Value the Company. Then Price Its Debt Three Ways.

Open Leverley See the features

Institutional valuation
and LBO financing.
Offline by default,
in one file.

Who it’s for

Anyone who has to say what a company is worth and how much debt it can carry.

Learn more

What it does

Value the company, price its debt three ways, test the covenants, then hand over a workbook anyone can audit.

Learn more

How it works

From a company name to a live-formula workbook, in ten steps.

Learn more

Security and privacy

One file, your browser, your disk. No account, no telemetry, no certification claimed — just how it is built.

Learn more

Start with the built-in template.

Open Leverley

Six features

What it does

Value the company, price its debt three ways, test the covenants, then hand over a workbook anyone can audit.

01
Five-method valuation
EBITDA and revenue multiples, a DCF with comps, the venture-capital method and liquidation value, side by side.
Learn more
02
Sixteen LBO structures
Bank, hybrid and private-credit ladders priced side by side on the results tape, then an LBO analysis of sixteen deal types — traditional LBO, take-private, carve-out, roll-up, MBO, secondary, recaps, HoldCo / OpCo, asset-based, sale-leaseback, PIK toggle, loan-to-own, continuation vehicle, preferred, founder buyout.
Learn more
03
Covenant tests and headroom
Leverage and interest cover are tested each year per structure, with breach year and headroom.
Learn more
04
Live-formula Excel workbook
Every model is written into the bundled template as real formulas, ready to audit and extend.
Learn more
05
Batch study and Stata bundle
Drop in many Orbis exports at once; every priced company feeds the study tables and a dated Stata bundle.
Learn more
06
One file, offline by default
One HTML file with the spreadsheet engine bundled. Statements and assumptions never leave the laptop.
Learn more

Three roles

Who it’s for

Anyone who has to say what a company is worth and how much debt it can carry.

Advisory
Financing consultant
Shows a sponsor, a family office or a founder what a deal returns under bank, hybrid and unitranche terms.
Lending
Private-credit analyst
Screens a pipeline of names for whether a private-credit structure clears the sponsor’s hurdle and keeps covenant headroom.
Research
Finance researcher
Builds a firm × structure panel across dozens of companies and runs the regressions from a reproducible dated bundle.

Ten steps

How it works

From a company name to a live-formula workbook, in ten steps — Data, Model and Analyse — and the Study across companies.

01
Data · steps 1–4
Name the company and say whether it is listed. A listed company arrives with its statements from the directory or the lookup; a private one starts from an Orbis export or its own accounts. Then the assumptions you set for it — multiples, growth, margins, tax, the three financing ladders and the method settings — and its operating metrics, its market data and the findings.
02
Model · step 5, Valuation
Five methods side by side — EBITDA and revenue multiples, a DCF with its grid, the venture-capital method, liquidation value — the listed peers, the precedent transactions you type in, and the range they span. One method sets the entry multiple.
03
Model · steps 6–7, Deal model and 16 structures
What the assumptions price for the three financing ladders — the sources and uses, the sponsor returns, the debt schedule and the operating model — then the sixteen deal types.
04
Analyse · steps 8–9, and Study
The scenario landscape, base, upside and downside cases, a custom change typed in words and the seeded Monte Carlo, which runs by itself; the sensitivity grids, the return bridge and the debt repayment for one company; Tables 1–10, the regressions and the comparisons across the saved study.
05
Analyse · step 10, Reports
The committee summary, the valuation summary, the downloads — the live-formula workbook, the one-pager, the appendix, the JSON, the whole-batch files — and the save history.

Four topics

Security and privacy

One file, your browser, your disk. No account, no telemetry, no certification claimed — just how it is built.

What it sends
Nothing, unless you switch lookup on
Online lookup is off as shipped and the file makes no network request at all. Switched on, it sends only the company name, the country (if you type one), the sector you tag and the ticker you pick to the lookup address you set — never statements, assumptions or results.
Company names, to this server only
This page comes from a Leverley server, so online lookup starts switched on: the company name, the country (if you type one), the sector you tag and the ticker you pick go to this server, which answers from public company registers and, where its operator has set one up, a market-data provider. It also gives this page its latest exchange-rate table; asking for it sends nothing about the company. Statements, assumptions and results never leave the browser. Switch lookup off on the Select company screen.
Where data lives
In this browser, on this computer
The study, a loaded directory, your settings and saved mappings sit in the browser’s local storage for this file; the save history, when on, in its IndexedDB. Statements and priced companies live only in the open tab.
How to clear it
One button in the Guide
Guide › Security and privacy lists every stored key with its size and clears all of it in two clicks; clearing the browser’s site data for the file does the same.
What is inside
Libraries inlined, nothing fetched
JSZip, SheetJS, TradingView Lightweight Charts, pdf.js and PptxGenJS, with Manrope and JetBrains Mono, are embedded in the file under permissive licences — versions and licences in the Guide.

Feature 01

Five-method valuation

EBITDA multiple, revenue multiple, an intrinsic DCF with comps as the cross-check, the venture-capital method and liquidation value, computed side by side for every company.

Solves: An owner or a buyer asks what the company is worth, and a single multiple is not an answer anyone can defend.
For: Owners and founders, corporate-finance advisers, PE associates screening targets

Step 1 · Select company asks for the company’s name, an optional country, its sector and whether it is listed. A listed company brings its market data and peers from the directory; a private one starts from its statements, with an optional private-company discount. Under Step 5 · Valuation the method you choose sets the entry multiple all three financing structures start from, beside the peers, the precedent transactions you type in and the range chart; the workbook carries every method as live formulas on its Valuation sheet.

Feature 02

Sixteen LBO structures

Bank debt, syndicated hybrid and private-credit fund priced side by side from one set of assumptions, with leverage, spread, amortisation and OID per tranche; then the LBO analysis prices the company under sixteen deal types, each with editable terms, funded from those three ladders.

Solves: A client asks “what if we used unitranche instead of the bank?” and the answer means rebuilding the debt schedule three times.
For: Financing consultants, leveraged-finance analysts, private-credit associates

Each structure is its own ladder under Step 4 · Assumptions. Set the ladder once, change any tranche, and the results tape recomputes IRR, MOIC, entry leverage and minimum ICR for all three on every edit. On Step 7 · 16 structures, the LBO analysis prices sixteen deal types — traditional LBO, take-private, carve-out, roll-up, management buyout, secondary buyout, dividend and leveraged recaps, HoldCo / OpCo, asset-based, sale-leaseback, PIK toggle, loan-to-own, continuation vehicle, structured preferred and founder buyout with earn-outs — and opens any one of them in detail: deal terms, sources and uses, tax structure with stock or asset deals, the 338(h)(10) election and NOL / §382 limits, exit, debt paydown with covenants by tranche, simulation and stress test, returns by equity class, cash flows, situations and the operating model.

Feature 03

Covenant tests and headroom

Maximum net debt / EBITDA and minimum interest cover are tested each year per structure, with breach year and headroom reported, or switched off for cov-lite terms.

Solves: IRR alone hides the year a structure stops working.
For: Private-credit analysts, bank leveraged-finance teams, restructuring advisers

Covenant limits default to 1.3× entry leverage and 1.1× EBITDA / interest and can be edited per structure. The results tape shows the first breach year; the study records breach as a flag for Tables 8 and 10.

Feature 04

Live-formula Excel workbook

Every model is written into the bundled three-scenario template as real formulas, so the workbook can be traced, audited and extended after it leaves the app.

Solves: Web tools hand over static numbers that a committee or a client’s finance team cannot check.
For: Advisory boutiques presenting to IC, lenders’ credit teams, auditors

Generate one workbook per company, or All models as a .zip for the batch, from the results tape or Step 10 · Reports. Inputs land in named cells, the debt schedule and returns are formulas, and a JSON summary sits alongside for anything downstream.

Feature 05

Batch study and Stata bundle

Drop in one or many Orbis .xlsx / .csv exports; each workbook becomes a priced company and adds three firm × structure observations to the study.

Solves: Pricing twenty names by hand takes days, and a cross-company regression built from hand-made models cannot be reproduced by a reviewer.
For: PE associates, transaction-services teams, MSc and PhD researchers, faculty

Step 2 · Financial statements accepts Orbis (BvD) statements or any labelled P&L and balance sheet with year columns; split statements are merged, unit scale and currency are read from the file, and the sector, which is also the industry the study groups by, is detected from NACE, SIC or NAICS codes or the shape of the statements. Under Study, Study tables and Regressions build Tables 1–10 (paired within-firm tests, determinants of sponsor IRR, downside resilience, industry interactions and a robustness grid) and export a dated bundle whose do-file loads the dataset by relative path and runs clustered by firm.

Feature 06

One file, offline by default

A single HTML file with the spreadsheet engine, the template and the fonts bundled. It opens with no network at all; no install, no account.

Solves: Client financials cannot go to a SaaS tool, and IT will not approve one for a single mandate.
For: Independent consultants, banks with strict information security, students on locked-down machines

Open the file in Chrome or Safari and it runs. Online lookup for listed companies stays off until you switch it on; then it sends only the company name, the country (if you type one), the sector and the ticker you pick to the lookup service you name, never statements or assumptions.On a page served by a Leverley server, online lookup starts on and sends only the company name, the country (if you type one), the sector and the ticker you pick to that server, never statements or assumptions; the downloaded file still opens with no network at all. TradingView links are plain links that open only when clicked. Light and dark modes, the saved study and the batch persist in the browser’s local storage, and charts and tables export to XLSX and PDF from inside the file.

Guide

Reading the screens

ⓘ means more information. An ⓘ holds the detail that would otherwise crowd the screen: the periods read from the statements, what a page or a field does, a table’s notes (“Notes ⓘ” under it), how a switched-off feature is turned on. Hover over it, or click it (tap it on a phone), and a small window opens beside it; move away, click elsewhere or press Esc to close it. A window opened by a click stays open until then. Where the figure is computed, the window ends with How it is calculated: the formulas the engine uses (a subscript is the year — y runs over the hold, LTM is the latest twelve months — or the tranche i), and, where they depend on the open company, its own values. Try it:

Info buttons ⓘTap an ⓘ for the detail behind any label. Everything else stays short.

Dashes —No value yet. It fills in once the earlier steps are done.

Bank · Hybrid · PC FundThree ways to fund the same deal. They keep the same colours on every screen.

Results tapeAt the foot of Valuation, Deal model and Reports, so you see the effect of each change.

The steps. Ten numbered steps in three groups — Data: 1 Select company, 2 Financial statements, 3 Operating metrics, 4 Assumptions; Model: 5 Valuation, 6 Deal model, 7 16 structures; Analyse: 8 Scenarios, 9 Analytics, 10 Reports — and Study: Study tables, Regressions and Compare, which read the saved study rather than one company. Choose the company first, then read its operating metrics and set its assumptions, ladders and method settings; the rest reads from those. The steps stand in the left-hand column (on a phone, the bar under the logo — Step 3 of 10 · Operating metrics, with ◀ and ▶ — opens them from the foot of the screen): a step you have passed is a ring with a check, only the open one is filled, and a click opens any step. A step that needs a company says so until the batch holds one. Every screen begins with its title and an ⓘ beside it, which holds the step’s line, what the page is for and what it holds; the content fills the same width on every step; the secondary bubbles are folded, a chip beside their titles, and open with a click. A bar fixed to the foot of the window holds ← Back, which returns to the page you were on before, step by step as far as you went (with nowhere to return to, the last page of the first steps, and on Select company the welcome page), Start over once the batch holds a company (a second click confirms it: the batch empties, the saved study stays) and, at the right, the step’s call to action, named after the step it opens — Financial statements, Operating metrics … —, which goes on once what that step needs is there: a company in the batch for every step from Operating metrics on; until then it is greyed and a click on it says why at the foot of the screen. Reports and the Study screens have none. On a company’s steps a bar above the bubbles holds the company, chosen from a drop-down, and the page’s own buttons. Welcome, the light / dark switch and this Guide sit at the right of the logo bar, which scrolls away with the page. The parts follow Google’s Material Design 3: pill-shaped buttons — filled for the call to action, outlined for the others, text buttons for the lesser ones —, outlined fields, one drop-down size everywhere (280 by 40 pixels), checkboxes, radio buttons, switches, segmented buttons with a check on the chosen segment, chips, cards with rounded corners and the ⓘ windows as rich tooltips; the typeface and the colours stay Leverley’s. Everything you can click is sienna — the filled buttons that move the work on (the call to action at the foot, Add file(s), Generate workbook, the Download menus) with navy words, all the other buttons as a sienna outline or in sienna words, the links, chips, choices, checkboxes, radio buttons and switches and the rows that open something, and in light mode the ⓘ and the drop-downs too. In dark mode the ⓘ and the windows they open, the fields you type in (the Country field’s suggestions too) and the drop-downs are light blue. Text in other colours is there to be read. The mark of the step open is filled in sienna and the marks of the steps done are sienna rings; the titles of the steps and of the bubbles are in the text colour, black in light mode and white in dark mode, though a click on a step opens it and a click on a bubble’s title folds it; and the arrow beside a bubble’s title, which shows whether it is open or folded, is sienna in both modes. Every field you type in or choose from has an outline in that colour too — sienna in light mode, light blue in dark mode — two pixels wide while you type.

Downloads. Under Study, Study tables, Regressions and Compare each carry one ⬇ Download menu holding that page’s own exports — Tables 1–5 or 6–10 as XLSX or PDF, the Stata bundle, or the open comparison’s charts, slides, tables and chart data — plus the dataset as XLSX or CSV and the open page as a PDF. The 16 structures page has its own: the analysis as a workbook of engine values, as one PDF or as PowerPoint slides, and the JSON. The company’s own files — the workbook, the IC one-pager, the appendix, the JSON — and the whole batch’s are on the results tape and under Step 10 · Reports › Downloads. Each table keeps its own Copy, XLSX and PDF buttons, and each chart its XLSX and PDF.

The (i) by each field of the LBO analysis works the same way on hover. A click on it also keeps the field’s note open under the field — what it is, which way it moves the numbers, which cards change and how it is calculated — until you click it again.

Offline by default

This page was served by a Leverley server. Online lookup starts switched on here and talks to that server (the address shown under Select company › Online lookup); switch it off there and nothing goes online. A downloaded copy of the file behaves exactly as described below.

Nothing leaves this computer unless you switch on online lookup. The file carries its own spreadsheet engine, template, chart library and fonts, so it opens and prices without a network connection.

Online lookup is the only feature that goes online. It is off until you switch it on under Step 1 · Select company › Online lookup; then it sends the company name, the country (if you type one), the sector and the ticker you pick to the address you set, and never statements, assumptions or results.
TradingView links are plain links. They open a new tab only when you click them.
The batch, the study and your settings stay in this browser’s local storage. Clearing the browser’s site data removes them, so export the study dataset to keep it. Every stored key, and a button that clears all of them, is under Security and privacy.

Online lookup (optional)

On this page nothing needs installing. It came from a Leverley server and looks companies up through it: public company registers (SEC EDGAR, ESEF filings, GLEIF, Companies House) and the market-data provider its operator has set up. If the server asks for an access key, enter it once under Select company › Online lookup; this browser keeps it.

With the downloaded file, online lookup needs a lookup service: the address of a Leverley server (with its access key, if it has one), or the small service leverley_api.py on your own computer. That one answers name searches with a listed company’s statements, market data and a year of prices from a free public market-data source (for your own research only: free market data is licensed for personal use), plus SEC EDGAR filings if you give it an e-mail address.

Set it up once, in a terminal, in the folder that holds leverley_api.py, directory_builder.py and requirements_local.txt. Current Pythons (Homebrew on macOS, most Linux systems) refuse a system-wide pip3 install, so the scripts get an environment of their own:

# once
python3 -m venv .venv
source .venv/bin/activate        # Windows: .venv\Scripts\activate
pip install -r requirements_local.txt
# each session, in a terminal window you keep open
source .venv/bin/activate
python3 leverley_api.py

Add --edgar-email with your e-mail address to include SEC EDGAR filings. Then switch on Select company › Online lookup. The default address is http://127.0.0.1:8787; change it there if the service runs elsewhere, and press Check. Avoid --break-system-packages: it can damage Homebrew’s own Python.

Every answer is stored in the local directory, so a company found once stays available offline.

Directory of listed companies

A directory is a dated JSON snapshot of listed companies (statements, market data and, optionally, a year of prices) that Financial statements loads, Select company searches and the comparables under Step 5 · Valuation draw their peers from. Every company keeps its source and as-of date.

python3 directory_builder.py, run from the same environment as the lookup service (Online lookup), builds it from free public market data, with SEC EDGAR and ESEF filings where available.
--tv-csv adds a TradingView screener CSV export. Screener exports carry no price history.
--prices adds one year of daily closes for the Market card’s chart.

Only tickers leave the computer while the script runs; Leverley reads the finished file offline.

Workbook cell map

The pages use plain labels. The generated workbook keeps every input in a fixed cell, so a reviewer can trace each figure (since v1027 a Cover sheet comes first; the cells below are on the model sheets and have not moved):

Input	Cell
Revenue (LTM)	J8
EBITDA (LTM)	J9
D&A (LTM)	J10
CapEx (LTM)	J11
Net working capital	J12
Existing net debt	E10
Entry multiple (×EBITDA); Valuation!B86 when a valuation method sets it	E8
Enterprise value	E9
Exit multiple (×EBITDA)	E16
Fees and expenses (rate × revenue, J8)	E13
Revenue growth, margin expansion, tax rate	J15 · J16 · J17
Financing scenarios: the three ladders	L7:O8
Private-company discount	Valuation!B25

Formula fidelity

Step 4 · Assumptions › Advanced switches the debt-schedule formulas between two modes.

Hardened (default) caps each paydown at the tranche balance and adds a Cash Balance row (94), so net debt at exit = O93 − O94. Mandatory amortisation, sweep %, OID financing fees, the tax floor and covenant tests apply in this mode only.
Faithful reproduces the original template exactly, including its uncapped second-tranche sweep. It is there to reconcile with the template.

Every saved observation records its mode, and the study notes any saved in faithful mode.

Study tables and Regressions

Each company saved to the study adds three observations, one per financing structure (bank, hybrid, private credit); generating a workbook saves it automatically.

Study tables (Study) builds Tables 1–5 and the screening funnel; Regressions holds the consistency checker.
Regressions builds Tables 6–10: paired within-firm tests, determinants of sponsor IRR, downside resilience, industry interactions and a robustness grid (+100 bps on rates, −1.0× exit multiple, −2.0 pts revenue growth a year by default).
The Stata bundle holds the estimation workbook leverley_dataset.xlsx, the full tables_6_10.do and a variable dictionary. The do-file loads the dataset by relative path and clusters by firm.

Security and privacy

This is one HTML file that runs in your browser. It has no account, no sign-in and no telemetry: nothing is counted, logged or reported anywhere.

On this page. It was served by a Leverley server, so online lookup starts on and sends the company name, the country (if you typed one), the sector you tagged and the ticker you pick to that server (and the access key, when the server asks for one). It also asks that server for its latest exchange-rate table, a request that carries nothing about the company. Statements, assumptions, results and the study never leave the browser.
What the file sends. Nothing, unless you switch on online lookup under Select company › Online lookup. With it on, the file sends the company name, the country (if you typed one), the sector you tagged and the ticker you pick to the lookup-service address shown on that screen — and nothing else (when that address is a Leverley server with a newer exchange-rate table than the file’s, the file also asks for the table; that request carries nothing about the company). Statements, assumptions, results and the study never leave this computer. With it off (as shipped) the file makes no network request of any kind: not for fonts, charts, PDF reading or any export.
Where data is stored. In this browser only, for this file’s origin: the study, a loaded directory, your settings and the saved mappings in localStorage; the save history (when switched on) in IndexedDB. The table under Stored data below lists every key with its size. Everything else — imported statements, priced companies, the batch — lives in the open tab and is gone when you close it (export what you want to keep).
How to clear it. The button at the end of this guide, Clear everything this file stored in the browser, removes all of it (two clicks). Clearing the browser’s site data for the file does the same.
Libraries inlined in the file (no download at run time): JSZip, SheetJS, TradingView Lightweight Charts, pdf.js, PptxGenJS and the two fonts — versions and licences under Licences.
No certification is claimed. The statements above describe how the file is built; they are not an audit, a compliance attestation or a security certification.

Stored data

Data · steps 1–4

Four pages, in the order of the work: Select company, Financial statements, Operating metrics and Assumptions. The section holds the companies themselves — who they are, their historical statements, the assumptions you set for them, the figures the models start from — and the batch of companies loaded in this tab. Choose the company first — a listing found by name, or its statements —, check what was read on Operating metrics, where every company opens, and then set its operating assumptions, financing scenarios and method & risk settings on Assumptions; the valuation, the deal model, the scenarios and the reports follow from those.

Select company asks first whether the company is listed — two radio buttons, Public and Private —, then for its name and its sector (one of twelve valuation sectors, or “not sure”) side by side and, for a listed company, an optional country. A listed company is found by name in the loaded directory or, with online lookup on, through the lookup service: press Enter in Company name, or click the search icon in it. The country, if you give one, is sent with the name and narrows the search; leave it empty to search by name alone. Every match shows a percentage. The match with the highest percentage is marked best match; when the top rows show the same percentage none is marked, so tell them apart by exchange and country. Pick the listing and the company opens at once on Operating metrics, priced with the batch template’s assumptions until you set its own, with its statements, market data and a TradingView link. A private company goes on with the call to action at the foot, Financial statements, which takes the name and the sector you typed along, to upload its accounts and from there, with Operating metrics, to the same page; the sector you tag seeds the valuation sector. Several companies at once: open Financial statements and drop all their files in one pick; each company starts from the batch template.

Financial statements. Drop one or many files, or pick them with Add file(s). Orbis (BvD) exports, any labelled P&L and balance sheet with year columns as .xlsx / .xls / .csv, and — with the v1021 switch on — .pdf (text layer) and .docx (Word tables). Files that share a company name merge into one company; untick “One company per workbook” to merge a whole pick into one company, and tick “Review mapping” to check the mapped lines before a company is built — two checkboxes side by side under the uploader. Every uploaded file is listed under them with the company it went to. Its delete button (a second click confirms) takes the file out and reads the company again from the files it has left — or takes the company out of the batch when no file is left, or when what is left no longer makes a company —, and Delete all empties the batch (the saved study stays).

PDF: the text layer of every page is rebuilt into rows; a page whose text holds two or more years on one line is a statement page, its numbers are placed under the nearest year column, and “EUR thousand”-style tokens set currency and units. A scanned PDF (no text) is refused with a message asking for xlsx or csv.
docx: every Word table is read as a sheet; the paragraphs before the first table become the cover (name).
Review table: always for PDF and docx, and for xlsx / csv when the switch under the drop area is on. It lists every line the parser maps, the source row (file, sheet or page, row), its values by period, and lets you re-point a line at another row, type values or unmap it; the company name can be corrected too. Corrections apply through the same parser, so provenance records the file, page and matched label.
Mapping memory: corrections are saved per layout (source type plus the set of row labels) and reapplied on the next import of that layout — the import log and the findings say “mapping reused”, and Mapping… in the bar over the bubbles of Step 6 · Deal model reopens the table. Saved layouts are listed under Label mapping, where they can be forgotten.
The batch is every company loaded in this tab: listed under Financial statements (Mapped) and in the company drop-down above the bubbles of a company’s steps; choose one to open it. Directory: a dated JSON snapshot of listed companies built offline by directory_builder.py (Directory), searched from Select company and the source of the peers under Step 5 · Valuation.

Assumptions is step 4, the last of Data — reached with Next from Operating metrics, from the stepper or from the batch template — so that a company’s assumptions are set before anything is read from the valuation, the deal model, the scenarios or the reports; Next goes on to Step 5 · Valuation. Four boxes, each with its title and its arrow inside it; Method & risk and Advanced are folded as shipped, and a click on a title folds or opens its box. In each ladder the tranches (multiple, rate, amortisation, OID) stay in view; the second tranche, the debt cap, the cash sweep, the covenant tests and the advanced terms fold under More terms. Each explanation has one ⓘ: a field’s note opens from the field’s own ⓘ, and the same field in the next ladder shows its ⓘ only while the first one is out of view. Four sections, as in the batch template; inside a box every field has its name above it and, where it has one, a short note under it, the fields in one to three columns that share the bubble’s width (three on a wide window, as wide as the three ladders), and the size preset’s four sizes under its name; the bubbles run across the page from edge to edge, like the pages around it. No results tape on this page (the Deal model shows what the values price). The sections: Operating assumptions (size preset, entry and exit multiples, EV, growth, margin expansion, tax, fees as a share of revenue); Financing scenarios, the three ladders side by side — bank, hybrid (bank + private credit) and private-credit fund, each with tranche multiples, all-in rates (base rate + spread; a hand-edited rate drops the link), amortisation, OID, a second tranche, the EV cap, the cash sweep, covenant settings and the Advanced block (revolver, PIK, step-up, equity cure, haircut, fees, default detection); Method & risk settings — the ATAD / CIR cap, the mixed-ladder draws, each a checkbox beside its name, and, at its end, the Optional features, a checkbox each, side by side (Optional features); Advanced — hardened or faithful formulas (Formula fidelity).

The batch template. Every company added to the batch starts from one set of defaults: the size preset (Auto, as shipped, gives each company the preset of its LTM EBITDA in US dollars — Small cap below USD 25m, Mid cap USD 25–100m, Large cap USD 100m and above, converted at the average exchange rate of its LTM year (monthly ECB euro reference rates since 1999, built into the file; the Federal Reserve’s for the Taiwan dollar) — and sizes it again when its EBITDA, units or mapping change, keeping what you typed; Small / Mid / Large hold one preset and Custom clears the fields; a preset sets the entry multiple, pre-import growth and margin placeholders and the three financing ladders; the dollars only choose the band — every figure stays in the company’s own currency), the operating assumptions, the three ladders and the method settings. Batch template… — in the bar over the bubbles of Assumptions, or beside Find a company → at the foot of Financial statements — opens it on this page, for a batch whose companies should start from the same values; with no company open, Assumptions says what to add first. Reset to defaults restores the shipped values; statements imported later replace growth, margin and tax with per-firm derived values unless you typed them. A company already in the batch keeps its own values, edited on Assumptions once it is open.

Operating metrics is where a chosen company arrives — from Select company once a listing is picked, from Financial statements with Next, or from the batch list. It is the open company: the bar over the bubbles (the company drop-down, to open another company of the batch, and the findings chip), the Company bubble on one row (the industry — it follows the sector —, the currency, the country and its base rate, the source units, the periods behind their ⓘ), Last twelve months (the LTM inputs: revenue, EBITDA, D&A, capex, NWC and net debt, each figure’s source and years behind its ⓘ), then, folded as shipped, the Market card of a listed company, the historical statements and Findings & provenance: the validation panel, folded until opened — the severity in one column and the message beside it, its location, a “Go to” link that opens the page and the card and focuses the field, and the mapped lines (Findings codes). The Industry box of the Company bubble follows the Sector box of the Valuation card: it offers the same choices, and changing either one changes the other. The study tables group companies by this same industry, that is by sector, whether the sector was detected, stated on the first screen or set by hand. Companies saved to the study before v1029 take the sector they were valued with, or the sector of their old industry tag; the old four-industry grouping stays in the dataset as the industry4 column.

The results tape at the foot of the steps that read the company — Valuation, Deal model and Reports (steps 5, 6 and 10); not Assumptions, Operating metrics, 16 structures, Scenarios or Analytics — prices all three structures on every edit: debt raised, leverage, sponsor equity, exit net debt, min ICR, covenant status, headroom, MOIC, IRR — and, once a Monte Carlo run exists, P(IRR ≥ hurdle) and P(breach), greyed with “re-run” when inputs changed since the run. Its buttons are the downloads of Step 10 · Reports.

Step 5 · Valuation

One page. Valuation — five methods: the EBITDA multiple, the venture-capital method, the DCF with comps as the cross-check, the revenue multiple and the liquidation value, side by side, with the sector, the mode (public, private, intrinsic) and the method that feeds the entry multiple — or Manual, which leaves the entry multiple as typed or preset (Valuation methods). The method in use is marked Selected in the table and ● on the valuation range, in sienna. The DCF grid is a card of its own: nine EV / EBITDA multiples at WACC ± 1 pt (rows) and terminal growth ± 0.5 pt (columns), the one in use in the middle, its row marked in the same colours, with the build-up above it — the WACC from the cost of equity (risk-free rate, β × ERP, size premium) and the cost of debt, and the terminal value as a multiple of year-5 EBITDA against the exit multiple. Under Detail: the comps’ median and interquartile range and the liquidation table. The valuation is the base case’s in every operating case: Upside and Downside keep its entry multiple, and the Valuation sheet of their workbooks is this one. Next goes on to Step 6 · Deal model. The valuation range and the DCF grid stand side by side on a wide window; the valuation settings and the precedent transactions are two bubbles at the foot of the page, folded as shipped.

Valuation range: every method’s EV on one axis. The DCF band is the WACC × growth grid, the EBITDA-multiple band the peers’ interquartile range when enough peers price; the venture-capital method is a cap and the liquidation value a floor, so each is one mark. The entry EV in use is the dashed sienna line; a listed company’s market EV (market cap + net debt) is dotted.
Trading comparables: the peer set the valuation used — from the loaded directory on ex-ante criteria (same sector, then region, then size band) — with each peer’s EV / EBITDA, EV / revenue, growth and margin, and the criteria and the medians above the table. Without a directory there is no peer set and the comps-based values fall back to the base multiple and the sector default; the card says so.
Precedent transactions: deals you type in — target, date, EV / EBITDA, EV / revenue — kept with the company. Their median and interquartile range are drawn on the valuation range for reference; nothing else reads them: not the entry multiple, the tape, the study or the workbook. The file holds no transaction database.
Valuation settings: the method default, the sponsor hurdle (one setting for the VC method, the pass/fail flag and the Monte Carlo share), the risk-free rate and its basis (the ten-year government bond yield of the statements’ currency, as shipped, or the domicile base rate), the equity risk premium, the size premium, terminal growth and its cap at the risk-free rate, a WACC override, the default β, the control premium and the minimum number of peers, the private-company discount and the liquidation recoveries. They are one set for the whole batch, not per company, until a company is given values of its own: Change per, at the top of the card — Whole batch or This company — is chosen before a field can be changed. Whole batch sends a change to every company, replacing a value a company had of its own; This company keeps it to the open company, and each value of its own says so under its field, with the batch’s; Follow the whole batch again drops them. The choice ends when another company or page opens. The valuation method is the default for companies added later, so it changes for the whole batch only. The same rows sit in the batch template’s Method & risk settings, where a change reaches the whole batch.

Steps 6–7 · Deal model and 16 structures

Two pages: Deal model — what the assumptions set under Step 4 · Assumptions price for the three financing ladders — and 16 structures, the LBO analysis of sixteen deal types (LBO analysis).

Optional changes (a bubble at the foot of the Deal model, folded as shipped): the four sections of Step 4 · Assumptions — the same values, so a change made here is made there; open it to change an assumption beside what it prices. Sources & uses and Sponsor returns stand side by side on a wide window; the operating model is folded as shipped. Deal model, every title above its bubble, for the three ladders: Sources & uses at close — the equity purchase price, the net debt refinanced, the fees on one side; each tranche, the preferred, the earn-outs and the sponsor equity on the other; Sponsor returns — sponsor equity, exit EBITDA and EV, exit net debt, the equity value at exit, MOIC and IRR, against the hurdle, with the covenant status; Debt schedule — for the ladder you pick, each tranche’s opening balance, interest, paydown and closing balance by year, the revolver, the cash balance, net debt, net leverage and interest cover; Operating model — revenue, growth, EBITDA, margin, D&A, EBIT, interest, taxes, net income, capex, working capital and levered free cash flow by year. Every figure is the engine’s, as the tape and the workbook carry it; the workbook writes the same schedule and model as live formulas. A bar over the bubbles holds the switch of Operating cases, Case in the Excel workbook — Base, Upside, Downside —, and says what the page shows: with Upside or Downside chosen, every figure of the Deal model and of the results bar (its name beside Show results) is that case’s, priced as Generate workbook prices it — at the base case’s purchase price —, while your inputs and the valuation stay as entered and the study rows and the other exports stay on Base.

16 structures (step 7; also from the tape) opens the LBO analysis: the company priced under the sixteen LBO deal types — traditional LBO to founder buyout — with one in detail: deal terms, sources and uses, tax structure (stock or asset deal, 338(h)(10), NOL and §382), exit, debt paydown with covenants by tranche, simulation and stress test, returns by equity class, cash flow breakdown, situations and the operating model. From the Deal model, Next goes on to it; the sixteen stand as tiles over the chosen one, whose card holds its sections as chips and its leverage, IRR and MOIC; the overview of all sixteen is folded at the foot.

Steps 8–9 · Scenarios and Analytics, and Study

Five pages: Scenarios and Analytics for the open company, then Study tables, Regressions and Compare on the saved study. Every title stands above its bubble.

Scenarios. Scenario landscape first: the three structures by leverage and IRR, each sized by sponsor equity. Operating cases (on as shipped): Base is the deal model as entered; Upside is the management case — growth and margin typed, blank = base; Downside is growth −2.0 pts a year, exit −1.0×, rates +100 bps as shipped, the study’s robustness shocks. Each case is priced for all three ladders, side by side in Cases × structures. Case in the Excel workbook chooses the case that Generate workbook and All models (.zip) write into the Excel file: the table marks it In the Excel workbook, and a note at the foot of the screen confirms each change. Every case keeps the base case’s valuation and purchase price: Upside and Downside change the operating model, the exit multiple and the rates, never the entry — the workbook’s Valuation sheet and the entry multiple of its scenario sheets are the base case’s, as the table prices them. The Deal model and the results bar show the chosen case too, as a preview of the workbook, with the case’s name on each; your inputs, the valuation, the other pages, the study rows and the other exports stay on Base. What if… is in development: its bubble stays folded, its title greyed with In development beside it, and its switch is off for now. It is to be the custom case: one change at a time, typed in words — pc leverage 6x, growth -2pts — shown before it applies, applied through the same setters the inputs use, with one-step Undo (grammar). Downside — Monte Carlo: seeded paths per structure with P(breach), P(default), P(MOIC < 1×), P(IRR ≥ hurdle) and the IRR percentiles (Monte Carlo). Its settings — N and the seed, the shock switches (a delayed exit, a refinancing test at the bullet) — are the block at the top of the card. The simulation runs by itself a moment after every change of the inputs, so its figures, the tape’s probabilities, the committee summary and the IC one-pager are never stale; the same seed gives the same result, and Run N paths re-runs it by hand. Next goes on to Step 9 · Analytics.

Analytics. Sensitivities (on as shipped): two-way IRR and MOIC grids per structure, re-run for every cell. Return bridge: equity in to equity out per ladder — EBITDA growth ((exit EBITDA − entry EBITDA) × entry multiple), the change in multiple ((exit − entry multiple) × exit EBITDA), net debt paid down (debt raised − net debt at exit) and, as the remainder, fees and other items — as a waterfall and a table; equity proceeds are MOIC × equity invested, as the engine reports them. Debt repayment: net debt, the share of the opening debt repaid, net leverage and interest cover year by year for the three ladders, with the covenant caps dashed on the chart.

Study tables. Each saved company adds three observations (firm × structure). Tables 1–5 (sample composition, summary statistics, structure characteristics, main outputs, outputs by industry) build live; the screening funnel applies the ex-ante filters in order and lists exclusions. Import brings in generated workbooks (re-run by the engine) or dataset exports; the saved-firms list re-tags industries. With save history on, a Study history card keeps versions of the whole dataset (Method note). Its exports — Tables 1–5 as XLSX or PDF, the dataset, the page as a PDF — are in the Download menu.
Regressions. Tables 6–10: paired within-firm tests, determinants of sponsor IRR (OLS, firm-clustered SEs), downside resilience, industry interactions and the robustness grid (rates +100 bps, exit −1.0×, growth −2.0 pts/yr by default; editable above Table 10). Every card carries its own self-contained Stata snippet; the consistency checker re-reads quoted figures from the dataset. The Stata bundle holds leverley_dataset.xlsx (dataset and robustness sheets), tables_6_10.do (loads the workbook by relative path, clusters by firm) and a README with the variable dictionary (Variable dictionary) — and history.json when save history is on. The bundle, Tables 6–10 (XLSX, PDF) and the dataset are in the Download menu.
Compare. Five views on the saved study: two companies, the attribution bridge (Bank → PC Fund gap into leverage, pricing and amortisation / sweep effects), two industries, all companies and all industries, which reports, per industry and structure, the number of firms, the mean, median and interquartile range of IRR, MOIC and minimum ICR, and the covenant breach rate. The bar that stays at the top of the window while the page scrolls names the open view and holds a button for each of its charts and tables — a chart ranked by the Selection’s measure by that measure (MOIC, IRR …), a table by its number (Table C5) — with Expand all and Collapse all, as the table index of Study tables and Regressions. Exports, in the Download menu: all tables (XLSX or PDF), chart data (XLSX), every chart (PDF), the page as PDF and a PowerPoint file with a title slide and one 16:9 slide per chart and table, as native charts and tables in the same series colours. Each table and chart also has its own buttons.

Step 10 · Reports

One page, with the results tape; every title stands above its bubble. Summary for the committee (on as shipped): one rule-based paragraph per structure, every figure read from the engine, with Copy in its header; it also goes on the IC one-pager. Valuation summary: the five methods with the one in use and the three ladders’ debt, leverage, equity, MOIC, IRR and covenant, side by side.

Downloads, one card with two parts. This company: Generate workbook (live formulas; also saves to the study), Save to study, the IC one-pager PDF (A4, or US Letter from the optional features), the sourced assumptions appendix, the JSON summary, and the link to the 16 structures page, whose Download menu holds the analysis as a workbook of engine values, as one PDF, as PowerPoint slides and as JSON; the same buttons sit on the results tape of the pages that read the company. Whole batch: All models (.zip, with history.json when save history is on), Save all to study and the batch report PDF.
Save history (v1021, off as shipped): versions of this company with one-click restore, kept in this browser’s IndexedDB; the Study history card under Study · Study tables keeps versions of the dataset.
The logo on what you export (v1027). Every PDF, Excel file and PowerPoint the app writes carries the Leverley logo and the line “These are the stats from Leverley”. In a PDF they sit in the top margin of each page; in a PowerPoint, on the title slide and at the foot of each slide; in an Excel file, on a Cover sheet that comes first and also gives the date, the build and a list of the sheets. In the dataset files that Stata, R and Python read, that sheet is called About Leverley and comes last, so the data stay on the first sheet. A page printed from the print dialog carries them too. CSV, JSON and .do files carry no logo, because other programs read them as data. The logo is the three bars and the letters at one height: the letters’ cap height equals the height of the mark.
Generate workbook is disabled while an error that would make the workbook wrong is open (NEG_EQUITY, SCALE_MISMATCH, NEG_EBITDA and a blocked tape); the tape note and Findings & provenance (Step 3 · Operating metrics) say why.

Label mapping rules

Every sheet (or page, or table) is classified from its name and first rows as P&L, balance sheet, cash flow, ratios or cover. The first row with two or more dates (dd/mm/yyyy, yyyy-mm-dd, FY2023, 2023 or an Excel date) is the period header; periods are sorted newest first and the first sheet’s periods become the master. Currency and unit tokens (“EUR thousand”, “th USD”, “(000s)”, “million”) are read near the header; without a token, figures above 10⁸ are full units, below 2 × 10⁴ thousands, and anything between raises UNITS_AMBIGUOUS.

Each row’s label is normalised (lower case, punctuation removed) and matched against an ordered list per line: a label matches when it equals the pattern or starts with it followed by a space. Rows containing “per employee”, “ratio”, “(%)”, “(x)”, “margin”, “growth”, “cash flow” or “per share” are skipped. The first matching row with at least one number wins; a second hit in another sheet replaces it only when more complete (DUP_FIELD). Statement hints keep balance-sheet patterns off P&L sheets and the reverse; a sheet with no classification (a csv, a PDF page) is open to every line.

Line	Patterns, in order
Revenue	operating revenue (turnover) · operating revenue · turnover · total revenue · revenue · net sales · sales revenue
EBITDA · EBIT · D&A	ebitda · operating profit (loss) [ebit] / operating profit / ebit / operating income · depreciation & amortization (and spellings) / d&a
CapEx (cash flow)	additions to fixed assets · capital expenditure(s) · purchase of tangible · purchases of property · capex; capitalised development from “increase/decrease in other long term assets” (outflows only)
Working capital	debtors / trade receivables / accounts receivable · stock / inventories · creditors / trade payables / accounts payable · working capital (memo)
Net debt	long term debt / non-current borrowings · loans & short-term debt / short-term debt / current borrowings / loans · cash and cash equivalents / cash at bank / cash
Other lines	cost of sales, gross profit, other operating expenses, profit before tax, income tax, net income, interest paid, R&D, tangible and intangible fixed assets, goodwill, total assets, current and total liabilities, shareholders’ funds, the Orbis exchange-rate line

Derived inputs: EBITDA = reported line, else EBIT + D&A, else revenue − COGS − opex + D&A; CapEx = 3-year cash-flow average as a share of revenue (balance-sheet-implied as the last resort, flagged above 15% of revenue); NWC = debtors + stock − creditors; net debt = long-term debt + short-term loans − cash; growth = average of the last three year-on-year rates, in local currency when the Orbis FX line is present; margin expansion = average change in the EBITDA margin; tax = the domicile’s statutory rate (25% default).

Saved mappings

Method note

Hardened versus faithful. Faithful reproduces the bundled template exactly: tranche 1 pays down MIN(levered FCF, balance) and tranche 2 MIN(LCF, LCF − tranche-1 paydown) — the template’s uncapped quirk, which can send tranche 2 negative (NEG_TRANCHE); taxes are EBT × rate with no floor; nothing else applies; exit net debt is gross debt. Hardened (default) caps each paydown at the tranche balance, pays mandatory amortisation first (a share of original principal per year, capped at the balance), sweeps a set share of the excess levered FCF, treats OID and sponsor fees as uses of funds, floors taxes at zero, applies the ATAD / CIR interest cap, carries a cash balance (row 94; exit net debt = O93 − O94), funds shortfalls from the revolver, accrues PIK to principal, tests the maintenance covenants every year with the breach actions (step-up, equity cure), detects a payment default and runs the recovery waterfall, and writes every mechanic as live formulas (rows 109 onwards). With every mechanic at its inert value the hardened formulas reduce to the template’s arithmetic.

Template layout (the rareliquid three-scenario workbook; one sheet per structure, named BankLBO, Syndicate LBO and PC Fund LBO): columns J–O hold years 0–5. Anchors validated on load: B7 LTM EBITDA, B19 Sources and Uses, B35 Revenue, B74 Debt Schedule, B95 IRR, B105 MOIC. Inputs sit in E7:E16 and J8:J17 (cell map); the capital structure in L7:O8; sources and uses in rows 23–29; the operating model in rows 35–71; the debt schedule in rows 78–94 (tranche 1 rows 78–81, tranche 2 rows 84–87, totals 90–93, cash 94); the returns block in rows 97–106; the live mechanics block from row 109 (blue inputs in column I). Extra sheets: Model Methods (the run’s parameters and sources), Valuation (live formulas for the five methods; E8 links to its B86 when a method is applied) and, since v1021, Sensitivities (engine values).

Consistency checker (Study · Regressions): a CSV of figures quoted in the text (metric,value,tol) is re-read from the saved dataset and diffed. firm:<name>|structure:<key>|field:<column> looks up one cell (the run ID is shown, so the figure is traceable to a run); mean:<field>|structure:…|industry:… takes a column mean over a slice (industry takes a sector key such as software; industry4: one of the four old industries); mean_gap:<field> is the mean within-firm PC − Bank difference (the paired contrast the tests rest on); n_firms and n_obs check the sample counts. The tolerance defaults to the larger of 0.5% of the quoted value and 10⁻⁶. Each test exists so that a reviewer can confirm any number in the write-up against the dataset fingerprint, without the app.

Monte Carlo (details), covenants and the interest cap (details) and the five valuation methods (details) have their own sections.

Covenants and the ATAD / CIR cap

Maintenance covenants (hardened mode, unless cov-lite): every projection year tests net debt / EBITDA ≤ cap and EBITDA / interest ≥ floor. The cap defaults to 1.3 × the structure’s entry leverage; the floor to 2.0× (bank), 1.5× (hybrid) and 1.1× (private credit); both are editable. The first failing year is the breach year; headroom is cap − actual (turns) and actual − floor (turns), and the tightest values go to the study. Breach actions: a pricing step-up in basis points from the following year, and an equity cure that restores net debt to the cap (an interim equity cash flow in the IRR). Cov-lite means springing-only, no maintenance test.

Payment default: when cash on hand + levered FCF + the undrawn revolver cannot meet mandatory debt service, the schedule freezes at that year and the company is sold at EBITDA × (exit multiple − haircut, 2.0 turns by default); proceeds go to the revolver, then the tranches in order, and the equity takes the residual. MOIC and IRR are measured on that recovery.

ATAD Art. 4 / UK CIR: deductible interest in a year = min(interest + carried-forward disallowed interest, ratio × max(EBITDA, 0)), the excess carried forward; ratio 30% by default and from the domicile table when detected (Netherlands 24.5%, Finland and Norway 25%, Switzerland none — asset-based thin-cap instead, exported as missing). Taxable income = EBIT − deductible interest − financing and monitoring fees; tax = max(taxable, 0) × rate. Interest cover is tested on the full interest charge.

The five valuation methods

1 · EBITDA multiple: EV = EBITDA × M, where M = comps median EV/EBITDA × (1 + control premium) × (1 − private-company discount) when at least the minimum number of peers (5) price, else the base multiple (size preset or typed). Peers come from the loaded directory on ex-ante criteria: same sector, then region, then size band.
2 · Venture-capital method: EV = exit value / (1 + hurdle)^N, with exit value = year-N EBITDA × exit multiple, or year-N revenue × EV/revenue when EBITDA is still negative at exit. The most a buyer can pay unlevered and still clear the hurdle — a cap, not a price.
3 · DCF (comps as the cross-check): k_e = r_f + β × ERP + size premium, with r_f the ten-year government bond yield of the statements’ currency unless typed (the domicile base rate when the file holds no yield, or when the risk-free basis says so) and the size premium 3.0 % Small cap, 1.5 % Mid cap, 0.5 % Large cap by the Auto band unless typed (v1046); WACC = (1 − w_d) k_e + w_d k_d (1 − t), with w_d from market net debt / (net debt + market cap) or the default; unlevered FCF_t = EBITDA_t − t × max(EBIT_t, 0) − CapEx_t − ΔNWC_t on the engine’s projection; EV = Σ FCF_t / (1 + WACC)^t + TV / (1 + WACC)^N, TV = FCF_N (1 + g) / (WACC − g) (Gordon), with g no higher than r_f while the cap is on, or EBITDA_N × exit multiple when chosen or when WACC ≤ g + 0.5 pt. A WACC ± 1 pt × g ± 0.5 pt grid ships with it, and the card sets the terminal value as a multiple of EBITDA_N against the exit multiple.
4 · Revenue multiple: EV = revenue × R × (1 + premium) × (1 − discount), with R from an OLS of peer EV/revenue on growth and margin (bounded to 0.5–2× the peer median), else the peer median, else the sector default (disclosed).
5 · Liquidation value: Σ book value × recovery rate (receivables 75%, inventory 50%, PP&E 40%, intangibles and goodwill 0%, other 25%; × 0.6 when forced) − costs (5%) − non-debt liabilities; the equity floor subtracts debt. A floor, never a price.

Selection: the sector rule (software and media: revenue multiple above 20% growth, else EBITDA multiple; infrastructure, health and staples: DCF; energy and real estate: EBITDA multiple with the liquidation floor; financials excluded ex ante; growth above 30%: VC method; distressed: liquidation floor) when the method is “auto”, or the method you choose. The implied EV/EBITDA feeds the entry multiple; typing a multiple stores an override. Manual (as shipped) leaves the entry multiple as typed or preset. The ⓘ after Valuation — five methods sets out the rule step by step with the reason for each step, what the Method choice and the mode do, and the open company’s case.

Monte Carlo

N seeded paths per structure (1,000, seed 42 by default). Four shocks — revenue growth (sd 3 pts), margin expansion (sd 1 pt/yr), exit multiple (sd 1.0×) and the base rate on every coupon (sd 75 bps) — are drawn as correlated normals: Box–Muller on a mulberry32 stream, correlated through the Cholesky factor of [[1, .30, .40, −.10], [.30, 1, .20, −.10], [.40, .20, 1, −.20], [−.10, −.10, −.20, 1]]. Each path re-runs the full engine. Reported: P(covenant breach), P(payment default), P(MOIC < 1×), P(equity wipe-out), IRR and MOIC at P5 / P50 / P95 (IRR over the paths with an IRR), the expected sponsor loss given default and, since v1021, P(IRR ≥ sponsor hurdle) with paths that have no IRR counted as failing. Switches: a delayed exit (+2 years) and a refinancing test at the original bullet (fails above a maximum leverage). The same seed gives the same result; the run ID stamps the inputs it was run for, so a changed input shows as “re-run”.

LBO analysis — the sixteen deal types (v1024)

Under Step 7, the 16 structures page (also from the results tape) is the LBO analysis: the open company priced under the sixteen LBO deal types of the prototype — LBO models · 16 structures — with one deal type opened in detail below the grid. Each deal type is a set of deal terms with defaults (LBOCore.STRUCTURE_META), editable on the screen and kept per company. LBOCore.resolveStructure resolves the terms against the company’s inputs, its balance-sheet lines (receivables, inventory, fixed assets), its market value (a listed company in public mode) and the three ladders of the Deal model into the scenario and the input vector the engine prices; the size preset, the domicile base rate and every hand-edited ladder value carry through. Traditional LBO on the hybrid ladder prices exactly as the tape’s Syndicate/Hybrid; on the bank or PC ladder it equals Bank Loan LBO or PC Fund LBO.

What each term does (v1025). Every editable field on the screen carries an (i) button; it opens a note under the field with what the field is, which way it moves the numbers and which cards change (the Deal terms card has Explain every term to open them all). Nothing on the screen moves the results tape, the study rows or the generated workbook, except the exit multiple, which is shared with Operating assumptions. The same notes, for every deal type and term:

Groups and chart colours: A Buyouts (blue, the bank series: who sells and how the price and the equity are set), B Recapitalisations (violet, the hybrid series: no change of control), C Structure & financing (gold, the PC series: where the debt sits and what secures it) and D Special situations & equity (green, --sc-eq: distress, GP-led secondaries, structured equity, founders). Tranches keep their instrument colour in the charts: bank and ABL debt blue, notes and HoldCo or toggle paper violet, unitranche gold, vendor notes green.

What the prototype asked for beyond the list. Stock vs asset deals and the 338(h)(10) election: the Tax structure card offers Stock deal, Asset deal and Stock + 338(h)(10); the last two create a stepped-up basis (a share of EV) deducted over the amortisation period, and the carve-out brings its own step-up through its terms while the tax card is at its default. NOL and Section 382: an opening loss balance, an annual §382 limit (the hint shows equity value × a 3.5 % long-term tax-exempt rate) and an offset limit (80 % of taxable income under US rules for post-2017 losses) — losses are used each year within both limits, the balance rolls forward and the cash-flow table shows the line. Earn-outs: several per deal, each a share of EV due in its year when EBITDA has grown at the test rate; a shortfall of cash is funded by the sponsor as a dated injection (counted in the equity invested), so an earn-out never defaults the debt. Covenant modelling by tranche: every tranche may carry its own cumulative leverage limit (senior-only, total) or sit outside the maintenance tests (the HoldCo note), in which case the coverage test uses the cash coupons of the tested debt; the Debt paydown card shows the table.

The detail cards: Deal terms (the terms of the selected deal type with their defaults and hints, the financing ladder, and the notes that say how they resolved for this company) · Sources & uses (purchase price, debt, the equity cheque, the loan-to-own price paid, debt / equity; the uses including separation costs and bid costs, the sources including sale-leaseback proceeds, preferred, each earn-out; the equity invested over the hold) · Tax structure (deal type, entity, local surcharge, step-up, NOL and §382, the combined rate) · Exit strategy (exit type with its typical fee, holding period, exit multiple, transaction fees, exit EV and equity value) · Debt paydown (starting debt, interest, paid down, debt added by add-ons or a recap, ending debt; debt balance by tranche; cash-flow components by year; the debt structure with each tranche’s type, toggle years and covenant; covenants by tranche) · Simulation & stress test (base case, the stored-seed Monte Carlo for this deal type or all sixteen, the deterministic stress ladder, the rule-based deal assessment, distributions, EBITDA variation against IRR, IRR percentiles) · Returns by equity class (the deal type sets the rollover class and the incentive pool — management, founder, GP and rolling investors, existing owners — and the company-level fields override them) · Cash flow breakdown (EBITDA to cash available for debt service with NOL, add-on draws, equity injections and the recap as their own lines) · Situations · Operating model.

What stays where. Everything typed on this screen — deal terms, tax structure, exit, situations, equity classes, the operating path — is kept per company (S.lbo, with the company in the batch) and prices this screen only. The results tape, the study rows and the generated workbook keep the Deal model, so every v1022 number is unchanged (golden test). A pill, deal settings differ from the tape, says when the two diverge; terms edited marks a deal type whose terms differ from the catalogue. The exit multiple is the one shared assumption: editing it here edits Operating assumptions under Step 4 · Assumptions. The screen’s exports are its own: JSON and XLSX of the grid (with the changed terms and the resolution notes), the selected deal type’s schedules, covenants by tranche, returns by class, the stress ladder, the Monte Carlo paths and the settings. The generated workbook gains a Structures sheet (engine values) only when its switch on that screen is on.

Engine mechanics, each inert at its default and applied in hardened mode only. v1023: a no-sweep tranche; preferred equity (PREF_SHORTFALL); earn-outs (EARNOUT_PAID / EARNOUT_MISSED); situations and the dividend recap (RECAP); exit transaction fees; the step-up deduction; the operating path. v1024: covenants by tranche (covLevMax, covExclude); a PIK toggle tranche that pays in kind at a higher rate when EBITDA / cash coupons falls under its trigger (PIK_TOGGLE); several earn-outs with the sponsor funding any shortfall; a sale-leaseback at close (proceeds as a source, rent off EBITDA); other sources and uses at close; a dynamic recap that re-levers to a target multiple in a year; debt draws and equity injections by year (add-ons); an equity paid override (loan-to-own: the claims bought plus new money replace the cheque in MOIC and IRR); a MOIC floor on the preferred; and NOL with the §382 annual limit and the offset limit. The stress ladder cuts EBITDA by 5–50 % in every projected year on top of the deal type’s own adjustments; the Monte Carlo keeps its seed.

Reading the grid. Financing names the Deal model ladder the deal is funded with; the asset-based deal is sized on the balance sheet instead (advance rates on receivables, inventory and fixed assets, estimated from working capital and D&A when the statements carry no such lines). Leverage is debt raised ÷ LTM EBITDA after the EV cap. Min ICR, the covenant column and the breach or default year come from the same covenant tests as the tape, plus the tranche limits. The engine’s payment-default rule is strict — FCF, cash and the undrawn revolver must cover mandatory debt service every year — so a thin-margin company defaults under the more levered deal types; the stress threshold and the Monte Carlo say how far from that line each deal sits. A deal type that cannot price with its terms says so in its row. The Monte Carlo columns appear once a run exists for a deal type and go stale with the inputs or the terms, like the Monte Carlo card under Step 8 · Scenarios.

What if… grammar

In development: What if… is off for now — its bubble under Step 8 · Scenarios stays folded and says so, and its switch is greyed; the grammar below is what it takes when it returns.

The command bar under Step 8 · Scenarios › What if… takes one change at a time: [structure] parameter value. It shows the parsed change before applying it, applies it through the same setters the inputs use, re-runs the tape and offers one-step Undo. This grammar is the interface a later language layer will call.

Part	Accepted
structure	bank · hybrid (synd) · pc · all — required for structure parameters, absent for company parameters
structure parameters	leverage · spread (rate, coupon) · amort · oid · pik · sweep · cap · covenant-lite (covlite) · levmax · icrmin · revolver
company parameters	growth · margin · exit · entry · tax · fees · ebitda · revenue · netdebt · capex · nwc · hurdle · mode
value	a number with a unit: 6x · 8% · +50bp · −2pts · 120 (MM) — a leading + or − is a change, a bare number (or =number) sets the value; on / off for covenant-lite; hardened / faithful for mode

Examples: pc leverage 6x (tranches scaled pro rata) · bank spread +50bp (every bank tranche; hand-set rates leave the base-rate link, as in the card) · all exit -1x (exit is company-level; the structure word is allowed with “all”) · growth -2pts · hybrid covenant-lite on · margin = -0.5pts · entry 11x (an override when a valuation method is applied) · hurdle 18% · mode faithful (whole batch, like Advanced under Step 4 · Assumptions). Anything else answers with the accepted forms.

Findings codes and the validation panel

Every finding has a severity (error, warning, info), a location (step, card, field) and — where the field is on screen — a “Go to” link that opens the card and focuses it. Errors marked as blocking disable Generate workbook and show why in the tape note. The classification:

Optional features and their defaults

Sensitivity tables, operating cases and the committee summary (since v1038) and the PDF / docx import with its mapping review table (since v1043) are on as shipped; save history stays off until switched on, and the What if… command bar is off while it is in development (since v1055). None of the six changes IRR, MOIC, minimum ICR, the covenant breach year or the valuation — the golden test below checks that against v1020. The switches live under Step 4 · Assumptions › Method & risk settings (for the open company and in the batch template) and here; they persist in this browser.

Default	Value	Where to change it
Sensitivity steps	±1.0× in 0.5× steps (exit multiple and total leverage); ±4 pts in 2-pt steps (growth); ±1 pt in 0.5-pt steps (margin expansion)	Step 9 · Analytics › Sensitivities (per company)
Sponsor hurdle	20%	Step 5 · Valuation › Valuation settings › Sponsor hurdle IRR (one setting for the VC method, the pass/fail flag and the Monte Carlo share)
Operating cases	Base = as entered; Upside (the management case) = growth and margin typed (blank = base); Downside = growth −2.0 pts/yr, exit −1.0×, rates +100 bps (the study’s robustness shocks)	Step 8 · Scenarios › Operating cases › Case in the Excel workbook, and the same switch over the Deal model (per company): Base unless chosen otherwise; the Deal model and the results bar show the case chosen, at the base case’s valuation and purchase price; your inputs and the study rows always carry Base
Save history	snapshots on Save to study and Generate workbook; last 50 per company and per study	Step 10 · Reports › Save history; Study · Study tables › Study history
Workbook gate	NEG_EQUITY, SCALE_MISMATCH, NEG_EBITDA and a blocked tape disable Generate workbook	fix the error; the code list above
Review table	always for PDF and docx; for xlsx and csv when switched on; corrections remembered per layout	Step 2 · Financial statements switch; Label mapping › Saved mappings
IC one-pager	A4 portrait	Step 4 · Assumptions › Method & risk settings › Optional features › IC one-pager page size (US Letter)
Compare › All industries	mean, median, IQR, firms and breach rate always shown in that view	Study · Compare

Stata variable dictionary

Rendered from the same list the bundle’s README is written from. Order in the dataset: every v1020 column first, the v1021 columns at the end; Tables 1–10 and tables_6_10.do read the same columns as before.

Walkthrough

Ten clearly fictitious companies in an Orbis-style CSV layout, two files each (a cover file with the name, city and country, BvD-style ID, NACE code and trade description; a statements file with four years of P&L, balance-sheet and cash-flow lines in thousands of the local currency). Every figure is invented.

Import the ten. Open Leverley, open Step 2 · Financial statements (or set a size preset first in the batch template — Batch template… beside Find a company) and drop all twenty files at once there with “one company per workbook” on. Files that share a name merge, so ten companies appear in the batch, each tagged with its sector from the NACE line and its domicile from the cover. Measured: the 20 files parsed and the ten companies built and priced in about 0.2 s.
Price them. Press Operating metrics →: the first company opens priced; every other company in the batch is priced as it is opened or saved. Measured: under 1 ms per company for the three structures; all ten re-priced together in about 3 ms.
Save to the study. Under Step 10 · Reports › Downloads › Whole batch press ★ Save all to study — thirty observations (ten firms × three structures). Measured: about 20 ms for the thirty observations.
Run Tables 1–10. Under Study open Study tables (Tables 1–5 and the funnel) and Regressions (Tables 6–10 with their Stata snippets and the checker; the robustness grid re-runs every observation three times). Measured: Tables 1–5 in about 4 ms; Tables 6–10 with the robustness re-runs in about 20 ms.
Export the Stata bundle. Under Regressions open ⬇ Download and choose Stata bundle: leverley_dataset.xlsx, tables_6_10.do, README.txt (and history.json when save history is on). Measured: about 0.1 s to build the zip (0.09 s in the page, 0.11 s until the download started).

Timings measured in the delivery test run on 14 October 2026 (headless Chromium 1440 × 900 on a Linux build machine, Playwright driving the page); a laptop will differ, usually within a factor of two or three.

Golden test against v1020

Rule for every change in v1021: with all new features off, the same inputs give the same IRR, MOIC, minimum ICR, covenant breach year and valuation as v1020. The delivery test loads the built-in sample (Sample CSV under Financial statements) into v1020 and v1021 at three financing-structure settings (the Small, Mid and Large presets — three different ladders) and compares, for all three structures, IRR, MOIC, min ICR, breach year, the five valuation multiples and every cell the workbook writer produces. Result: identical in every setting — 1,859 workbook cells per setting, three structures, no difference. v1022 (server mode, install steps, the Step 02 layout) changes no number either: the same comparison against v1021 is identical. v1023 adds the LBO analysis and new engine mechanics that are inert at their defaults, and v1024 replaces its catalogue with the sixteen deal types and adds their mechanics (also inert at their defaults): the delivery test re-ran the ten walkthrough companies at the Small, Mid and Large presets in both formula modes and compared IRR, MOIC, min ICR, breach year, default year, debt raised, sponsor equity, exit net debt, equity at exit, every cash balance, the covenant record and the flag list against v1022 — identical in all 180 runs for each build; the Monte Carlo paths, the attribution bridge, the valuation and the generated workbook (28,260 cells, only the engine stamp in the mechanics note differs) are identical too. The engine version moved to v1024 then because the engine priced new instruments and mechanics (seller notes, preferred equity, earn-outs, situations, exit fees, step-ups, operating paths; covenants by tranche, PIK toggles, dynamic recaps, sale-leasebacks, equity injections and debt draws, NOL and §382 limits); run IDs therefore change while every number stays. v1046 moves it to v1025: the DCF’s cost of equity takes a size premium and its terminal growth a cap at the risk-free rate, both off in the engine itself, and the DCF reports its terminal value as a multiple of year-5 EBITDA. With the risk-free basis on the domicile base rate, a size premium of 0 and the cap off, every valuation is v1045’s — the ten walkthrough companies at the three presets in both formula modes — and with the method on Manual, as shipped, no LBO figure moves either way. v1047 keeps engine v1025: while no company has valuation settings of its own, every figure is v1046’s — the ten walkthrough companies at the four size presets (Auto, Small, Mid, Large) in both formula modes, with the optional features off and on; the workbook’s Valuation sheet gains row 79, whose settings it used. v1048 keeps engine v1025 and changes navigation and layout only — the section menu, the order of 01’s pages, their buttons and the bars that stay or scroll — touching no pricing code. v1049 keeps engine v1025 too and changes layout only — the top bar scrolls away with the page, and the button to the next page stands at the right end of the toolbar of Operating metrics, Assumptions, Valuation and Scenarios, and under 16 structures → on the Deal model — touching no pricing code. v1050 keeps engine v1025 too and changes layout only — the toolbar’s buttons stand at its right, and the only line left between the bars and the page is the page bar’s — touching no pricing code. v1051 keeps engine v1025 too and changes layout only — a way back on every company page, Inputs folded to a rail on Operating metrics and Valuation, the Deal model’s buttons reordered and the company’s name without “Target” — touching no pricing code. v1052 keeps engine v1025 too and changes navigation and layout only — one ← Back on every page, the rail of Valuation titled Valuation settings and a section bar on Assumptions — touching no pricing code. v1053 keeps engine v1025 too and changes colour and the menu only — everything you can click in sienna, the section menu by its titles and light and dark as a switch — touching no pricing code. v1054 keeps engine v1025 too and changes colour only — in dark mode everything you can click is light blue — touching no pricing code. v1055 keeps engine v1025 too and changes layout and colour only, with one card switched off — Assumptions across the window, What if… in development and, in light mode, the fields, the method in use and the DCF grid's middle row in sienna — touching no pricing code. v1056 keeps engine v1025 too and changes navigation, layout and copy only — a vertical stepper of ten steps and Study, one header and one footer on every screen, the explanations behind ⓘ, the welcome page on one screen — touching no pricing code: the tape, the study rows and the workbook are those of v1055. v1057 keeps engine v1025 too and changes the welcome page and colour only — the welcome page of v1055 again, the filled buttons sienna in dark mode too, the titles of the steps and the bubbles in the text colour with their arrows in sienna — touching no pricing code. v1058 keeps engine v1025 too and changes colour only — in dark mode everything you can click is sienna again, as in light mode: the buttons (the outlined ones too), the links, drop-downs, switches and fields, and the chosen states such as the method in use — touching no pricing code. v1059 keeps engine v1025 too and changes colour only — in dark mode the ⓘ and their windows, the fields (the suggested ones too) and the drop-downs are light blue again, while every button and link stays sienna — touching no pricing code. v1060 keeps engine v1025 too and changes the screens only — Google’s Material Design 3 for the buttons, fields, drop-downs (one size everywhere), checkboxes, radio buttons, switches, segmented buttons, chips, cards, tooltips and dialogs; one content width on every step; the calls to action in a bar fixed to the foot of the window, Start over among them and no “Next:”; the header without its eyebrow, its line and its rule; Select company with radio buttons; every toggle of Method & risk settings a checkbox; the uploaded files listed with their delete buttons; the company drop-down; one ⓘ per explanation; the case the Excel workbook carries named and marked under Operating cases — touching no pricing code. v1061 keeps engine v1025 too: every operating case keeps the base case’s valuation and purchase price — an Upside or Downside workbook carries the base case’s Valuation sheet and entry multiple, where a valuation-driven entry multiple (DCF, VC) used to move with the case, as the Cases × structures table never did; the case the Excel workbook carries is shown on the Deal model and the results bar, priced as the workbook prices it but on copies of the inputs, so the inputs, the batch, the study rows and every other export stay on Base; and a workbook generated with Upside or Downside no longer leaves that case’s valuation in the company, so the study rows and the history version that Generate workbook and All models (.zip) save carry Base, as they always should have. Base-case workbooks and every other export are v1055’s, touching no pricing code.

Licences of inlined libraries

Library	Version	Licence	Used for
JSZip (with pako)	3.10.1	MIT (dual MIT / GPLv3; MIT applies)	reading the template, writing workbooks and every .zip; reading .docx
SheetJS Community Edition	0.18.5	Apache-2.0	reading .xlsx / .xls statements, writing analysis workbooks
TradingView Lightweight Charts™	5.2.1	Apache-2.0 (attribution logo on the chart)	the Market card’s price chart
pdf.js (Mozilla)	3.11.174	Apache-2.0	reading the text layer of PDF statements (v1021); the worker runs from a blob URL built from the inlined source
PptxGenJS	3.12.0	MIT	the PowerPoint exports of Compare (v1021) and of the 16 structures (v1032)
Manrope · JetBrains Mono	variable · 400/500	SIL Open Font License 1.1	the two typefaces, embedded

Everything else in the file — the engine, the parser, the workbook writer, the PDF writer and the statistics — is Leverley’s own code.

Exchange rates (v1045). The monthly history inside the file — US dollars per unit of each currency, January 1999 to 23 October 2026 — is built from the ECB’s euro foreign exchange reference rates (Source: ECB statistics; reused under the ESCB’s policy on the reuse of its statistics) and, for the Taiwan dollar, the Federal Reserve Board’s H.10 release (public domain), by leverley_fx.py, which also feeds a Leverley server’s /api/fx. The rates serve analysis: the ECB publishes them for information and discourages using them for transactions.

Long-term interest rates (v1046). The DCF’s blank risk-free rate is a ten-year government bond yield built into the file, by the statements’ currency — monthly averages for September 2026 (August for the Danish krone and the Czech koruna): the OECD’s long-term interest rates (Main Economic Indicators), read from FRED, the euro taking the German Bund; for the US dollar, the Federal Reserve Board’s H.15 ten-year Treasury constant-maturity yield (public domain). A newer build carries newer rates.

Leverley v1061 · engine v1024 ·

JavaScript is not running.
This usually means the file was opened in a document preview (e.g. the iOS Files "Quick Look" screen) instead of a browser.

Open this file in Safari or Chrome: share/open it with the browser app, or serve it from any link and open that link.

v1061 · offline

Financial statements

One or many companies in a single pick — each workbook becomes its own three-scenario model.

Drop statements here

xlsx · xls · csv Orbis (BvD) exports or any labelled P&L and balance sheet — .xlsx, .xls or .csv.Add more companies — same formats, new files join the batch.

One company per workbook Review mapping

Mapped 0 filesMapped from the statements:RevenueEBITDAD&ACapExNWCNet debtOne or many companies in a single pick — each workbook becomes its own three-scenario model.

Nothing uploaded yet.

Directory none loaded

A dated snapshot of listed companies, built on your own computer and read here offline. Every company keeps its source and as-of date; TradingView links are plain links that open a new tab. How to build one: Guide › Directory.

Find a company →Batch template…

Last twelve months LTM inputs: revenue · EBITDA · D&A · capex · NWC · net debt

Companies in this batch

Valuation settings method · hurdle · rates · WACC · discounts

Precedent transactions typed in · reference only

Sections

Operating assumptions multiples · growth · margins · tax

Financing scenarios bank · hybrid · private credit

Method & risk settings tax cap · ladder draws · optional features

Advanced formula fidelity

Debt-schedule formulas Hardened (default) caps each paydown at the tranche balance and applies amortisation, sweep %, OID fees, the tax floor and covenant tests. Faithful reproduces the original template exactly. Guide › Formula fidelity

Historical statements the mapped lines by period

Market — listed company

Findings & provenance

Valuation — five methods

DCF grid WACC ± 1 pt × terminal growth ± 0.5 pt

Valuation range

Trading comparables

Sources & uses at close, three ladders

Sponsor returns equity in · equity out · MOIC · IRR

Debt schedule by tranche and year

Operating model revenue to levered free cash flow

Scenario landscape x leverage · y IRR · area sponsor equity

Operating cases base · upside · downside

What if… In development

Downside — Monte Carlo

Settings · paths · seed · shock switches

Sensitivities IRR and MOIC grids, re-run for every cell

Return bridge equity in → EBITDA · multiple · debt · fees → equity out

Debt repayment net leverage and cover by year

Summary for the committee rule-based, every figure from the engine

Valuation summary the five methods and the three ladders at a glance

Downloads this company · the whole batch

This company

The three-scenario Excel workbook with live formulas: Cover, BankLBO, Syndicate LBO, PC Fund LBO, Model Methods, Valuation and, when switched on, Sensitivities. Generating also saves the company to the study.

Three observations, one per financing structure, into the study under Study.

One page for the committee: the results tape, covenant headroom, Monte Carlo quantiles, the valuation bridge, the sourced assumptions and the rule-based summary. A4, or US Letter from the optional features.

Every assumption of the run with its value, low, high, source and date.

The run’s inputs and results as JSON, for anything downstream.

The sixteen deal types as a workbook of engine values, one PDF, PowerPoint slides or JSON, from that page’s Download menu.

Whole batch every priced company

One .zip with a three-scenario workbook per priced company, and history.json when save history is on; every company is saved to the study too.

Every priced company in the batch, three observations each.

The dataset fingerprint, the screening funnel and exclusion log, the All-industries comparison and every priced company’s IRR, MOIC and breach year per structure.

Save history versions of this company

This company

Building the three models

Preparing…

Who it’s for

What it does

How it works

Security and privacy

Start with the built-in template.

What it does

Five-method valuation

Sixteen LBO structures

Covenant tests and headroom

Live-formula Excel workbook

Batch study and Stata bundle

One file, offline by default

Who it’s for

Financing consultant

Private-credit analyst

Finance researcher

How it works

Data · steps 1–4

Model · step 5, Valuation

Model · steps 6–7, Deal model and 16 structures

Analyse · steps 8–9, and Study

Analyse · step 10, Reports

Security and privacy

Nothing, unless you switch lookup on

Company names, to this server only

In this browser, on this computer

One button in the Guide

Libraries inlined, nothing fetched

Five-method valuation

Sixteen LBO structures

Covenant tests and headroom

Live-formula Excel workbook

Batch study and Stata bundle

One file, offline by default

Loading the built-in LBO template…

Operating assumptions multiples · growth · margins · tax

Financing scenarios bank · hybrid · private credit

Method & risk settings tax cap · ladders · Monte Carlo · valuation

Drop statements here

No observations yet

No observations yet

Nothing to compare yet

Nothing to analyse yet

This company

Whole batch every priced company

No company is open