Pharma Valuation

Drug-level Sum-of-the-Parts valuation for pharmaceutical companies. Each drug is valued independently through its full lifecycle, then aggregated with platform value and an equity bridge to arrive at a per-share target.

11-Stage Pipeline
DATA ACQUISITION 1. Classification Confirm pharma + archetype 2. Data Snapshot SEC + VA + market data 3. Web Research 6 parallel searches 4. Financial Extract XBRL structured data 5. WACC 9% fallback if error Data source LLM / agent call Deterministic Audit gate AI AGENTS 6. Drug Portfolio Agent Marketed drugs, revenues, LOE dates 7. Pipeline Agent Clinical drugs, phases, peak estimates 8. Reviewer Agent Cross-check, de-duplicate, correct ASSEMBLY & VALUATION 9. Assemble Input Merge agent outputs 10. Drug-Level Valuation Marketed DCF + Pipeline rNPV 11. Report Markdown investment memo Core Design: Drug-Level Sum-of-the-Parts The total company value is the sum of individual drug valuations. No portfolio-level revenue assumption is needed. Marketed Drugs — Discounted Cash Flow For each approved, revenue-generating drug: 1. Project revenue through growth, plateau, and patent expiry 2. Apply IRA price reduction if eligible (Medicare negotiation, ~20%) 3. Model post-expiry erosion from generic/biosimilar entry Pipeline Drugs — Risk-Adjusted NPV For each clinical-stage candidate: 1. Look up FDA success rates by phase and therapeutic area 2. Apply modifiers (biomarker, breakthrough, orphan designation) 3. Weight projected revenue by cumulative probability of approval

Sum-of-the-Parts

Each drug is valued with its own revenue trajectory, profit margin, patent expiry date, and erosion profile. The total enterprise value is the sum of all individual drug DCFs plus platform value, minus corporate overhead.

Agent Intelligence

Three LLM agents extract drug data from SEC filings and web research: Drug Portfolio (marketed), Pipeline (clinical candidates), and Reviewer (cross-checks and de-duplicates). The Reviewer is non-blocking — raw agent output is used as fallback.

6 Company Archetypes

Classification selects archetype-specific parameters: large-cap diversified, large-cap pure pharma, mid-cap specialty, small biotech (commercial stage), pre-revenue biotech, and biosimilar company. Each gets different beta bounds, materiality thresholds, and platform value settings.

Data Acquisition & Agent Pipeline
Stage 1: Classification Gate Confirms the ticker is a pharmaceutical company. Determines company archetype, which sets downstream parameters. Archetype controls: WACC beta bounds, materiality thresholds, expected drug count, platform value eligibility. Large-Cap Diversified Large-Cap Pure Pharma Mid-Cap Specialty Small Biotech (Comm.) Pre-Revenue Biotech Biosimilar Company Stages 2 & 4: Financial Data SEC EDGAR 10-K, 10-Q filings VA Data Supplemental filings yfinance Price, beta, shares, mkt cap XBRL parsing: revenue, R&D, COGS, SG&A, net debt, shares outstanding Stage 3: Web Research (Non-Blocking) Drug portfolio + revenues Pipeline + trial phases Patent expiry + LOE dates Competitive landscape Biosimilar threats Regulatory + IRA impact If web research fails, agents proceed with SEC data alone. Stage 5: WACC Re = Rf + beta * ERP. WACC = (E/V) * Re + (D/V) * Rd * (1-t). Beta floored/capped by archetype. Falls back to 9% on error. WACC bounds: clamped to [3%, 30%]. Values outside this range produce nonsensical valuations. Stages 6-8: AI Agent Trio Each agent reads the full context (SEC filings + web research) and outputs structured JSON. Drug Portfolio Agent For each marketed drug, extracts: Name, annual revenue, therapeutic area, modality, loss-of-exclusivity year, growth rate Pipeline Agent For each clinical candidate, extracts: Name, indication, current phase, peak revenue est., expected approval year, modality, designations Reviewer Agent (Non-Blocking) Cross-checks the two drug lists: Removes duplicates, corrects misclassifications, validates LOE dates. Falls back to raw output on failure.

XBRL Parsing

Financial extraction parses structured XBRL tags from SEC filings to get revenue breakdown, R&D spend, COGS, SG&A, operating expenses, net debt, and shares outstanding at machine precision.

Resilient Design

Only classification and financial extraction are hard gates. Web research, WACC, and the Reviewer agent are non-blocking: the pipeline continues with SEC-only data, a 9% default WACC, or raw agent output respectively.

Structured Extraction

Agents output validated JSON. Each marketed drug includes annual revenue, therapeutic area, drug modality, LOE year, and growth rate. Each pipeline drug includes current clinical phase, peak revenue estimate, and regulatory designations.

Marketed Drug Valuation — DCF Through Patent Expiry

Each Drug Gets Its Own DCF

For every approved, revenue-generating drug, the engine projects year-by-year revenue through four lifecycle phases: growth to peak, plateau, IRA price reduction (if eligible), and post-patent erosion from generic or biosimilar entry. Revenue is multiplied by a drug-specific net margin (looked up by therapeutic area and drug modality), taxed, and discounted at WACC using mid-year convention.

Revenue Lifecycle — Per Drug Years Growth Revenue grows at annual rate to peak Peak Revenue IRA Cut Medicare negotiation reduces price ~20% LOE Patent Expiry Generic Erosion Revenue declines as generics/biosimilars enter Per-Drug Net Margin Looked up by (therapeutic area, modality): Oncology + biologic: 78% | Cardio + small mol: 65% Rare disease + gene therapy: 80% | ~30 combos total Inflation Reduction Act (IRA) Medicare can negotiate prices for top-spend drugs. Small molecules: eligible after 9 years. Biologics: 13 years. Loss of Exclusivity (LOE) Patent expiry date for each drug. After LOE, generic or biosimilar competition erodes branded drug revenue.
Post-Patent Revenue Erosion — By Drug Type After patent expiry, revenue declines as competitors enter. The speed of decline depends on the drug type and barriers to generic/biosimilar entry. Values show fraction of pre-expiry revenue retained in each year after loss of exclusivity. Drug Type Year 0 Year 1 Year 2 Year 3 Year 5+ Typical Drugs Small Molecule — Rapid Erosion 100% 15% 8% 5% 2% Atorvastatin (Lipitor), clopidogrel (Plavix), omeprazole Small Molecule — Moderate Erosion 100% 35% 20% 12% 5% Extended-release, combination products, abuse-deterrent Complex Generic (Inhaler, Injectable) 100% 50% 35% 25% 12% Inhaled corticosteroids, injectable formulations, transdermals Biologic — Standard Biosimilar Entry 100% 65% 45% 35% 22% Adalimumab (Humira), infliximab, trastuzumab, bevacizumab Biologic — Complex / Limited Competition 100% 80% 65% 50% 30% CAR-T therapies, bispecific antibodies, ADCs Orphan Drug (7-Year Exclusivity) 100% 70% 55% 45% 30% Enzyme replacement therapies, rare disease treatments Ultra-Rare Orphan (<5,000 patients) 100% 85% 75% 68% 55% Ultra-rare enzyme therapies, gene therapies for tiny populations Gene Therapy — No Erosion Expected 100% 100% 100% 100% 100% One-time curative gene therapies with no biosimilar pathway Sources: DrugPatentWatch (2025), PubMed PMID 41166003, IQVIA Biosimilar Reimbursement Analysis, Lipitor/Humira empirical data

Why Erosion Matters

A blockbuster small-molecule drug like Lipitor lost 85% of branded revenue in the first year after generic entry. A biologic like Humira retains ~65% in year 1 because biosimilars face higher barriers (no automatic pharmacy substitution, interchangeability requirements).

Automatic Curve Selection

If an erosion profile isn't specified, the engine infers one from drug modality: small molecules get rapid erosion, biologics get standard biosimilar curves, ADCs and bispecifics get complex biologic curves, orphan-designated drugs get the orphan profile.

IRA Eligibility

The Inflation Reduction Act allows Medicare to negotiate prices for top-spend drugs. Small molecules become eligible 9 years after approval; biologics after 13 years. The default model applies a 20% price reduction in eligible years before patent expiry.

Pipeline Drug Valuation — Risk-Adjusted NPV

Discounting Revenue by FDA Success Probability

Most clinical-stage drugs never reach market. The risk-adjusted NPV (rNPV) method accounts for this by weighting projected revenue by the cumulative probability of surviving all remaining FDA approval stages. A Phase I oncology drug with 10% overall approval probability is worth far less than a Phase III cardiovascular drug with 56% probability, even with identical peak revenue estimates.

FDA Phase Transition Rates — By Therapeutic Area Probability of advancing from one clinical phase to the next. Cumulative = probability of reaching approval from Phase I. Therapeutic Area Phase I → II Phase II → III Phase III → NDA NDA → Approval Overall from P1 Key Drivers Oncology — Solid Tumor 57.6% 32.7% 60.0% 90.0% 10.2% High Phase II attrition; complex biology, heterogeneous tumors Oncology — Hematologic 65.0% 45.0% 62.0% 92.0% 16.7% Better-characterized biology; highest LOA in BIO 2022 Immunology / Autoimmune 69.8% 45.7% 63.7% 91.0% 18.5% Includes rheumatology, dermatology, GI autoimmune Cardiovascular 73.3% 65.7% 62.2% 90.0% 26.9% Large outcomes trials; high bar for hard endpoint data Rare Disease (Non-Oncology) 78.0% 55.0% 65.0% 95.0% 26.5% Orphan designation: smaller trials, accelerated pathways Infectious Disease 70.1% 58.3% 75.3% 92.0% 28.3% Clear endpoints (viral load, clearance); shorter trials Vaccines 76.8% 58.2% 85.4% 93.0% 35.5% Highest overall LOA; clear immunogenicity endpoints Success Rate Modifiers (additive to cumulative probability, clamped to 1%–99%) Increase: Biomarker-selected (+15pp) Breakthrough designation (+10pp) Strong prior data (+12pp) Orphan designation (+5pp) First-in-class (+5pp) Antibody modality (+5pp) Decrease: Failed prior Phase III (−20pp) Sources: Wong et al. 2019 (Biostatistics), BIO/Informa Clinical Development Success Rates 2011-2020, Nature Communications 2025, Tufts NEWDIGS 2023
rNPV Computation — Per Pipeline Drug rNPV = NPV( Success Probability x Revenue x S-Curve Ramp x Margin x (1 - Tax) ) - NPV( Staged Development Costs ) Revenue Side Each future year's revenue is weighted by the drug's cumulative probability of reaching market. Revenue ramps from launch to peak via an S-curve (logistic function), then plateaus through exclusivity, then erodes. Discounted at WACC, mid-year convention. Development Cost Side Remaining costs split by phase, each weighted by probability of reaching that phase: Current phase cost x 100% (certain) Next phase cost x P(advance) e.g. 35% Final phase cost x P(reach) e.g. 21% Costs by area: Oncology $650M from P2, Rare disease $300M, CNS $750M. S-Curve Revenue Ramp Revenue builds gradually from launch. Logistic function: slow start, acceleration, then deceleration near peak revenue. Launch Peak

Why Phase II Is the Killer

Across most therapeutic areas, the Phase II to Phase III transition has the lowest success rate (often 30-50%). This is where efficacy signals from small trials often fail to replicate. The rNPV model captures this by making Phase II drugs worth dramatically less than Phase III drugs.

Biomarker Effect

Programs that use selection biomarkers (patient enrichment) have 3x higher approval rates (25.9% vs 8.4% per BIO 2022). The +15pp modifier to cumulative probability reflects this well-documented advantage in precision medicine trials.

Cost-to-Complete

Development costs vary widely by area: CNS drugs cost ~$750M from Phase II, oncology ~$650M, rare disease ~$300M. These costs are spread across remaining phases and probability-weighted — you only "spend" Phase III costs if you reach Phase III.

Aggregation, Sensitivity & Confidence
Platform Value — Going-Concern R&D Engine Captures value of future drugs not yet in the portfolio. Two methods, blended 50/50 (adjustable by archetype). R&D Productivity Perpetuity Value = (Annual R&D Spend x IRR x Fade) / WACC IRR by size: Large cap 7% | Mid cap 10% | Small biotech 12% Fade factor: 0.85 (R&D productivity declines over time). Pre-revenue biotech: $0. Terminal Value (Gordon Growth Model) TV = (Revenue x 40% steady-state x 20% FCF margin x 1.025) / (WACC - 2.5%) Discounted back from end of 15-year forecast horizon at WACC. 40% of current revenue assumed to persist long-term through pipeline replenishment.
Equity Bridge — Enterprise Value to Per-Share Price Marketed Drug DCFs + Pipeline rNPVs + Platform - Overhead 20% = EV Enterprise Value - Net Debt + Preferred + Minority = Equity ÷ Diluted Shares = $/Share Patent cliff flag: raised when any single drug >40% of revenue and LOE within 5 years. Residual revenue (unidentified drugs) assigned default margin and reduces confidence score.
Sensitivity Analysis — 6 Independent Dimensions Each parameter varied independently to produce bull and bear per-share values. WACC ±200bps Base Margins ±5pp Pipeline Success ±20% Platform Method R&D Only Blended Terminal Only LOE Dates ±2 years Peak Revenue ±30% Center line = base case | Left/top = bear | Right/bottom = bull
Confidence Score — 4 Sub-Scores x 25% How much to trust the per-share estimate. Each dimension 0-100, equally weighted. Maps to: High (80+), Moderate-High (60-79), Moderate (40-59), Low (<40). Data Quality (25%) Revenue coverage of identified drugs Number of individually modeled drugs Unidentified revenue fraction penalty Source reliability and data freshness Portfolio Risk (25%) Revenue HHI (concentration index) Patent cliff exposure (3-year lookout) Pipeline buffer against cliff losses Top-drug revenue dependency Sensitivity Range (25%) Bull-bear range as % of base value WACC sensitivity magnitude Narrower range = more confident <15% range → 100, >60% → 0 Structural Certainty (25%) Platform value as % of total (lower=better) Unidentified revenue as % of total Pipeline as % of total (speculative) More data-driven = higher score

Equity Bridge

EV = Sum(marketed DCFs) + Sum(pipeline rNPVs) + platform value - 20% corporate overhead drag. Equity = EV - net debt - preferred stock - minority interest. Per share = equity / diluted shares.

Patent Cliff Detection

Automatically flags when any single drug exceeds 40% of total revenue and faces patent expiry within 5 years. This directly penalizes the portfolio risk sub-score and is highlighted in the report.

15-Year Horizon

Marketed drugs valued through LOE + erosion tail. Pipeline drugs modeled for up to 20 years (launch + exclusivity + erosion). Platform value captures beyond-horizon value through the R&D perpetuity and terminal value methods.