Real Options Analysis (ROA) for Biotech Valuation & Portfolio Management

Executive Summary

Valuing biotechnology ventures is an exercise in storytelling and probability rolled into one: a molecule's arc from discovery to approval is defined by research signals that arrive in fits and starts, regulatory checkpoints that can stop or accelerate progress, and commercial contingencies that amplify small scientific advantages into large market rewards. Standard discounted-cashflow methods capture expected payoffs but, by design, treat managerial choice as passive. Real Options Analysis (ROA) changes that frame. It recognizes the value of managerial choice and makes it quantifiable: the choice to defer, expand, abandon or stage a program is an option whose value depends on the stochastic evolution of the project's underlying commercial worth. For investors and CEOs who must allocate scarce capital across pipelines, ROA offers a disciplined way to translate uncertainty and flexibility into actionable funding rules.

The first half of this article explains why ROA matters for biotech, what the core models are, and how staged development maps naturally to compound options. The second half is a practitioner's workshop: precise implementation recipes, numerical architectures (binomial/recombining lattices, trinomial variants, Monte Carlo with Longstaff–Schwartz regression), and worked example templates.

1. Why Real Options Change the Valuation Conversation

Imagine two drug candidates that appear identical when aggregated into a single expected cashflow stream. One has a clear biomarker-driven Phase II readout that will rapidly reveal efficacy; the other relies on long, expensive, late-stage outcomes. A classic rNPV treats both as equivalent if their expected cashflows coincide. In reality, however, the first candidate lets management make an informed go/no-go choice after a low-cost learning investment; the second does not. By treating the ability to stop or continue as an actual financial quantifiable option, ROA often assigns a higher decision value to early, well-instrumented projects. This premium exists because you have valuable flexibility in the face of uncertainty. You can make a small initial bet to gather information, and then decide later whether to make the full investment. This option to either capitalize on good news or walk away from bad news is what creates mathematical convexity in payoffs under uncertainty and therefore, the extra value. Taking it together, such reframing has three major practical consequences.

• ROA identifies where additional small investments in information (biomarkers, pilot studies) have high value because they unlock optionality.
• ROA changes stage sequencing and tranche design by making explicit how holding back capital reserves and setting milestone triggers create or destroy option value.
• At portfolio scale ROA alters ranking: programs that look subscale under rNPV can become high-priority if they generate platform optionality (extension into multiple indications, licensing pathways, or partnership options). And that is a key difference when valuating early-stage companies as conventional rNPV methodologies can heavily undervalue their potential (Figure 1).

FIGURE 1: Flexibility Value as % of Base NPV

This graph shows how the option premium (flexibility value) increases with volatility, demonstrating that ROA captures additional value from managerial flexibility. As volatility rises from 10% to 80%, the option premium can exceed 40% of base NPV, illustrating why high-uncertainty biotech projects benefit most from real options analysis.

2. The Mathematical Foundation Step-by-Step

Before we present equations, it is useful to clarify modeling choices, their economic meaning, and the tradeoffs they imply. Valuation models live on a spectrum of realism and tractability. At one end are continuous-time models that yield transparent comparative statics and closed‑form thresholds; at the other end are fully numerical, high‑dimensional simulations that can capture jump risks, path dependence and rich decision sets. Choosing the right model is a pragmatic decision based on the project's features: the number of stochastic drivers, the discreteness of readouts, the finite horizon imposed by patents, and the ability (or inability) to find market proxies for hedging.

Conceptual mapping: what the option valuation model pieces represent

• The underlying S(t) is the present value of future post‑launch cashflows conditional on success. Economically, it bundles epidemiology, price, market share and net margins into a single stochastic state variable whose evolution reflects market and competitive dynamics; conceptually S is what the firm stands to gain if the drug is ultimately successful.
• The cost to exercise (I) corresponds to the next tranche of development investment; in practice this is the expected sunk amount committed to run a trial or fund development to the next decision node.
• Volatility σ captures combined uncertainty in development and commercial outcomes: clinical effect sizes, competitor entry, pricing pressure, and uptake uncertainty. When you hold an option, you have the right, but not the obligation, to make a future decision. This creates a powerful "heads I win, tails I don't lose much" scenario. When volatility is high, the chance of both extremely bad and extremely good outcomes increases. Therefore, your maximum loss, the actual cost of creating the option itself (e.g., the cost of a pilot project, R&D, or acquiring land), is capped while your upside potential is unlimited. σ is not simply historical equity volatility, instead it should, where possible, be estimated from bottom‑up scenario analyses that reflect project‑specific drivers so that specific portfolio decisions can be made.

Modeling choices and why they matter

• Geometric Brownian motion (GBM) vs jump models: GBM is analytically convenient and provides clear intuition (higher σ raises option value). But clinical trials produce binary, jump‑like outcomes; for some programs, jump‑diffusion or discrete‑state models provide a better fit.


FIGURE 2: Geometric Brownian Motion vs Jump-Diffusion Models

Comparison of continuous Geometric Brownian Motion (teal) versus Jump-Diffusion model (red) for biotech asset valuation. The jump-diffusion model captures discrete clinical trial outcomes (Phase I, II, III readouts) more realistically than continuous GBM, showing sudden value changes at key milestones rather than smooth evolution.

• Perpetual vs finite‑horizon: perpetual models are clean analytically but can be misleading when patent expiration matters. A perpetual model may advise infinite waiting if certain parameters drive β → 1; finite‑horizon lattices (a financial tool that maps out the possible future values of an asset by using a discrete, tree-like structure) or Partial Differential Equations (PDEs) with explicit terminal conditions avoid that pitfall.
• Risk adjustment: because projects are non‑traded, risk‑neutral arguments are approximations. Practitioners choose between risk‑neutral pricing with a proxy hedging asset, certainty‑equivalent cashflows, or indifference pricing, each of which should be tested in sensitivity analysis.

In essence, these components form the vocabulary of our valuation model. The specific choices we make in defining their behavior, such as using a continuous process like Geometric Brownian Motion or accounting for a finite patent life, determine the model's fidelity to the strategic reality of the project.

Having established this conceptual architecture, our next step is to build the quantitative engine. We will begin with the most fundamental piece: modeling the uncertain path of the underlying asset, S(t).

2.1 Modeling the Underlying Asset: The Random Walk of Value

At the heart of our model is the underlying asset, denoted as S(t). This represents the total, risk-adjusted present value of all future cash flows if the drug is successfully launched. Think of it as the ultimate prize. Since the future is uncertain, we can't model this value as a straight, predictable line. Instead, we model it as a stochastic process, a path with a built-in random component (Figure 2).

A powerful and widely used model for this is the GBM previously mentioned. It describes a path that has a general upward trend but is also subject to constant, random shocks. The formula describing its tiny, instantaneous change, dS(t), is:

Let's break this down:

• μS(t)dt (The Drift): This is the predictable part. It represents the expected growth or "drift" of the project's value over a tiny time step, dt. The parameter μ is the expected rate of return in the "real world."
• σS(t)dW(t) (The Diffusion): This is the random part. It represents the unexpected volatility or "diffusion" of the project's value. The parameter σ is the volatility, and dW(t) is the random shock from a standard Brownian motion, which is the mathematical engine of randomness in the model.

FIGURE 3: Combined Effect of Drift and Diffusion in Asset Evolution

Demonstration of how Geometric Brownian Motion combines predictable drift (trend) with random diffusion (volatility). Three sample paths show different realizations starting from S=$100M, with the dashed line representing pure drift. The shaded regions indicate paths above and below the drift, illustrating the stochastic nature of asset value evolution.

The Crucial Switch: From the Real World to the "Risk-Neutral" World

GBM describes project evolution using the real-world growth rate μ. For option valuation, however, μ cannot be used. To value an option, we can't use the real-world growth rate, μ. Why? Because it's subjective and includes a risk premium that is difficult to estimate for a non-traded early-stage asset like a drug program.

Instead, finance theory provides a powerful alternative: we switch to a synthetic "risk-neutral" world. In this world, we can pretend that all assets grow at the risk-free rate, r. To do this, we replace the real-world drift μ with a risk-neutral drift, (r−q).

• r is the risk-free interest rate.
• q is the convenience yield, which represents the "cost of waiting" or the value lost by not having the asset today (e.g., from potential competitor entry).

Because a biotech project has no market price, there's no perfect way to do this. The analyst must be transparent about the approach used, whether it's finding a traded proxy asset, adjusting cash flows to their certainty-equivalents, or using another pricing framework.

2.2 Valuing the Option: The Black-Scholes-Merton PDE

Once we've defined the random walk of our underlying asset S(t), we need a tool to determine the value of our option on that asset, V(S,t). That tool is the celebrated Black-Scholes-Merton Partial Differential Equation (PDE).

Using the rules of Itô calculus—a special form of calculus designed for stochastic processes—we can derive a master equation that the option's value must always satisfy. This PDE brilliantly connects the change in the option's value to the random movement of the underlying asset.

Think of this equation as a fundamental law of financial physics for options. Each term represents a distinct economic force acting on the option's value:

• ∂V/∂t (Time Decay or "Theta"): The rate at which the option's value erodes as its expiration date approaches.
• (1/2) σ² S² (∂²V/∂S²) (Volatility & Convexity or "Gamma"): This term is the unique and powerful insight from Itô's calculus. It represents the value created by uncertainty. The term (∂²V/∂S²) (Gamma) measures the curvature or convexity of the option's value. Because of this curvature, the option gains more from a large positive price swing than it loses from a large negative one. Volatility (σ) makes these large swings more likely. This entire term, therefore, captures the profit generated by the option's non-linear payoff in a volatile environment. It is the mathematical reason why options have value even if the underlying asset is expected to go nowhere.
• (r-q)S(∂V/∂S) (Financed Growth or "Delta"): This term represents the change in option value from the underlying asset's growth, but in a special, risk-free world. The component ∂V/∂S is Delta, which measures how much the option's value changes for a $1 change in the asset price. So, S(∂V/∂S) reflects the option's total exposure to the asset's movement. The factor (r−q) is the net cost of carry, that is: the risk-free interest rate (r) you'd pay to finance holding the asset, minus any dividend or yield (q) you'd receive. In essence, this term models the gain on a financed position in the underlying asset as it grows at the risk-free rate.
• -rV (Discounting): This term ensures that the entire equation is properly discounted at the risk-free rate.

This single equation is the bedrock of option pricing. Whether you're using a closed-form formula, a binomial lattice, or a numerical solver, you are, in effect, finding a solution to this PDE. It provides a complete map of how risk, time, and cost interact to determine value.

FIGURE 4: How PDE Parameters Impact Option Value

Four-panel visualization showing the sensitivity of option value to key parameters. Top left: Impact of volatility (20-50%). Top right: Time to expiry effects (0.1-1 year). Bottom left: Delta surface showing price sensitivity. Bottom right: Risk-free rate impact (1-6%). Higher volatility, longer time, and higher interest rates generally increase call option values.

2.3 The Perpetual Option to Invest: Finding the Optimal Trigger

To build intuition, we start with a classic teaching problem: the decision to invest in a project that has no expiration date. This is called a "perpetual option." While biotech projects are finite due to patents, this simplified case is the best way to understand the core logic of real options valuation.

Our goal is to find the optimal investment trigger, a specific project value S* at which it makes sense to pay the sunk cost I and launch the project. A simple "Net Present Value" rule would say "invest as soon as the payoff A⋅S is greater than the cost I", (where in this example S represents the projected peak annual sales of a new drug, and A the present value of all sales over the drug's patent life). But this ignores a crucial fact: by investing today, you give up the option to wait for potentially better information tomorrow. This waiting option has value. So, we should only invest when the project's value S is not just greater than the breakeven point but is sufficiently high to justify killing the valuable option to wait.

So, how high is "sufficiently high"?

From PDE to ODE: Simplifying for a Timeless Problem

The general Black-Scholes PDE from section 2.2, V(S,t), values an option based on the underlying value S and time t. Because a perpetual option has no expiry date, its value doesn't depend on time. This means the time-decay term, ∂V/∂t, is zero. This simplifies the powerful PDE into a more manageable Ordinary Differential Equation (ODE) for the option value, which we'll call O(S):

This equation governs the value of our waiting option, O(S), in the "continuation region", the range of S values where it's better to wait than to invest.

The "Value Matching" and "Smooth Pasting" Conditions

To solve the optimal trigger S*, we need to define what happens at the exact moment of decision. Two elegant economic principles, known as boundary conditions, guide the mathematics:

• Value Matching: At the trigger point S*, the value of the waiting option must equal the net value you get from investing. If it were worth more, you'd keep waiting; if it were worth less, you should have already invested. This gives us our first condition: O(S*) = A⋅S* − I
• Smooth Pasting: This is the key to optimality. At the trigger S*, the slope of the option value curve must seamlessly meet the slope of the investment payoff curve. If there were a "kink" at the junction, it would mean the trigger wasn't truly optimal. This condition ensures a smooth and economically rational transition from waiting to investing, and it gives us our second condition: O'(S*) = A

Solving for the Trigger (The "Worked Algebra")

With the economic logic established, the rest is mathematics. We need a solution to the ODE that also satisfies our two boundary conditions. A standard approach for this type of ODE is to guess a solution of the form:

Think of this as a general template for our option's value. β is the crucial exponent that captures the project's risk and reward profile, while C is a constant that scales the formula to the right size. At this stage, both are unknown. Plugging this guess back into the ODE and simplifying gives us the characteristic equation for β:

This is a quadratic equation that yields two possible values (roots) for β. We choose the positive value (which is always greater than 1), as it represents the growing, leveraged nature of our investment opportunity. After this step, β is now a known number that reflects the project's volatility and the cost of waiting.

Now we have the form of our solution, O(S)=CS^β, and we can use our value-matching and smooth-pasting conditions to find the two remaining unknowns: the constant C and our ultimate goal, the trigger S*. Solving these two simultaneous equations yields the famous optimal investment rule:

Interpreting the Result and the Special Case of q = 0

Let's break down this powerful result. The term (I/A) is the simple breakeven point from a static NPV analysis. The term (β/(β-1)) is the "option premium multiple." Since β>1, this multiple is always greater than 1. It tells you how much of a premium or safety buffer you should demand over the simple breakeven point before committing capital. This multiple increases with uncertainty (σ), reflecting that with more volatility, the option to wait becomes more valuable.

This also reveals a critical limitation. What happens if there is no "cost to waiting," i.e., the convenience yield q=0? If q=0, the algebra shows that the root β becomes exactly 1. This sends our trigger S* to infinity!

Economic intuition: If the option never expires and there's no cost to holding it (no dividends or competitive erosion you're missing out on), why would you ever exercise it? You could always wait for a slightly higher value. The model correctly tells you to wait forever.

For a biotech project that is not yet launched but is protected by strong patents, the cost of delay is effectively zero (q=0). Here's why:

• No Missed Profits: The drug isn't on the market, so there are no cash flows or profits to miss out on by waiting to make the next big investment.
• No Competitive Erosion: The patent prevents competitors from launching a similar product and stealing potential market share.

So, while the project is in the "waiting" phase, the company isn't losing anything. If the option never expires and there's no cost to holding it (no dividends or competitive erosion you're missing out on), why would you ever exercise it? You could always wait for a slightly higher value. The "Perpetual Model" correctly tells you to wait forever, a mathematically logical yet completely unrealistic conclusion. Think of it from a practical standpoint considering the critical aspects of biotech projects: the single most important feature of a biotech project is the patent cliff, the date the patent expires. After that date, generic competition enters the market, and the value of the project plummets dramatically. A perpetual model completely ignores this deadline.

Therefore, using a perpetual model for a project with a finite patent life is inappropriate. It fails to capture the urgency of launching the drug and maximizing profits before the patent expires, leading to a flawed valuation. For biotech projects protected by patents, the assumption of q=0 is common, which means the perpetual model is often inappropriate. It serves as a clear signal that for a realistic valuation, the analyst must use a finite-horizon model that explicitly accounts for the patent cliff.

2.4 Compound Options: Staged R&D as Options-on-Options

Biotech research and development is a journey of sequential investments. A successful Phase 1 trial doesn't directly generate cash flow; instead, it creates the opportunity, but not the obligation, to invest in a Phase 2 trial. A successful Phase 2, in turn, creates the option to fund Phase 3, and so on. This "one success unlocks the next opportunity" structure is the quintessential real-world example of a compound option.

A compound option is simply an option on an option. Think of it like this:

• Phase 1 is an option to acquire a second option (the Phase 2 program). The "exercise price" is the cost of the Phase 1 trial.
• Phase 2 is the underlying asset for the Phase 1 option. If acquired, it is itself an option to acquire the Phase 3 program.
• This continues until the final stage, which is an option to finally launch the drug and capture the commercial value, S(t).

FIGURE 5: Value Build-up: How Compound Options Create Value

Visualization of cumulative value creation through staged drug development. Starting from preclinical ($20M total), each successful phase adds incremental option value, reaching $500M at launch. The graph shows costs (red bars), probability of success (PoS), and how the option value (teal line) grows as uncertainty resolves through each development stage.

Valuation: Why Simple Formulas Fall Short

While elegant closed-form solutions for simple European options exist (like Black-Scholes), they break down in this multi-stage reality. The pioneering work by Geske (1979) provided a formula for a European option on a European option, but this is too restrictive for biotech projects which feature:

• Early Exercise: Decisions to proceed or abandon can be made at multiple points (American-style options).
• Discrete Jumps: Trial readouts are often binary "success/fail" events, not the smooth diffusion of a stock price.

The Practical Solution: Numerical Backward Induction

Given these complexities, the most robust and practical valuation method is numerical backward induction. Instead of starting from today and moving forward, we start at the end and work our way back to the present. It's a logical, step-by-step process:

1. Start at the Final Stage (e.g., Phase 3): At the end of Phase 3, the company holds a simple option: pay the launch costs to receive the commercial value (S). We can value this final-stage option using standard techniques (like a lattice model or finite differences).
2. Step Back to the Previous Stage (e.g., Phase 2): Now, consider the decision point at the start of Phase 3. The "payoff" from a successful Phase 2 trial isn't cash; it's the value of the Phase 3 option we just calculated. So, the Phase 2 program is itself an option where the underlying asset is the value of the Phase 3 option. We then solve for the value of this Phase 2 option.
3. Continue Stepping Back: We repeat this process, stepping back through Phase 1 and any pre-clinical stages. At each step, the value of the subsequent stage's option becomes the underlying asset for the current stage's option.

This backward-looking approach correctly captures the value at each decision node by embedding the value of all future potential choices. It provides a flexible framework that can accommodate the unique features of a biotech pipeline. A detailed, two-stage numerical example of this process follows in Section 3.5.

3. A Worked Example: Valuing a Project with a Binomial Lattice

The previous sections established the theory. Now, we'll walk through a complete, step-by-step numerical example to show how these concepts are applied in practice. We will build a simple model to value a single-stage R&D investment opportunity.

Our tool of choice is a binomial lattice (or "tree"). Think of it as a road of possibilities where instead of using a single complex equation to describe the continuous random path of an asset (like the Black-Scholes model does), we break down time into a series of small steps. At each step, the asset's price is assumed to have a limited number of possible outcomes, typically moving either up or down. This popular method is intuitive because it approximates the continuous, random movement of the project's value as a series of discrete "up" or "down" steps over time. It's a powerful way to visualize uncertainty and decision-making and fits well with conventional portfolio and project management strategies in drug discovery and development (e.g.: flowcharts and underlying decision trees).

The valuation process involves three main phases:

1. Build the Tree Forward: Map out all possible future paths of the underlying asset's value.
2. Calculate Payoffs at the End: Determine the option's value at the final time step for each possible outcome.
3. Work Backward to Today: Move back through the tree, node by node, to find the option's value at the very beginning.

3.1 Setup: The Project's Vital Statistics

Let's consider a hypothetical biotech drug candidate about to enter a new development stage. Here are our key assumptions:

• Underlying Asset Value (S₀): The current estimated present value of all future cash flows if the drug is successfully launched is $200 million.
• Investment Cost (I): The cost to complete the next stage of development (our "exercise price") is $30 million.
• Time Horizon (t): We will model this decision over 3 years, using annual steps (Δt=1 year).
• Risk-Free Rate (r): The annual risk-free interest rate is 4%.
• Volatility (σ): The annual volatility of the commercial asset's value, reflecting market and clinical uncertainty, is high at 60%.
• Convenience Yield (q): We assume 0%, meaning there is no explicit cost to holding the option (e.g., no expected value erosion from competitors during this stage).

3.2 Step 1: Building the Asset Tree

First, we map out how the $200M project value could evolve. We use the Cox-Ross-Rubinstein (CRR) method to calculate the magnitude of the annual "up" and "down" movements.

• Up-step size (u): This is the "good news" multiplier. If a year brings positive clinical results or favorable market changes, the project's estimated value gets multiplied by about 1.82. So, a $200M project would become worth $364.4M. This factor represents a good outcome.
  u = e^σ√Δt = e^0.60√1 ≈ 1.822
• Down-step size (d): This is the "bad news" multiplier. If a year brings disappointing results, the project's value is multiplied by about 0.55, cutting its value nearly in half. A $200M project would become worth about $109.8M.
  d = 1/u = e^-σ√Δt ≈ 0.549

With these, we can build the tree of possible project values over 3 years, starting from S₀=$200M:

• Year 0: $200
• Year 1: 200⋅u=364.4 (up); 200⋅d=109.8 (down)
• Year 2: 663.9 (up-up); 200.0 (up-down); 60.2 (down-down)
• Year 3: 1209.9 (uuu); 364.4 (uud); 109.8 (udd); 33.1 (ddd)

FIGURE 6: Step 1: Build the Asset Tree Forward

Binomial tree construction showing how asset value evolves over time. Starting from $200 at Year 0, the tree maps all possible future paths with up-moves (u=1.822) and down-moves (d=0.549) at risk-neutral probability p=0.39. This forward construction creates the foundation for backward option valuation in the next step.

3.3 Step 2: Calculating Terminal Payoffs

Now we jump to the end of the tree (Year 3) and calculate the project's net payoff at each of the four possible final nodes. The payoff is simply the project's value minus the investment cost, but it cannot be less than zero (since the company can choose not to invest).

The formula is: Payoff = max(Final Asset Value - $30M, 0)

• S_uuu=1209.9 ⟹ Payoff = 1209.9−30=$1179.9M
• S_uud=364.4 ⟹ Payoff = 364.4−30=$334.4M
• S_udd=109.8 ⟹ Payoff = 109.8−30=$79.8M
• S_ddd=33.1 ⟹ Payoff = 33.1−30=$3.1M

These four values represent the worth of our project in three years, depending on the path taken.

3.4 Step 3: Backward Induction and Decision Making

This is where the option valuation truly happens. We work backward from Year 3 to today, calculating the option's value at each node. To do this, we first need the risk-neutral probability (p) of an "up" move and the discount factor.

• Risk-Neutral Probability (p): This isn't the real probability; it's a synthetic probability that allows us to discount expected future values at the risk-free rate (r).
  p = (e^(r-q)Δt - d) / (u - d) = (e^(0.04-0)(1) - 0.549) / (1.822 - 0.549) ≈ 0.3865
  (Obviously, the probability of a down move is 1−p≈0.6135.)
• Discount Factor: This simply means that one dollar one year from now is only worth about 96 cents today, based on the 4% risk-free interest rate.
  e^-rΔt = e^-0.04(1) ≈ 0.9608

Now, at each node, we perform two calculations:

1. Calculate "Continuation Value" (Wait value): This is the expected value of the option in the next period, discounted back to the current node. It's the value of keeping the option alive. We calculate the expected value of the option in the next year (using our special probability p) and then discount that value back to the present node.
   Continuation Value = [ p⋅(Value at up-node) + (1−p)⋅(Value at down-node) ] ⋅ Discount Factor
2. Compare with "Exercise Value": This is the value of stopping and exercising the option immediately. It's the straightforward profit you'd make: the current asset value minus the investment cost.

The option's value at the node is the higher of these two. Let's do this for one node in Year 2, the "up-down" node where S=$200M:

• Continuation Value: [ 0.3865⋅$79.8M (udd payoff) + 0.6135⋅$3.1M (ddd payoff) ] ⋅0.9608=$31.4M
• Exercise Value: 200M−30M=$170M

FIGURE 7: Step 3: Backward Induction - Working Back to Today

Option valuation using backward induction through the binomial tree. Starting from terminal payoffs at Year 3, the algorithm works backward calculating option values at each node as max(Exercise Now, Continue Holding). The initial option value of $173.4M exceeds the immediate NPV of $170M, indicating the optimal strategy is to wait rather than invest immediately.

Here, the Exercise Value ($170M) is much higher than the Continuation Value ($31.4M), so if we reach this node, we should invest immediately.

By repeating this process for all nodes back to Year 0, we find the final result:

This value is significantly higher than the simple NPV of 200M−30M=$170M. The extra $3.4 million is the option value, that is, the value of having the flexibility to wait and decide later. Since the option value at the start ($173.4M) is greater than the immediate exercise value ($170M), the optimal decision today is to wait.

3.5 Worked Example: A Two-Stage Compound Option

Now, let's apply this step-by-step logic to the compound option structure that is so central to biotech. We are at the beginning of a project (t=0) and face a decision: should we invest $10 million in a one-year Stage-1 program?

The program has a 60% probability of success. If it succeeds, it doesn't generate cash directly. Instead, it unlocks a new investment opportunity at Year 1: the option to proceed to Stage 2.

Here are the parameters for that potential Stage-2 option:

• Underlying Asset (S₁): If Stage 1 succeeds, the project's conditional value at Year 1 will be $300 million.
• Investment Cost (I₂): The cost to execute Stage 2 will be $40 million.
• Volatility (σ₂): The volatility of the Stage-2 asset is 50% (annual), Stage-2 is handled with a one-period CRR.

To value the initial $10M decision, we must work backward. First, we need to determine the value of the prize we're trying to win: the Stage-2 option.

Step 1: Value the Prize (The Stage-2 Option at Year 1)

Assuming we reach Year 1 with a successful Stage-1 result, we will hold an option to invest $40M to capture an asset worth $300M. We can value this with a simple one-period binomial model:

• Up-step (u₂): e^0.50≈1.649
• Down-step (d₂): 1/u≈0.607
• Risk-Neutral Probability (p₂): ≈0.417
• Possible Outcomes at Year 2:
• - Up-move: S_up=$300M⋅1.649=$494.7M
• - Down-move: S_down=$300M⋅0.607=$182.1M
• Payoffs at Year 2 (net of $40M cost):
• - Payoff_up = $494.7M−$40M=$454.7M
• - Payoff_down = $182.1M−$40M=$142.1M
• Expected Value (E) at Year 1 (Stage-2 payoff at t=1):
• - (0.417⋅$454.7M) + ((1−0.417)⋅$142.1M) = $272.2M

So, the prize is clear: if Stage 1 succeeds, we will be holding an option worth $272.2 million at Year 1.

Step 2: Value the Bet (The Stage-1 Investment at Year 0)

Now we can return to today and make our decision. Is it worth paying $10M for a 60% chance at winning an asset worth $272.2M in one year?

• 1. Calculate Expected Outcome: There is a 60% chance of success (yielding $272.2M) and a 40% chance of failure (yielding $0).
  Expected Value at Year 1 = (0.60 ⋅ $272.2M) + (0.40 ⋅ $0) = $163.3M
• 2. Discount to Today: We discount this expected value back one year at the 4% risk-free rate.
  Present Value of Expected Payoff = $163.3M⋅e^-0.04≈$156.9M
• 3. Calculate Net Present Value (NPV):
  NPV = (PV of Expected Payoff) - (Upfront Investment Cost)
  NPV = $156.9M - $10M = $146.9 million

3.6 Practical Notes: Cross-Checking the Valuation with LSM

While our lattice model provides a clear and accurate value, it's good practice, especially in complex cases with multiple uncertainties, to validate the result with a different method. The Longstaff-Schwartz Monte Carlo (LSM) simulation is the industry standard for this kind of cross-check.

FIGURE 8: Monte Carlo Simulation - Each Path = Different Random Outcome

Monte Carlo simulation showing 20 sample paths of asset value evolution over 3 years. Each path represents a possible future scenario starting from S=$100, with the strike price at $120 (red line). The diversity of paths illustrates the stochastic nature of asset values and why Monte Carlo methods are valuable for capturing complex option dynamics.

The key to a reliable LSM model is ensuring that the regression step—our "smart shortcut" for estimating the continuation value—is well-calibrated. This is done by choosing a set of basis functions. Think of these as the building blocks for the statistical formula. A robust and common starting point is a simple polynomial basis: [1, S, S²]. This instructs the model to find a continuation value that behaves like a quadratic function of the asset value S. For most problems, this is sufficient, but analysts can test more complex functions to see if the fit improves.

Seeing Convergence in Action

The most important diagnostic is checking for convergence. A simulation is based on random paths, so a small number of simulations can produce a noisy, imprecise result. As we increase the number of paths (M), the random noise should average out, and the estimated price should converge toward a stable value. At the same time, our statistical confidence in that estimate should increase.

Below are the illustrative results of running an LSM simulation on our worked example. Notice how the estimated price homes in on our lattice value (~$173.39M) as we increase the number of paths, while the error measure (Standard Error, or SE) shrinks:

FIGURE 9: LSM Convergence - Approaching Lattice Value

Convergence analysis of the Longstaff-Schwartz Method showing how the option value estimate stabilizes as the number of simulation paths increases. The LSM estimate converges to the true lattice value of $173.39M, with error bars showing decreasing standard error. This validates the accuracy of the Monte Carlo approach for American option valuation.

These illustrative results provide strong validation for our initial lattice valuation. They show convergence towards the lattice price (≈ $173.39M) as we increase the number of paths (M) and the reduction of Monte Carlo error. For a final report, these convergence tables and diagnostics are essential for demonstrating the robustness of the valuation.

4. ROA vs rNPV: Comparative Framework and Decision Rules

Real options analysis and risk‑adjusted net present value are not competitors in the sense of mutually exclusive methods but rather complementary lenses. rNPV translates current expectations about future cashflows and stage probabilities into a single present dollar figure. In essence, it answers the question "if we force ourselves to commit now and do nothing differently in the future, what is the expected present value?". For a more in-depth discussion on rNPV check our article at https://inbistra.com/en/blog/biotech-valuation-framework. ROA, on the other hand, reframes that question to something more closely resembling: "given we can make future choices in response to signals, what is the value of retaining that flexibility?" The difference between the two, the option premium, captures managerial flexibility and the value of staged learning.

To make this comparison operational for investment committees, begin with rNPV as a shared baseline. rNPV is transparent, auditable and fast; it lays out the revenue assumptions, the stage costs and the probability tree. It is a straightforward calculation everybody can rapidly follow and understand. When the baseline rNPV is small relative to the size of the next-stage cost or when managerial options (abandon, defer, expand) are present, compute ROA. The decision then becomes evidence-based: if ROA exceeds rNPV significantly, the manager's right to wait, to scale or to partner should be explicitly priced into funding terms. In practice, several clear patterns hold. In early, platform-oriented programs where σ is high and where small pilot studies can update beliefs, ROA typically materially increases the value relative to rNPV. By contrast, in predictable late‑stage programs where clinical outcomes are largely about enrollment and execution rather than binary scientific uncertainty, rNPV and ROA converge because managerial optionality is limited.

A practical reconciliation is to compute and present rNPV as the baseline expectation, and then compute ROA for projects with staging and high uncertainty. For each project show the rNPV, the ROA value under a base‑case σ and PoS, and the option premium percentage. Then show the same numbers under stress cases (higher σ, lower PoS). A credible governance policy is to adopt tranche sizes and documentation thresholds that are consistent with the size of the option premium: If the option premium exceeds an agreed materiality threshold (e.g., 10% of rNPV), prefer ROA-informed stage gating and reserve sizing; otherwise use rNPV for straightforward budgeting and reporting. This dual‑metric approach achieves two goals: it keeps the conversation grounded in the transparent numerics of rNPV, while communicating the strategic weight of optionality through ROA.

5. From Projects to Portfolios: Optimization Under a Budget

So far, we have valued projects in isolation. In reality, a fund or company must select a portfolio of projects from a list of candidates, all while staying within a fixed budget. This section demonstrates how to move from individual project valuation to strategic portfolio construction.

A simple and powerful starting point is to rank projects by their capital efficiency. Essentially, their "bang for the buck." We can define a metric called Expected Option Value (EOV) Density to capture this:

However, a simple ranking can be misleading. It ignores diversification and the hidden risks of correlation between projects. A truly robust selection process must also account for portfolio-level risk.

FIGURE 10: Portfolio Selection Under Budget Constraint

Optimal portfolio selection under a $100M budget constraint. Left panel shows individual projects with their costs, EOV, and value density. Right panel demonstrates the optimal allocation (Projects B, D, A) that maximizes portfolio EOV ($180M) within budget, achieving a portfolio density of 1.68x. This illustrates how value density guides capital allocation decisions.

A "Toy" Selection Example

Let's illustrate with a simple example. A fund has a budget of $100M and must choose from three candidate projects:

• Project A: Cost = $40M, EOV = $60M → EOV Density = 1.50
• Project B: Cost = $30M, EOV = $25M → EOV Density = 0.83
• Project C: Cost = $50M, EOV = $40M → EOV Density = 0.80

Approach 1: A "Greedy" Selection Based on Density

The most straightforward approach is to greedily select projects with the highest EOV density until the budget runs out.

1. Select Project A (Density = 1.50). Cost: $40M. Budget remaining: $60M.
2. Select Project B (Density = 0.83). Cost: $30M. Budget remaining: $30M.

We cannot afford Project C, so we stop. This simple method yields a portfolio of A, B.

• Total Cost: $70M
• Total EOV: $60M + $25M = $85M
• Unused Capital: $30M

This seems like a reasonable outcome, but it completely ignores the risk profile of the portfolio.

FIGURE 11: Efficient Frontier - Optimizing Risk-Return Trade-off

The efficient frontier showing optimal portfolio combinations that maximize expected return for each level of risk. The current portfolio (yellow square) can be improved by moving to the optimal portfolio (teal star) on the frontier, achieving higher returns with similar risk. Portfolios below the frontier are suboptimal as they can achieve either higher returns for the same risk or lower risk for the same return.

Approach 2: Constraining the "Tail Risk" with CVaR

Now, let's introduce a risk constraint. A common tool for this is Conditional Value at Risk (CVaR). In simple terms, CVaR answers the question: "When things go really bad, what is my average expected loss?" A CVaR_95 constraint, for instance, forces the manager to limit the average loss in the worst 5% of possible scenarios.

Let's assume a scenario analysis reveals a critical insight: Projects A and B are highly correlated. They both rely on the same biological pathway. If that pathway proves un-druggable (a "tail risk" scenario), they will likely both fail, leading to catastrophic losses. Project C, however, has a completely different mechanism of action.

• The A, B portfolio has a very high CVaR. It is fragile because of the hidden correlation. In the worst scenarios, the losses are compounded.
• The A, C portfolio, while more expensive (costing the full $100M budget), is far more robust. A failure in Project A's pathway has no bearing on Project C's outcome. Its CVaR would be significantly lower.

A risk-aware manager, bound by the CVaR constraint, would be forced to override the simple density ranking. They would have to replace the highly correlated Project B with the independent Project C. The resulting portfolio is A, C.

• Total Cost: $90M
• Total EOV: $60M + $40M = $100M

This toy example demonstrates two practical lessons:

1. Start with Efficiency: Ranking projects by EOV density is a simple and useful first pass to identify capital-efficient opportunities.
2. Optimize for Risk: This initial ranking must be followed by a scenario-based risk analysis to guard against concentration and protect the portfolio from devastating tail-risk events.

6. Garbage In, Garbage Out: A Practical Guide to Model Inputs

The mathematical machinery of real options is powerful, but its output is only as reliable as the inputs you feed into it. This section provides a practical guide for estimating the three most critical and often most challenging parameters for any biotech valuation: the project's underlying value (S₀), its probability of success (PoS), and its volatility (σ).

How to Build a Defensible Project Value (S₀)

The value of the underlying asset, S₀, is the fundamental bridge connecting the science of a drug to its financial worth. Building a credible S₀ is an exercise in modular transparency. Instead of presenting a single, black-box number, you should construct the value piece-by-piece so that every assumption can be examined and debated.

A best-practice workflow looks like this:

1. Start with the Patients: Begin with epidemiology to determine the total addressable market. How many patients have the disease?
2. Model the Market Share: Sketch realistic "uptake curves" that project how quickly the drug will be adopted by physicians and patients, considering factors like competition and reimbursement dynamics.
3. Determine the Net Price: Translate market share into revenue by applying a realistic price, making sure to account for rebates and discounts (the "gross-to-net" adjustment).
4. Calculate the Cash Flow: Convert revenues into unlevered free cash flows by subtracting the operational costs required to market and sell the drug.
5. Discount to Today: Finally, bring all expected future cash flows back to a single present value using a discount rate that reflects the risks of the commercial phase of the project.

The goal is a "glass box" model where every lever—from patient numbers to pricing—is clearly documented.

How to Think About PoS and Its Uncertainty

Stage-transition probabilities should never be a single, unqualified number. They represent the analyst's beliefs and should be presented as distributions with explicit priors grounded in a clear, evidence-based process. Use public meta-analyses (industry success rates by phase) as informative priors, then update them based on program-specific signals such as biomarkers, preclinical models, mechanistic plausibility and so on. The most robust approach is Bayesian in nature:

1. Start with a Baseline (The "Prior"): Begin with objective, historical industry data for success rates of similar drugs in the same therapeutic area and phase. This provides an unbiased starting point.
2. Incorporate Specific Evidence (The "Update"): Adjust this baseline belief up or down based on evidence specific to your project. Is there a strong biomarker? Are the preclinical results unusually predictive? Is the biological mechanism well-understood?
3. Form an Informed View (The "Posterior"): The result is your new, evidence-based PoS.

In any governance setting, this process must be transparent. Document the baseline data, the specific evidence used for the update, and test how sensitive the final investment decision is to alternative expert opinions.

Estimating Volatility σ in Biotech

Volatility in biotech is a bottom-up object. It is the combined uncertainty of commercial outcomes conditional on approval and the uncertainty about achieving approval itself if one treats S as conditional on success. The most defensible approach is a scenario-based Monte Carlo simulation: sample distributions of efficacy, market share, price and time-to-peak; compute the resulting distribution of S and use the standard deviation (or standard deviation of log S) as σ. Document how variance partitions across drivers: if σ is dominated by price assumptions, then the modeler should expose price sensitivity rather than defaulting to a single σ value. Finally, report bootstrap confidence intervals for σ so decision-makers understand parameter quality.

Robustness and the Value of Information

Finally, a good valuation isn't complete until it's been stress-tested. Two important tools are:

• Tornado Analysis: This is a powerful visualization tool that systematically tests each input assumption to see how much it impacts the final option value. The resulting chart instantly shows which levers "move the needle" the most, focusing leadership's attention on what really matters.
• Value of Information (VOI): As discussed at the beginning, ROA identifies where additional small investments in information (biomarkers, pilot studies) have high value because they unlock optionality. Therefore, VOI represents one of the most profound applications of real options analysis. It helps answer critical strategic questions like, "Should we spend $5M on a small biomarker study before committing $50M to a full Phase 2 trial?" The analysis quantifies how much a better-informed decision is worth. If the value of reducing uncertainty (e.g., improving your PoS estimate or lowering σ) is greater than the cost of the experiment, the analysis provides a clear economic rationale to fund it. It quantifies the strategic value of buying knowledge.

FIGURE 13: Tornado Analysis of Key ROA Parameters

Comprehensive sensitivity analysis showing how changes in key parameters affect option value. Critical parameters (Peak Market Size, Phase III PoS, Volatility) show the largest impact bars in red, while important parameters (Phase II PoS, Time to Market, Development Costs) show medium impact in orange, and moderate parameters show smaller impacts in gray/teal. This helps prioritize due diligence efforts.

First, run tornado analyses to find which parameters move the option value most. Use global sensitivity (Sobol or PRCC) for complex models. Perform VOI calculations for proposed small experiments: if an early biomarker trial would reduce σ or update PoS in a way that materially changes ROA, the VOI may justify the upfront trial cost even if it reduces short-term rNPV.

A Step-by-Step Example: Building the Project Value (S₀)

To make the construction of S₀ concrete, we'll walk through a hypothetical calculation. The goal is to translate a set of commercial and clinical assumptions into a single, defensible present value.

The Setup: Project Assumptions

First, we lay out all our key assumptions transparently.

Step 1: Project Annual Free Cash Flows (FCF)

Next, we build a forecast of the cash flow the project will generate each year post-launch. Let's calculate Year 1 in detail:

• Patients: 100,000 × 5% = 5,000 patients
• Gross Revenue: 5,000 × $50,000 = $250M
• Net Revenue: $250M × 80% = $200M
• Less COGS: $200M × 30% = ($60M)
• Less M&G&A: $200M × 20% = ($40M)
• Year 1 FCF: $200M - $60M - $40M = $100M

We repeat this for all years. For this example, we'll assume FCF peaks in Year 4 and then, from Year 6 onward, declines by 10% annually to model the effects of patent expiration or new competitors:

• Year 3 net revenue = $1,200M → FCF = 1,200 − 360 − 240 = $600M.
• Year 4 net revenue = $1,600M → FCF = 1,600 − 480 − 320 = $800M.
• Year 5 (plateau) FCF = $800M.
• Years 6–10 decline by 10% per year: Y6=720, Y7=648, Y8=583, Y9=525, Y10=472 (M$).

FIGURE 14: Annual Cash Flow Build-up

Detailed breakdown of annual cash flows showing revenue composition (Net Revenue in teal), costs (COGS in red, M&G&A in orange), and resulting free cash flow trajectory (teal line). The analysis spans 10 years with peak free cash flow of $800M in years 4-5, declining as patent expiration approaches. This cash flow model forms the basis for calculating the underlying asset value S₀.

Step 2: Discount FCF to Present Value (PV)

A dollar tomorrow is worth less than a dollar today. We need to discount each year's FCF back to its present value using our 10% commercial discount rate. The formula is: PV = FCF / (1 + 0.10)^Year

FIGURE 15: Future vs Present Value and Time Value of Money

Two-panel visualization of discounting mechanics. Left panel shows future cash flows (orange bars) and their present values (teal bars) over 10 years, demonstrating the declining present value of distant cash flows. Right panel illustrates the discount factor curve at 10% rate, showing exponential decay from 0.909 in year 1 to 0.424 in year 10. This fundamental concept underlies all DCF and option valuations.

Step 3: Sum the PVs to Find S₀

Finally, we sum the present values of all the annual cash flows to arrive at the total value of the project's commercial phase.

7. Case Summaries: Evidence from Public Examples

The following case studies illustrate how ROA reframes decisions in real corporate settings, shaping milestones, tranche design, and deal terms:

• Consider the example discussed by Kellogg and Charnes in their 2000 paper: the protease inhibitor Viracept from the (back then) startup Agouron. Investors in that period were effectively buying a portfolio of stage‑gated options. The lattice that Kellogg and Charnes constructed is a way to translate public announcements and evolving trial evidence into a sequence of conditional payoffs that explain why the market assigned a high valuation well before commercial revenues arrived. The lesson is that when a single program contains asymmetric upside (breakthrough potential) and relatively small incremental trial costs, ROA will typically assign a sizable option premium.
• CAR‑T platform deals provide a different lesson. The acquisitive partner often pays modest upfront sums with significant milestone payments. From a ROA perspective those milestones are explicit exercise prices: the acquirer retains the right to continue funding as efficacy signals emerge. Translating a milestone schedule into an option tree clarifies how much optionality the acquirer is buying at each step and helps justify the deferred contingent payments in economic terms rather than as negotiation artifacts.
• Platform bets such as mRNA scaling are instructive because they show how the optionality of a single technical capability can produce a multiplicative effect across multiple indications. In that sense, a platform is more like a growth option: a modest initial investment buys the right to scale into many future programs at low marginal cost. Market reactions to platform validation events (successful early clinical readouts that materially lower σ for multiple potential programs) show how quickly optionality can crystallize into enterprise valuation.

These narratives share a common practical message: ROA makes explicit the tacit managerial logic that investors often use implicitly. By making this logic numeric, ROA turns qualitative intuition into governance criteria and deal terms.

8. Governance and Communication: Our Advice

Quantitative rigor alone will not change behavior in boards and funds unless outputs are communicated in a way that connects to governance. The most useful ROA outputs are those that translate into clear decision rules and audit trails.

Begin with a short assumptions provenance statement that the board can read in five minutes. This single paragraph should state how S₀ was constructed (topline epidemiology and price), what PoS priors were used (and their sources), how σ was estimated (briefly describing the bottom‑up model), and what risk adjustment was applied. The assumptions box statement is the single most effective tool: it prevents the ROA number from being treated as a black box and forces the analyst to own the key judgments.

Second, present a three‑panel dashboard: the baseline rNPV, the ROA-adjusted value, and a compact sensitivity display that shows how the option premium responds to plausible shifts in σ and PoS. For board conversations, the narrative should be concise: "If we proceed now, expected NPV is X; if we preserve the option to wait until the Phase II biomarker, our ROA value is Y and the option premium is Y−X, which motivates a staged funding policy with a reserve of Z." The numbers anchor a governance recommendation that translates directly into an approval vote.

Third, turn model outputs into operational rules. For example: "Require follow-on approval only if posterior PoS > θ and the observed biomarker effect size exceeds threshold δ; otherwise, recommend a partial expansion of capacity but no full commit." A policy like this is actionable because it ties model thresholds to observable metrics and assigns responsibility for decisions.

Fourth, incorporate ROA outputs into term sheets and milestone structures. Milestones should be structured to preserve optionality: a series of smaller, clearly observable triggers is often superior to one large final payment because it allows the buyer and seller to re-assess and renegotiate as new information arrives without destroying optionality.

Finally, document everything. Every change in assumptions after the initial valuation should be logged with a clear reason. Decision committees should require both the updated ROA outputs and a short appendix showing the sensitivity of the decision to the updated assumptions. This audit trail builds trust and prevents ex‑post rationalization.

9. Limitations, Common Misuses, and Defensive Practices

ROA can mislead when inputs are arbitrary. Common misuses include:

• using firm equity volatility as a proxy for single-molecule σ without checking correlation assumptions.
• using closed-form Black–Scholes on non-traded projects without documenting hedging proxies.
• reporting single-point option values without uncertainty bands.

Defensive practices are straightforward: triangulate σ, bootstrap confidence intervals, always present ranges, and validate lattice results with LSM.

10. Conclusions and Immediate Actions

Real Options Analysis is the right framework for translating managerial flexibility into monetary value where biotech's staged, high‑variance discovery and development processes produce decision nodes. It is therefore especially suitable for early-stage programs and companies with high potential yet bearing a high degree of uncertainty. This report supplies both the language and the numeric tools to adopt ROA responsibly: build defensible S₀, estimate σ by bottom‑up simulation and bootstrap, present rNPV and ROA side‑by‑side, and use EOV ranking plus CVaR constraints at the portfolio level to allocate capital.

Acknowledgments

I am deeply grateful to my INBISTRA co-founder, Anabel Perez-Gomez. Her critical reading, insightful comments, and invaluable discussions about the optionality of early-stage projects were instrumental in improving and shaping the final manuscript.

This work was also profoundly inspired by the seminal reports on Real Option Analysis by David Kellogg and his colleagues. Their research opened my eyes to the power of a "systems biology" approach for valuing complex, stochastic scenarios, providing a perspective far beyond the reach of traditional DCF and rNPV methods.

Last, and perhaps less directly connected but still meaningful, my thoughts on ROAs are also influenced by financial theory, specially the book "A Random Walk Down Wall Street" published in 1973 by Burton Gordon Malkiel, which inspired section 2 of this article.

References

1. Black, F. & Scholes, M. (1973). The Pricing of Options and Corporate Liabilities. Journal of Political Economy. PDF available (Foundational article)
2. Merton, R. C. (1973). Theory of Rational Option Pricing. Bell Journal of Economics and Management Science. JSTOR (Foundational)
3. Cox, J. C., Ross, S. A., & Rubinstein, M. (1979). Option pricing: A simplified approach. Journal of Financial Economics, 7(3), 229-263. DOI: 10.1016/0304-405X(79)90015-1 (CRR lattice)
4. Geske, R. (1979). The Valuation of Compound Options. Journal of Financial Economics. DOI: 10.1016/0304-405X(79)90022-9 (Compound options)
5. McDonald, R., & Siegel, D. (1986). The Value of Waiting to Invest. The Quarterly Journal of Economics, 101(4), 707-728. DOI: 10.2307/1884175 (Perpetual option insights)
6. Dixit, A.K., & Pindyck, R.S. (1994). Investment under Uncertainty. Princeton University Press. Publisher page (Canonical text)
7. Trigeorgis, L. (1996). Real Options: Managerial Flexibility and Strategy in Resource Allocation. MIT Press. Publisher page (Applied real options)
8. Longstaff, F.A., & Schwartz, E.S. (2001). Valuing American options by simulation: a simple least-squares approach. Review of Financial Studies, 14(1), 113-147. DOI: 10.1093/rfs/14.1.113 (LSM method — implementation reference)
9. Wong, C. H., Siah, K. W., & Lo, A. W. (2019). Estimation of clinical trial success rates and related parameters. Biostatistics. https://pmc.ncbi.nlm.nih.gov/articles/PMC6409416/ (open-access empirical priors)
10. Kellogg, D., Charnes, J. M., & Demirer, R. (2000). Real options valuation for a biotechnology company. https://www.realoptions.org/papers1999/Kellogg.pdf (case study)
11. BIO Industry Analysis. (2021). Clinical Development Success Rates 2011–2020. https://go.bio/rs/490-EHZ-999/images/ClinicalDevelopmentSuccessRates2011_2020.pdf
12. Analysis Group. (2024). Biotech Asset Valuation Methods: A Practitioner's Guide. https://www.analysisgroup.com/globalassets/insights/publishing/2024-biotech-asset-valuation-methods.pdf
13. Tufts CSDD cost and timelines summaries (industry context) — see Tufts Center reports for accessible summaries.
14. FDA New Drug Therapy Approvals annual summaries (public): https://www.fda.gov/drugs/new-drug-therapy-approvals
15. DiMasi, J.A., Grabowski, H.G., & Hansen, R.W. (2016). Innovation in the pharmaceutical industry: New estimates of R&D costs. Journal of Health Economics, 47, 20-33. https://doi.org/10.1016/j.jhealeco.2016.01.012
16. Glasserman, P. (2003). Monte Carlo Methods in Financial Engineering. Springer. https://link.springer.com/book/10.1007/978-0-387-21617-1

Related Articles