Once you have a valuation range, the recovery analysis distributes it. This is mechanical work, which is why it is worth doing precisely: the errors that occur here are arithmetic and entity-mapping errors, and they are avoidable.
Running the waterfall
Begin with enterprise value at your chosen scenario. Deduct administrative and priority claims first — professional fees, which in a large case run to tens or hundreds of millions; Section 503(b)(9) claims for goods delivered in the twenty days pre-petition; priority employee wage and benefit claims; and any DIP facility, which sits at the top. What remains is available to pre-petition claims in priority order.
Allocate to each secured class up to the lesser of its claim and the value of its collateral. A secured claim exceeding collateral value splits: secured up to the collateral, unsecured for the deficiency, and the deficiency claim ranks alongside general unsecured claims. This bifurcation is frequently mishandled and it materially changes unsecured recoveries.
Doing it by entity
A consolidated waterfall is wrong whenever the group has meaningful entity separation. Value must be allocated by legal entity, with each entity's creditors paid from its assets before value flows upward through the equity of that entity to its parent. Guarantees change this: a subsidiary guarantee gives a parent-level creditor a direct claim at the subsidiary, ranking alongside that subsidiary's own creditors.
The practical procedure is to build an entity-by-entity value allocation, map every tranche to its issuing entity and its guarantors, and then run the waterfall at each entity. This is tedious. It is also where recoveries are actually determined in complex groups, and it is the analysis that separates a professional model from a summary.
Base rates from the default studies
Moody's, S&P, and Fitch publish annual corporate default and recovery studies, free with registration. They provide long-run average recovery rates by instrument type and seniority, default rates by rating, and time series across cycles. These are your priors: if your model produces a 70% recovery for a second lien claim, the base rate tells you that outcome sits well above the historical average and invites you to identify what specifically justifies it.
Recovery rates are cyclical. Recoveries in a broad default wave are lower than in an idiosyncratic default, because asset sales compete with other distressed sellers and buyers are scarce. A model built on through-cycle averages will be optimistic in exactly the environment where you need it to be right.
The liquidation analysis
Every disclosure statement contains a liquidation analysis: what creditors would receive in a Chapter 7 wind-down. It exists to satisfy the best-interests test, which requires that each dissenting creditor receive at least its liquidation value. It applies recovery percentages to each asset class — receivables at perhaps 70–85%, inventory at 40–70% depending on type, real estate at appraised value less costs, intangibles usually near zero — and deducts wind-down costs and trustee fees.
These documents are free on any claims agent docket and they are the best available worked examples of the analysis. Read several before you build your own.