Project Caldbeck — 60-Minute Take-Private LBO

The plan is handed over from Adjusted EBITDA down and there is nothing to get wrong in it. Everything expensive happens before the first projected column: the premium and the share count, the treatment of existing debt, the cash on the balance sheet, the fee split, the appraisal reserve. Four of the five mistakes this case is built around live above the first year of the model, which is the opposite of where a candidate who has practiced private-company LBOs expects to find them. Budget the clock accordingly — the offer and the sources and uses are worth twelve of the sixty minutes between them, and none of that time is spent on projections.

When an offer is a premium to a traded price, the diluted count is not the count on the cover of the last filing. Options struck below the offer are all in the money, so all of them are exercised, and the strike proceeds the company is deemed to receive buy shares back — at the offer price, because that is what a share costs on the day the deal closes and the price the options are actually cashed out at. Running the method at the unaffected close instead is the quiet version of this error: it uses the right method and the wrong price, and it repurchases more shares than the proceeds actually buy. Nothing on the page looks wrong afterward. The tell is that the deemed repurchase price and the price the options are cashed out at are two different numbers sitting in the same model.

A restricted stock unit is a promise of a share, not a right to buy one. There is no exercise price, so there are no proceeds, so the treasury stock method has nothing to repurchase with and every unit counts in full. That is one sentence of theory and it is worth real money on the first tab, because units are the line candidates skip: they read the option table, run the method properly, and never scroll to the row underneath it. On a company with a live equity compensation program the units are always there, and their whole count lands in the price the acquirer pays.

A private-company LBO prices the enterprise value off a multiple and derives the equity. A take-private runs the bridge the other way: the offer per share times the fully diluted count is the equity purchase price, and the enterprise value is that plus existing gross debt less balance-sheet cash. The entry multiple is then something the model REPORTS rather than something it is told, which means it is a check on the price rather than an input to it. Candidates who have only built the private-company direction reach for the multiple first, and there is no multiple to reach for.

A term loan is repaid at par and there is nothing to think about. Bonds are different: an indenture typically gives holders a change-of-control put, here at 101, so the merger triggers a right to be taken out at a premium to face. Two consequences follow and both are easy to miss. The redemption is a use of cash at the put price, not at face. And the difference between the two is a transaction cost rather than part of the enterprise value, which is struck at face — so the same instrument appears at two different amounts in two different places in the model, correctly. This is the only error in the case that is a reading error rather than a mechanical one: the put is a line on the capitalization page and nothing in the workbook prompts you for it.

The buyer acquires the target's cash along with everything else, and can use it to fund the purchase. What it cannot use is the cash the business needs to keep running, so the split is between the operating minimum — which is funded straight back onto the closing balance sheet and becomes the opening cash balance — and everything above it, which is a genuine source. Leave the whole balance out of the sources column and the entire amount lands on the sponsor's check for exactly the same asset. This is the one error in the case that makes the deal look WORSE: a candidate who makes it can reach a defensible-sounding answer off a broken model, and reaching the right answer for the wrong reason is not the same as being right.

In any cash merger, stockholders who dissent can petition a court to determine the fair value of their shares instead of taking the merger consideration. Some fraction perfects that right, and the reserve against a potential uplift over the offer price is real money that has to be funded at close. The case prices it rather than assuming it away, and the reserve appears in uses alongside the fees and the call premium. A model that omits it has an entry that does not tie, and a candidate who has never seen it will not think to ask.

The two halves of the fee load are treated differently. Financing fees are a cost of putting the debt in place, so they are capitalized and amortized over the life of the facility — which means they reduce EBIT and taxable income a little every year, generating a real cash tax shield, and then come straight back above the working-capital line because they are non-cash. Advisory fees are a cost of doing the transaction, expensed at close, in the year before the projections start. Two errors follow from getting this wrong and the second is worse than the first: expensing the whole load at close throws away the shield, and putting the advisory fee into the first projected year instead of the close year is a period error on top of a treatment error.

Most quick models hold the base rate flat and hope nobody asks. Here the base rate is swapped to a fixed rate for the full five years at close, which is why there is no rate curve, no floor and no forward strip in the model: the hedge is part of the capital structure rather than a simplification taken for time. That is a one-sentence answer to the most predictable follow-up in the room. Add the swapped base rate to each spread and you have the all-in cost of each floating tranche; the notes are fixed and are not swapped.

This sweep is a step function of net leverage rather than a constant percentage: a high percentage while leverage is above the top threshold, a lower one in the middle band, and nothing below the bottom one. Which means a deleveraging company sweeps less as the hold goes on, and the cash it stops sweeping does not disappear — it accumulates on the balance sheet, where it still reduces net debt. Two mechanics matter here. The step is measured off the PRIOR year-end net leverage, which is what keeps the schedule sequential: read it off the current year and the sweep depends on the leverage that the sweep itself determines, and you have rebuilt the circularity that interest-on-beginning-balances was designed to remove. And a model whose sweep percentage is flat runs perfectly well, throws no error, and gets the entire back half of the debt schedule wrong.

Bank debt is prepayable at the borrower's option; high-yield notes usually are not, at least not without a make-whole, and the term sheet's prepayable column says so in one word. So the cascade can only ever reach the facilities the documentation allows it to reach, and sweeping the notes breaches the agreement the sponsor has just signed. Build the sweep once, multiply it by the prepayable flag, and copy it. That way the term sheet decides where the money goes and anyone reading the file can see that it did. Write three bespoke formulas instead and you have hard-coded your reading of the documentation somewhere nobody can find it.

A post-close option pool does two things and they are not symmetric. Management pays a strike price IN, which enlarges the equity pool before it is divided; management takes a percentage OUT, which is deducted after the division. Do both in the right order and the cost to the sponsor is the share taken out net of what was paid in. Ignore the pool entirely and the return improves, because a slice of the equity has been handed back to the sponsor. Take the dilution but forget the proceeds and the return worsens, because management has been charged for shares it was never credited with buying. The strike here is the sponsor's own entry price per unit, which means you have to work out how many units the sponsor's check bought before you can price the option — an ordering that catches people who reach for the exit price instead.

Any valuation running off an earnings multiple has to say what is in the earnings, and on a public company that is never a zero-length answer. Two questions do the work. First, is stock-based compensation charged or added back? It is charged here, which means every multiple and every leverage figure in the case is struck on a base that already bears the cost of paying management in equity — the conservative choice, and the flattering alternative is available to anyone who wants to quote it. Second, is each add-back something that has happened or something forecast to happen? The public-company costs eliminated on a take-private are a PRO FORMA add-back: nothing has been saved yet, and the saving arrives only because the company stops being listed. It is customary and it is defensible, and it is also the one line on the reconciliation where a candidate who strikes it out is paying a visibly different multiple. Where the whole basket sits inside the customary 25% cap a credit agreement applies, it is a disclosure point rather than a covenant point — but it is still a point, and it costs four seconds to make.

Interest coverage — EBITDA over interest — is the ratio everyone quotes and it is the wrong one for this company. Capital expenditure runs at close to thirty percent of Adjusted EBITDA here and never falls far below it, so a credit committee sizes the structure on fixed-charge coverage: EBITDA less capital expenditure less cash taxes, over interest. The general rule generalizes past this case. When a company's defining operating feature is X, no coverage statistic that ignores X may stand alone. Quote both, name which one the counterparty actually underwrites to, and compare it against a customary maintenance covenant rather than against nothing.

Total gain decomposes into growth in earnings, expansion in the exit multiple, and deleveraging and cash generation — plus, in a structure with a pool, what leaks to management, and minus the transaction costs funded before the hold begins. Growth is the change in Adjusted EBITDA capitalized at the ENTRY multiple; multiple expansion is the change in multiple applied to EXIT-year earnings; the remainder is the change in net debt. Holding exit equal to entry makes the middle term zero by construction, and saying that out loud is the point: none of the return depends on the market re-rating the asset, which is the honest way to underwrite and the thing that makes the rest of the bridge attributable. The transaction-cost line at the front of the bridge is the one nobody expects to matter and it is not small — every dollar of fees, call premium and appraisal reserve is a dollar of the sponsor's own equity that leaves the deal on day one and compounds at nothing.

'What does it return?' and 'what would you pay?' are different questions with different answers, and the second one is the follow-up to any returns number that lands anywhere near a hurdle. Holding the exit multiple equal to whatever entry multiple the price implies, and the debt quantum fixed at the term sheet, solve for the highest offer price that still returns the hurdle. That converts a readout into a negotiating position and it is a far more useful sentence than the return itself. Be ready for the structural version of the same question too: how much more leverage would close a gap, and whether the commitment letter would actually fund it — because a minimum equity contribution is a hard constraint, and an answer that reaches past it is not an answer.

It is, if the structure is already familiar — which is what the format is testing. The template hands over every label, column header, number format and print setting, plus the whole capitalization page, the whole term sheet, the sweep grid and the whole management plan, so nothing is spent on layout or on typing inputs. The measured workload is 477 cells, which is 175 distinct formulas once the fill-right and copy-down groups are collapsed: almost every row in the operating build, the cash flow and the debt schedule is written once in the first projected column and filled right across four more years, the grid is one formula copied across its cells, and the attribution bridge is one column copied down. That is about 21 seconds per authored formula against 8 seconds per raw cell. Both numbers are measured from the file rather than estimated, and both are printed on the template Cover so neither can be presented as the whole story.

Four minutes reading the capitalization page and the term sheet before you type anything. Six on the offer and the share count, six on sources and uses, six on the operating model, six on the cash flow. Fourteen on the debt schedule, which is more than a quarter of the cells and the only tab where the clock really bites. Four on the returns, four on the attribution, seven on the sensitivities. Three at the end to write the recommendation. That is sixty minutes across all eight build tabs plus the one thing you hand back that is not a cell, and it is the same clock printed on the template Cover. The Assumptions tab needs none of it: it is given in full and nothing is cleared from it.

Because a sponsor's first pass at a deal does not have one either, and because an hour that builds a balance sheet never reaches the returns. EBITDA-down is the defining scope of this tier: the plan is handed over at the Adjusted EBITDA level and all the work is below it. That constraint is also why there is no purchase price allocation, no goodwill, no asset write-up and no deferred tax liability anywhere in the case — all three require the balance sheet this model does not carry. If you find yourself laying one out, you have started a two-hour exercise inside a sixty-minute one. Naming what the model leaves out, and why, is worth more than half-building it.

The entry, and nothing else. A private-company buyout prices the enterprise value off a multiple and derives everything from there. A take-private starts from a share price and a premium, so you have to build a fully diluted share count on the treasury stock method, work out what happens to existing debt that may not come off at par, decide how much of the balance-sheet cash is available, reserve for appraisal rights, split the fees between the half that is capitalized and the half that is expensed, and only then reach the sources and uses that the other exercise starts with. The bridge also runs the other way: the equity price is the given and the enterprise value is the output, which means the entry multiple is a check on the price rather than an input to it.

The offer price. The options are being cashed out in the merger at the offer, so that is the price at which the deemed proceeds repurchase shares. Using the unaffected close is the single most common way to get this line subtly wrong: the method is right, the arithmetic is right, and the price is the wrong one, so the count comes out light and nothing on the page indicates it. If you want a check that catches it, foot the equity purchase price the other way — basic shares at the offer, plus the intrinsic value of the options at the offer, plus the units at the offer. The two routes agree only when the repurchase price and the cash-out price are the same number.

Beginning, in this exercise, and the instruction sheet says so. Average-balance interest is more precise and is common in practice, but it makes the model circular and requires iterative calculation to resolve. Beginning balances remove the circularity by construction, which is why every check row in the file can be trusted and why there is no circularity toggle anywhere in the workbook. The same design decision is why the sweep steps off the PRIOR year-end leverage rather than the current year's — same principle, applied twice. If you are used to building it the other way, say so out loud and say what the convention costs; knowing why a simplification exists is worth more than the simplification itself.

Cash available for financing after mandatory amortization, multiplied by the percentage the term sheet's grid assigns to the prior year-end net leverage band. It then cascades in documented priority — the revolver first, then the term loan B — and it is multiplied by the prepayable flag, which is what keeps it away from the tranche the indenture protects. Bound it at both ends while you are there: floored at zero, and capped at the balance that survives mandatory amortization, so a tranche cannot be repaid below zero and whatever the cap leaves over cascades onward rather than evaporating. Whatever is not swept stays in cash, where it still reduces net debt.

As two different things, because they are two different things. The financing fees are capitalized at close and amortized straight-line over the hold: they reduce EBIT every year, which is a real cash tax saving, and because the charge is non-cash it is added straight back in the cash flow above the working-capital line, exactly as depreciation is. The advisory fees are expensed at close, in the year before the first projected column, and never touch the projections at all. The two errors to avoid are expensing everything at close, which throws away the shield, and — worse — putting the advisory fee into the first projected year, which is a period error stacked on a treatment error.

No. There is a single equity outflow at close and a single inflow at exit, so the return is the multiple of money raised to the power of one over the hold period, less one. Neither IRR nor XIRR nor NPV appears anywhere in the model or the template, and discount factors, where the model needs them, are chained by hand. It is also the safer habit under a clock: a closed form cannot pick up an extra cell unnoticed, and it is far easier to sanity-check in your head. Carry the market's own rule of thumb while you are at it — a multiple of money above 6.0x over five years implies a return above 40% and almost always means an arithmetic error somewhere upstream.

It is explicitly graded here, as it is on real modeling tests. Blue for a hardcoded input, black for a formula on the same tab, green for a pure link to another tab. A hardcoded number buried inside a formula is marked down even when the number is right, because the next person to open the file cannot see it. Build one related habit before you sit the test: pin every reference to a single-cell assumption with dollar signs. An unpinned scalar cannot be filled right at all — dragging it across four columns walks onto empty cells — so a model that fails this does not merely look untidy, it has to be retyped five times inside an hour that has no five minutes in it.

Cut the sensitivity grid, and cut it before anything else. Do not cut the check rows, do not cut the option pool, and do not cut the returns attribution — the attribution tab is twenty-three cells and carries more marks than the hundred and nine in the grid. Above all, do not cut the recommendation. A model that ties, with an attribution bridge and a few sentences on where the return came from and what would have to change, is a better answer than a full grid sitting on top of a return nobody can explain. The clock printed on the template puts the grid last because that is where the marks are thinnest per cell.

Some form of 'what would you pay?' or 'what would have to be true?'. The first is answered by holding the return at the hurdle and solving back for the offer price, which turns a readout into a negotiating position. The second is answered from the attribution bridge: name the source of value that dominates, say what has to be true for it to arrive, and say what happens if it does not. A third is worth preparing because it sounds like a lifeline and is not — whether more leverage would close a gap. The commitment letter is conditioned on a minimum equity contribution, so there is a ceiling on that lever, and knowing where it binds is a better answer than reaching past it.

Constantly. The sixty-minute take-private is the standard second-round modeling screen where the private equity ladder stops being arithmetic and starts being modeling: the first tier that makes the candidate build an offer, a diluted share count and a real sources and uses before any of the familiar steps start. It is cheap to administer, it fits inside an on-site session, and it is very hard to fake — a share count either uses the right price or it does not, and a sweep either respects its step, its cascade and its prepayable column or it does not. Thirty seconds of scrolling tells an interviewer which.