Project Puma

Project Puma was my capstone for the Master of Finance program at Penn State. It also happened to be a live deal on a real "company." That is what changed the texture of everything. The model was still there, the research was still there, but the center of gravity moved from completing an assignment to forming a view that could be defended in front of people who actually understood the business.

The most valuable part was learning how quickly clean finance breaks against messy reality. A model can make a company look legible too early. Diligence slows that down. It forces the obvious questions, then the uncomfortable ones, then the ones that only show up after management explains how the business actually works.

The deal details stay private. The point of this page is the experience: the research, the judgment, and the shift from classroom finance to a real transaction with real stakes.

Workflow

Building an APV Model for a Private Company

Why Private Companies Are Harder

Valuing a public company is already more work than people think. Beta is not pulled. Beta is calculated. You run a regression of the stock's returns against the market's returns and the slope of that line is the equity beta. Even then, the answer describes the company as it is financed today, not as it operates. Private companies push the problem further. No traded stock, so no returns to regress. No market price to sanity-check the answer against. When the target has an unusual capital structure, low debt, minimal taxes, a bunch of non-operating assets, a standard WACC-based DCF starts hiding more than it shows. That is why I went with APV.

The APV Framework

APV stands for Adjusted Present Value. The idea is to split the value of the business from the value of how it is financed, instead of blending them into one discount rate.

The equation

APV = Unlevered Firm Value + PV(Tax Shields)

Unlevered Firm Value = Σ [ UFCFₜ / (1 + rₐ)ᵗ ] + TV / (1 + rₐ)ⁿ

The unlevered firm value is what the business would be worth with zero debt. The tax shields are what debt adds on top, because interest is tax-deductible and that deduction has real economic value. Splitting the two makes the reasoning transparent. When the target barely pays taxes, the tax shield is nearly worthless, and APV shows it plainly. WACC would bury it inside a blended rate.

Unlevered Free Cash Flow

To value the unlevered firm you need unlevered free cash flow. That is the cash the business generates before any financing decisions.

UFCF = EBIT × (1 - t) + D&A - Capex - ΔNWC

Each piece earns its keep. EBIT × (1 - t) is operating profit after taxes as if there was no debt. D&A gets added back because it is a non-cash expense, not real money leaving the business. Capex is real money leaving the business to buy assets. ΔNWC is the cash tied up in working capital, inventory, receivables, payables. Interest is deliberately missing. APV values that separately through the tax shield.

The Beta Problem

Private companies do not trade, so there is no stock return series to regress. You build a beta from the outside in. The approach is called a bottom-up beta. Find five public comps in adjacent industrial segments, calculate each one's equity beta, unlever each to strip out its capital structure, average the asset betas, then price your discount rates off that industry-level operating risk measure.

Step 1. Calculate the raw equity beta for each comp

Beta is the slope coefficient from a regression of the comp's stock returns against a market index's returns. I ran the regressions myself, five-year monthly returns against the S&P 500, rather than using a published figure. Published betas use different windows, different indices, and sometimes different smoothing adjustments. Running them yourself means every comp is on the same basis.

β_equity = Cov(R_stock, R_market) / Var(R_market)

In Excel you can do this directly with SLOPE(stock_returns, market_returns) once you have the return series lined up. That is the equity beta. It reflects both the business risk and the financing risk of the comp.

Step 2. Pull the inputs you need from the 10-K

To unlever a comp's equity beta you need two things from its 10-K: its capital structure and its effective tax rate. Both are there if you know where to look.

Debt:    Balance Sheet → Long-Term Debt + Current Portion of LT Debt
         + Short-Term Borrowings + Capital Lease Obligations
         (use book value if market value of debt is not disclosed)

Equity:  Market Cap = Shares Outstanding × Share Price
         Shares Outstanding is on the cover page of the 10-K
         Price is as of the valuation date, not the filing date

Tax Rate: Income Statement or Tax Footnote
          Effective tax rate = Tax Expense / Pre-Tax Income
          Average the last 3 years to smooth out one-time items

Use market value of equity, not book value of equity. Market cap is what the market thinks the equity is worth right now. Book equity is an accounting artifact. For debt, market value is cleaner if the comp has public bonds. Most of the time you will end up using book value of debt as a reasonable proxy, because short-dated corporate debt trades close to par.

Step 3. Unlever each comp's equity beta

β_asset = β_equity / (1 + D/E)

This is the simplified unlevering formula. No tax term. I picked it on purpose. It is mathematically consistent with discounting tax shields at rₐ (cost of assets) instead of r_d (cost of debt). The other option is the Hamada formula, which includes a tax term.

β_asset = β_equity / [ 1 + (1 - t) × D/E ]

If you use Hamada, you commit to discounting tax shields at r_d. If you use the simplified version, you commit to rₐ. The two approaches give different answers and mixing them breaks the math. Most students do not know the choice matters. Most professionals forget to check. I stayed simplified and kept the tax shield discounting clean.

Step 4. Average and apply

Average the five asset betas. That is your estimate of the industry's operating risk, with capital structure noise stripped out. For a private target with minimal debt, the asset beta basically is the equity beta because there is almost nothing to relever. For a target with real leverage, you relever the industry asset beta at the target's D/E to get the equity beta used in CAPM.

CAPM and Terminal Value

Once you have the asset beta, CAPM turns it into a discount rate.

r_a = R_f + β_asset × MRP
r_d = R_f + β_debt × MRP

R_f is the risk-free rate. I used the 10-year Treasury. MRP is the market risk premium, a standard textbook value. For the terminal value at the end of the explicit forecast I used Gordon Growth.

TV = FCF_terminal × (1 + g) / (r_a - g)

The growth rate g has to sit below the long-run growth of the economy, otherwise the business eventually becomes bigger than the economy, which is impossible. I used 2%.

Mid-Year Convention

One detail that matters more than it should. Cash flows do not arrive at year-end. They show up throughout the year. The discounting should reflect that. Mid-year convention shifts each period by half a year.

Discount periods: 0.5, 1.5, 2.5, 3.5, 4.5
PV Factor = 1 / (1 + r_a)^(t - 0.5)

The terminal value sits at year 5 but gets discounted at 4.5 years, because it represents perpetual cash flows starting right after year 5. Sounds minor. It adds roughly 5% to valuation compared to year-end discounting. On a multi-million dollar deal that is real money.

Tax Shields and the Equity Bridge

Tax shields get valued separately and added on top.

Annual Tax Shield = Interest Expense × Tax Rate
Interest Expense = Debt Balance × r_d
PV(Tax Shields) = Σ [ Shield_t / (1 + r_a)^t ]

Discounting at r_a (not r_d) because the unlevering formula I picked earlier commits me to it. Then to go from enterprise value to equity value, you bridge.

Equity Value = APV EV + Non-Operating Assets - Net Debt
Net Debt = Total Debt - Cash & Cash Equivalents

Non-operating assets are things the business owns that do not show up in the cash flow projection. Real property the business does not use for operations, liquid investments, anything that has market value but is not generating the operating cash flows you just modeled. You add those back at the end.

How Due Diligence Changed the Model

The Setup

A model is only as good as its assumptions. Before talking to management, ours ran on textbook defaults. Federal statutory tax rate. Conservative growth. Only the obvious balance sheet items counted as cash. All defensible, all wrong for this specific business.

After one round of structured questions, five assumptions changed. The standalone equity value moved by double digits.

Non-Recurring Income

A large line item in Other Income did not exist the year before. Management confirmed it was a one-time government refund tied to employee retention during COVID. Without stripping it out, reported net income looked materially higher than the true recurring level, enough that any valuation built on P/E would have been wildly off.

The fix is not a formula correction. It is a definition correction.

Adjusted NI = Reported NI - Non-Recurring Items

Then you feed adjusted NI, not reported NI, into anything multiple-based. This is not about being conservative. It is about making sure the number you are multiplying by a market multiple actually represents what the business will earn year after year.

Asset Liquidity

The balance sheet showed an investment account that looked like a long-term holding. Management confirmed it was fully liquid within 72 hours. That changes how you classify it. If it is liquid, it belongs in cash. Moving it cut net debt by more than half.

Net Debt = Total Debt - Cash & Cash Equivalents

Smaller net debt means bigger equity value. That is a lot to leave on the table because you assumed a balance sheet line was what it looked like.

Revenue Pipeline

Management walked through three specific product development initiatives. Named end markets, projected revenue, estimated margins. That changed the growth assumption from flat (what last year's financials implied) into something grounded in what the business is actually trying to do. When you have real pipeline tied to real initiatives, you stop guessing and start modeling.

Cost Structure and Synergies

Management also identified specific overhead an acquirer could eliminate. Not vague "synergies." Named functions, named cost lines. That gave the synergy model real grounding.

Annual Synergy CF
  = Revenue × (ΔCOGS margin + ΔSGA margin) × Ramp Factor
  + New Revenue × Contribution Margin × Ramp Factor

The ramp factor is where people get sloppy. Synergies do not show up on day one. Year 1 is often negative because integration has real costs. Year 2 drifts toward breakeven. Years 3 through 5 scale up toward full realization.

Year 1: negative (integration costs)
Year 2: breakeven
Years 3-5: scaling to 100%

Treating synergies as immediate is one of the fastest ways to overpay for a deal. They show up after you have already paid to integrate, not before.

What the Conversation Actually Did

The model didn't change because we got better at Excel. It changed because we asked the right questions and knew where to look in the financials for things that did not fit. Due diligence is not about confirming what you already believe. It is about finding the three or four facts that make your existing model wrong.

Two Models, One Target

Why We Built Two

This was a two-person project. We each built the model independently, then sat down to reconcile. On paper that sounds inefficient. In practice, it is the best quality control you can run on a valuation. Two people, same financials, same framework, same target. Different assumptions, different cell references, different small decisions. When the two models disagree, you learn a lot about what each of you actually decided to do.

Both of our models worked. No #REF! errors. No circular references. No broken formulas. Every cell had a number. Both looked clean.

We still found four places where they did not match.

Referencing the Wrong Line

One version pulled revenue projections from gross sales. The other pulled from net revenue. Those are different lines on the income statement. Gross sales is before deductions for freight, discounts, and returns. Net revenue is what actually hits the top of the business. The difference was a few percentage points. Small. Compounded over five years plus a terminal value, small becomes real.

The formula also had a growth issue. Adding a growth rate as a dollar amount instead of multiplying. The two errors partially canceled each other out, which made the output look reasonable. Nothing in the final number told us anything was wrong. Tracing the formula back to the source was what revealed it.

Sign Convention on Working Capital

One version added the change in net working capital to free cash flow. The other subtracted it. Only one is right. When NWC increases, cash is tied up. Inventory grew. Receivables grew. Payables didn't keep pace. That is real cash leaving the business. It should reduce FCF.

NWC = Accounts Receivable + Inventory - Accounts Payable
ΔNWC = NWCₜ - NWCₜ₋₁

ΔNWC > 0  →  cash tied up   →  subtracted from FCF
ΔNWC < 0  →  cash released  →  added to FCF

The annual impact looked small in isolation. The kind of number you glance at and assume is fine. Over five years plus terminal value, it is not fine.

Tax Rate Assumption

One version used the federal statutory corporate rate. The other used something closer to the target's actual effective rate. The target is a pass-through entity. It pays minimal corporate-level tax. That single assumption moved after-tax operating income by a meaningful amount every year.

UFCF = EBIT × (1 - t) + D&A - Capex - ΔNWC

The tax rate sits inside the most important line. Using the wrong one doesn't show up as an error. It just shows up as a different number at the bottom.

Non-Recurring Items in the Earnings Base

Both versions used reported net income for the P/E cross-check. Reported NI included the one-time government refund mentioned in the previous section. The implied P/E valuations came out 7 to 8 times too high. This wasn't a formula correction. The formula was right. The input was wrong. The fix was adding an adjusted recurring earnings line that stripped out the non-recurring item.

Adjusted NI = Reported NI - Non-Recurring Items

Then feeding adjusted NI into the P/E valuation instead of reported NI.

What a Team Actually Catches

The real lesson from the reconciliation was how fragile a finished model looks. Both of ours looked done. Both produced valuations. Both would have been presented with confidence if nobody pushed back.

Team work does not mean two people building one thing together. It means two people each building something, then sitting down and asking why they do not match. The most dangerous errors in finance are not the ones that break a spreadsheet. They are the ones that produce plausible-looking numbers. You do not catch those by looking at a model. You catch them by explaining a model to someone who built their own version and did not make the same choices.

Three Buyers, Three Stories

Why the Standalone Number Is Not the Deal

A standalone valuation tells you what the business is worth on its own, run the way it is being run today. That number is useful, but it is not the deal. The deal is what the business is worth to a specific acquirer, because the synergies sit on top of the standalone value, and synergies are acquirer-specific. The same target is worth different numbers to different buyers. The work is figuring out how much different, and why.

Value to Acquirer = Standalone Value + PV(Synergies, acquirer-specific)

The Three Archetypes

We built acquirer-specific cases for three buyer profiles. No names here, just archetypes, because the shape of the logic travels.

Buyer A, the consolidator. A larger player in the same operating space. Synergies come mostly from cost. Overlapping overhead, shared back office, negotiating leverage with suppliers. Revenue synergies exist but are secondary. The integration risk is moderate because the two businesses do similar things the same way.

Buyer B, the adjacent platform. A buyer operating in a neighboring space where the target fills a product or capability gap. Synergies skew revenue. Cross-sell into an existing customer base. Bundling. The cost synergies are real but smaller, because there is less direct overlap. Integration risk is higher because two different operating models have to learn to work together.

Buyer C, the financial buyer with a platform thesis. A sponsor with a portfolio company in an adjacent vertical looking to tuck in. Synergies are a blend, with an added financing angle because leverage assumptions shift the returns math. The standalone quality of the business matters more here, because there is no strategic buyer absorbing operational weakness.

Max Price Logic

For each buyer, the max price is the standalone value plus the present value of their specific synergies, adjusted for integration risk and the share of synergy value a rational seller would expect to capture.

Max Price = Standalone Value + PV(Synergies) × Capture %
            − Integration Risk Adjustment

The capture percentage is a judgment call. Synergies are created by the buyer, but the seller negotiates for a share of the value. In competitive processes more of that value leaks to the seller. In bilateral negotiations the buyer keeps more. That one lever changes what each archetype can afford to pay.

What the Comparison Produced

The three buyers did not line up in one order. The consolidator had the cleanest cost story but limited upside past that. The adjacent platform had a bigger range, because revenue synergies are harder to size and more sensitive to assumptions. The financial buyer's number depended heavily on financing structure and how the existing portfolio company was valued.

The recommendation was not a single number. It was a ranking, a reasoning, and a band. Which buyer had the structural right to pay the most, which one was most likely to actually execute, and where the deal should land if each side negotiated from a defensible position. That is the part the standalone valuation cannot tell you.

Research Frame

A good model does not replace judgment. It reveals whether the judgment underneath it is coherent.

What I took from it

The main lesson was that valuation is a conversation between mechanics and reality. The mechanics matter. They impose discipline. But they only become useful when the business is understood well enough for the assumptions to carry weight.

This project forced the work closer to the machinery: imperfect information, moving assumptions, operational detail, buyer incentives, and the need to explain a view without hiding behind the spreadsheet. That is the part worth sharing.

It is also the part I want to keep doing. This is the kind of work I am building toward: live deals, real businesses, the space where the model meets the operating reality. Strategic finance, corporate development, PE portfolio operations. The work that sits between the spreadsheet and the decision.