An equity waterfall model is the set of rules that turns deal documents into cash outcomes for each equity holder. An “equity waterfall” is the order and math for who gets paid first, how much they get, and when the split changes as returns improve. A “catch-up” is the step that shifts more dollars to the promote or incentive class until the overall split reaches the target.
In private equity, the waterfall is mostly a governance tool. It forces adults in the room to say, in numbers, what happens if the exit is early, late, big, small, or messy. If your model can’t handle a recap dividend, a partial sale, or a down-round, you don’t have a model. You have a slide.
This article explains how to build and review a deal-level equity waterfall model that mirrors the documents, survives real-world cash events, and stays auditable under deadline pressure.
What a waterfall model is (and is not)
A waterfall model is not a valuation model and not a projection model, even if it lives in the same workbook. It starts where operating projections end. Inputs are distributable proceeds, the cap table, and the economic terms: return of capital, preferred return or hurdles, catch-ups, promote or carried interest at the deal level, management equity, and any priority classes.
The model has two jobs. It must mirror the governing documents without “interpretation” sneaking into formulas. And it must be auditable under stress cases such as partial exits, recapitalizations, add-ons, and interim distributions because that’s where disagreements show up and where timelines get tight.
Two mistakes show up again and again. People blend legal economics with management reporting conventions, then wonder why the spreadsheet drifts from the documents. And people build a single-exit toy, then patch it with ad hoc fixes; the patches become the model. If multiple cash events are plausible, the model needs dates and state variables from day one; otherwise IRR hurdles, capital accounts, and catch-ups won’t hold.
Separate the two waterfalls people confuse
In a typical buyout, there are at least two waterfalls that people talk about in one breath but should separate in modeling. The fund-level waterfall allocates distributions from the fund to LPs and the GP under the LPA. The deal-level waterfall allocates proceeds inside the portfolio company equity stack: sponsor equity, management equity, co-invest, SPVs, and any preferred equity at HoldCo or OpCo.
Here we’re talking about the deal-level model for one investment, while keeping an eye on where it ties to fund reporting. The sponsor often holds the deal through a fund vehicle, may have GP commitment, co-invest sleeves, or blockers, and those holdings can change what you report even if the deal economics are unchanged.
Boundary conditions matter. A deal-level model doesn’t allocate between a fund and its LPs unless you’re explicitly bridging parallel vehicles or sleeves. And a fund-level model shouldn’t try to recreate portfolio company instruments beyond what it needs to compute fund distributions. Mixing the two gives you a clean-looking output that nobody can reconcile, which creates bad optics and worse close certainty.
Terms people treat as synonyms like promote, carried interest, incentive equity, sweet equity, MIP, growth shares, and profits interests often are not economically equivalent. “Preferred return” gets abused the most. Sometimes it’s a true contractual preference. Sometimes it’s only a hurdle that switches on the promote. Your model has to follow what the paper does, not what the meeting called it.
Legal architecture sets the math you’re allowed to do
The waterfall is constrained by the legal form of the issuing entity and the instrument design. You can deliver similar economics with different legal tools, but the cash and liquidation rights are what drive allocation.
Common structures include Delaware LLCs issuing common units, preferred units, and profits interests, where the operating agreement can allocate distributions with broad flexibility. Delaware corporations issue common and preferred stock, where corporate law and tax treatment can tighten the range of workable mechanics. UK companies use alphabet shares, growth shares, and preference shares, usually paired with bespoke articles and shareholders’ agreements. Luxembourg holding vehicles (S.à r.l., SCSp) may implement economics across multiple levels because investors need different blockers.
The practical modeling point is simple. Don’t assume a preferred return “accrues and must be paid” unless the instrument creates an enforceable claim under the chosen law and the entity has legally distributable reserves. If the entity can’t legally distribute, your spreadsheet’s “accrued pref” is only a number on a screen. That matters for timing, disputes, and lender diligence.
Ring-fencing matters when investor groups sit at different levels. If management participates at an intermediate HoldCo but the sponsor holds senior preferred at that same level, management can be structurally junior even if the headline plan sounds generous. Model the actual entity that receives sale proceeds, then the actual path of distributions up and down the structure.
Define distributable value before you build tiers
Before you argue about tiers, define what counts as distributable value. Most people start with “exit equity proceeds” (sale price minus debt and fees). That’s necessary, but it’s not enough.
A robust distribution universe has three buckets. First, interim cash during the hold: dividends, tax distributions, intercompany distributions, equity repurchases, and any portfolio-level fees paid from operating cash. Many MIP plans don’t participate in ordinary dividends, or they participate only above a threshold. If you assume pro rata and the plan says otherwise, your outputs will be wrong in exactly the scenarios where mid-year payments happen, including the timing and IRR impact.
Second, exit proceeds: sale proceeds, recap dividends, and value delivered through debt-for-equity exchanges. If consideration includes rollover equity or an earnout, model it explicitly or disclose it as excluded. Earnouts and rollovers are where headline MOIC and cash MOIC diverge, and the divergence drives compensation conversations.
Third, liquidation proceeds. In weak outcomes, liquidation preferences, caps, and seniority drive the result. Your model must allocate correctly when proceeds fall short of invested capital. If it produces negative allocations or forces junior classes to “pay” seniors, rebuild it.
Also define permitted payment channels and blocks. Credit agreements often restrict dividends. Corporate law can restrict distributions. If your model assumes distributions that are not permitted, you will misstate sponsor and management outcomes and give lenders and auditors a reason to slow you down. At minimum, include a clear toggle, “interim distributions permitted: yes/no,” and show the effect.
Choose the paradigm that matches the documents
At the fund level, people talk about European vs American carry. At the deal level, the relevant choice is different. You either model a single global promote measured on total equity invested into the deal entity (net of distributions to date), or you model explicit instrument-level preferences and participation by class.
If instruments differ by class or issuance date, use the instrument-level approach. It tracks reality: each class has rights, and each issuance date changes the base for accruals and hurdles. A global promote approach can work when the legal documents explicitly define the promote as a split after hurdles with no intervening preferences. Otherwise it becomes a shortcut that breaks the first time there’s an add-on, a recap, or a second closing.
If the sponsor reports “carry” on a deal-by-deal basis, produce a bridge from deal-level outcomes to the fund-level framework. Don’t pretend the fund LPA applies directly to the portfolio company equity stack. It doesn’t, and that confusion is expensive when you’re closing a transaction under deadline.
Turn narrative terms into structured inputs
Waterfalls don’t fail because people can’t do algebra. They fail because key definitions live in narrative and never become structured inputs.
- Classes and holders: List every class and every holder, including co-invest sleeves, SPVs, and management vehicles.
- Contributions and dates: Capture amounts and timing by tranche because timing drives IRR hurdles.
- Invested capital definition: Specify whether it means original contributions, unrecovered capital, or net of certain distributions and fees.
- Priority and splits: Translate “capital back, pref/hurdle, catch-up, promote split” into a tier-by-tier rule set.
- Hurdle mechanics: Define IRR vs MOIC, compounding, day count, and the accrual base.
- Management terms: Build in vesting, forfeiture, leavers, repurchases, and any exercise assumptions for options.
- Dilution and conversion: Model conversion features and anti-dilution mechanics if they exist.
Don’t assume “standard.” Capital can be contributed in tranches, fees can be funded with equity, and interim distributions can reset balances. “Standard” language applied to non-standard facts is where disputes are born.
Build a flow-of-funds engine that can survive reality
A durable model is a flow-of-funds engine with a clear ordering of operations. Separate three layers.
- State variables: Track capital accounts or invested capital balances by class and holder, plus any accrued preferred return or hurdle balances.
- Allocation logic: Build a tier engine that allocates a distribution amount using current state and the stated terms.
- Outputs and checks: Produce distributions by event, proceeds by holder, implied MOIC/IRR, and reconciliations.
Time matters even if you expect one exit check. IRR hurdles require dates. Tax distributions and dividend recaps are common. The clean structure is an event table: each row has a date and a gross distributable amount. The engine allocates, updates balances, and moves to the next row. This approach survives add-ons, recaps, and continuation transactions without rewiring formulas, which reduces operational risk and speeds review.
A useful “freshness” rule for 2026 workflow is to treat your waterfall like production code, not a one-off spreadsheet. If you are using AI or script-based modeling to accelerate builds, lock a human-readable “terms dictionary” next to the model. Then test it with a fixed regression pack of scenarios (base exit, low exit, partial sale, recap + exit, add-on + exit). Regression tests catch silent formula drift before it becomes a compensation dispute.
Preferred return vs liquidation preference: name the animal correctly
“Preferred return” can mean three different things. It can be a contractual preferred distribution that accrues and must be paid before common (subject to legally available funds). It can be a hurdle rate that only determines when a promote turns on, with no separate “pref payment.” Or it can be a synthetic preference embedded in preferred equity paired with liquidation preference multiples and participation.
Your model must reflect what the documents do. If the instrument is preferred equity with a liquidation preference, apply it to liquidation proceeds at exit the way the paper says, not as a generic accrual line.
Key choices that move dollars include compounding (simple vs annual vs quarterly vs daily), day count (actual/365, 30/360), accrual base (original invested capital vs unrecovered capital vs net of distributions), and whether tax distributions reduce the accrual base. A small difference in accrual conventions can change hurdle achievement in a five-year hold, which means real money and real incentive consequences.
Catch-up tiers are where flattering spreadsheets appear
Catch-up provisions determine how quickly promote or incentive equity moves from zero participation to its target split. “100% catch-up” usually means the next dollars go entirely to the promote class until cumulative distributions reach the target overall split (say 80/20). It does not mean the promote takes 100% forever.
You must compute the catch-up amount explicitly. That means tracking cumulative distributions across the relevant holders and solving for the amount that equalizes the overall split per the defined base. Approximating catch-up as “20% of proceeds above the hurdle” is usually wrong, especially in moderate exits where catch-up is the convexity lever.
A quick illustration helps keep everyone honest. Suppose distributable proceeds are 200, invested capital is 100, and the hurdle requires the senior class to reach 120 before promote turns on, then the target split is 80/20 with full catch-up. The first 120 goes to the senior class. After that, proceeds may go entirely to the promote until the cumulative split matches 80/20 as defined in the documents; only then does it split 80/20 going forward. The catch-up amount depends on whether the initial 120 is included in the split base, so you must model, not guess.
Management incentive equity: model the instrument, not the slogan
Management participation often behaves like an option: out-of-the-money at entry, then steep participation above a threshold. But the modeling has to match the instrument type.
In US LLCs, profits interests use a “hurdle value” above which the profits interest participates. Model the hurdle value and the participation percentage above it, and reflect forfeiture, repurchase rights, and vesting that change the cap table at exit.
In the UK, growth shares and alphabet shares can have bespoke capital rights and growth thresholds. Treat them as separate classes with explicit participation mechanics. Calling them “options” is convenient, but convenience is not a term in the documents.
Leaver provisions often decide the outcome in mid-cases. A bad-leaver repurchase at nominal value can wipe management economics even in a successful exit if termination occurs before closing. If that governance risk exists, show sensitivities for good-leaver versus bad-leaver outcomes. It’s a small modeling effort with big clarity value.
Map documents and the fee stack so cash ties to paper
Reference the controlling documents and their hierarchy: shareholders’ agreement, LLC agreement or articles, MIP documents, subscription agreements and side letters, credit and intercreditor agreements, fee letters, and monitoring agreements. Waterfall disputes often come from inconsistent definitions of “net proceeds,” “invested capital,” or which fees reduce distributable proceeds.
Model what actually reduces distributable proceeds: transaction and monitoring fees, financing fees and OID, make-wholes, advisory and admin costs, trustee fees, and withholding taxes where applicable. Individually these can look small; near a hurdle they can decide who gets paid. If fees are variable, make them explicit inputs and run sensitivities.
If you need a broader context for where the waterfall sits in the capital stack, keep that framing separate from the allocation math. Likewise, if you are aligning the waterfall with fund incentive language, it helps to distinguish deal promote from promote and waterfall mechanics used in other asset classes and vehicles.
Governance and checks that hold up under deadline
Waterfalls fail more from sloppy term translation and weak change control than from complexity. Watch for ambiguous definitions, inconsistent treatment of interim distributions, multiple closings with assumed single dates, add-ons that dilute some holders but not others, recaps that return capital early, vesting or repurchases right before exit, and escrows or earnouts that shift timing.
- Single source cap table: Tie holder balances to executed exhibits and keep the cap table locked to versioned inputs.
- Terms freeze: Put a “terms freeze” date in the model so late edits are deliberate and traceable.
- Event reconciliation: Check that total allocated equals the gross distribution for every event, with disclosed rounding rules.
- Stress-case suite: Run partial exits, recaps, add-ons, and low outcomes to confirm no negative allocations appear.
- Audit trail: Store PDFs used for approvals and maintain version control so you can recreate what was shared.
If the structure is too complex to model cleanly, it’s usually too complex to administer cleanly. A simple common equity split with vesting can beat an ornate tier stack if your admin team can’t calculate distributions quickly and defend them with an audit trail when it matters.
When you finish a deal and you’re done sharing the model outputs, close it out with discipline: archive the index, versions, Q&A, users, and full audit logs; generate a hash of the archived package; apply the retention schedule; instruct vendor deletion and obtain a destruction certificate; and remember that legal holds override deletion. That sequence protects you when memory fades and stakes rise.
Conclusion
An equity waterfall model is only as good as its fidelity to the documents and its ability to handle real cash events. If you define distributable value clearly, model instrument-level rights when needed, compute catch-ups explicitly, and enforce governance with stress tests, you get a model you can defend in diligence, reporting, and compensation discussions.