preferred return actually accrues. Two questions get answered once, early, and rarely revisited: does the hurdle compound or accrue on simple interest, and does it accrue against committed capital or only capital actually called and deployed. Both are legitimate modeling choices, and both show up in market-standard documents. The problem isn't that one choice is wrong. It's that the choice has to be translated correctly from LPA language into a working model, and never drift from that intent across years, administrator handoffs, and model rebuilds.

Get the basis wrong at the outset — accrue on committed capital when the LPA specifies called, or compound when the LPA specifies simple — and the error is baked into every downstream calculation the waterfall will ever run.

Why nothing catches it earlier

The formula runs every quarter without complaint, because running the formula and distributing capital through it are different things. Quarterly reporting, NAV calculations, capital account statements — none of them require the hurdle to actually resolve into a distribution split. Nothing distinguishes the wrong assumption from the right one until the first distribution forces the model to produce a real allocation.

That changes at the first big exit, when there's real capital to allocate and a formula that has to resolve a specific number. If the hurdle basis was wrong, it's wrong at the worst possible time — with distribution checks about to go out and counterparties looking closely at how the number was derived.

Where the check belongs

The right moment to verify a waterfall's hurdle assumptions is at formation, before the model has ever distributed a dollar. It's also the moment most funds are least likely to question the assumptions they just finished negotiating.