StratOS — Strategy Intelligence Engine

♟

Strategy Intelligence Engine

Select an Analysis Theme above, describe your situation, and StratOS will deploy the right AI agents — each thinking as a senior expert in their domain — to generate a decision-grade intelligence brief.

📊 Excel Model Builder

AI-Driven Data Synthesis K-Means Clustering Monte Carlo Simulation Brand vs Competitor Dashboard

Model

Data

Configure

Generate

Select Analysis Model

🔵

K-Means Segmentation

Cluster HCPs, patients, or markets. Auto-detect optimal K. Outputs segment profiles + revenue potential.

🎲

Monte Carlo Simulation

10,000-run revenue forecast with probabilistic distribution. Confidence intervals & scenario P10/P50/P90.

⚔️

Brand vs Competitor Dashboard

Multi-dimensional performance scoring. Indexed benchmarks vs 3 competitors across 12 KPIs.

🗺️

Market Opportunity Heatmap

Territory × therapy × segment opportunity scoring. Prioritised resource allocation matrix.

💧

Revenue Waterfall & Bridge

Decompose revenue delta. Volume, price, mix, new product contribution. Dynamic waterfall chart.

🧩

Portfolio ROI Optimizer

BCG matrix + NPV scoring. Prioritise pipeline assets by risk-adjusted return & strategic fit.

🎛️

Scenario / What-If Simulator

Type Base, Bull, or Bear into one cell — every driver and the combined output recalculates live via formulas.

🎯 AI Deck Generator

McKinsey / Bain / Big 4 Quality AI-Decided Structure No Fixed Templates

📝

Source Material

Paste article, brief, report, or strategy document

Deck Style

Slides

Audience

🔮 Forecast Builder

Mode → Technique → Inputs → Model Real Calculation AI-Written Rationale

Mode

Technique

Inputs

Model

What are you forecasting?

📈 Pharma Revenue Forecasting — Bootcamp

From Zero to Forecast-Ready in 5 Weeks

Master the art and science of pharma revenue forecasting — from epidemiology-driven patient pools and competitive intelligence to multi-scenario models used by top-tier commercial teams at Roche, Novartis, and Sun Pharma.

📅 5 Weeks 🧮 6 Real Models Built 🏥 Oncology · Immunology · Dermatology 💼 Portfolio-Ready Output 🌍 WHO / SEER / GLOBOCAN Data

What You Will Learn

WK1

Epidemiology & Patient Flow

Build patient pools from global disease registries · Funnel modelling

▼

Build patient pools using WHO / SEER / GLOBOCAN data

These are the three primary sources forecasters lean on for defensible sizing. WHO's Global Health Observatory publishes disease burden statistics across nearly every therapeutic area worldwide. SEER (Surveillance, Epidemiology, and End Results) is the US NCI's cancer registry, covering roughly 48% of the US population with incidence, stage-at-diagnosis, and survival data — the gold standard for US oncology sizing. GLOBOCAN (IARC) gives global cancer incidence, mortality, and prevalence estimates by country, essential for ex-US and global forecasts. Outside oncology, therapeutic-area-specific registries (IDF for diabetes, the GBD Study for broader disease burden) play the same role. Example: Forecasting a new NSCLC therapy — pull SEER's stage-specific lung cancer incidence for the US number, then cross-reference GLOBOCAN for ex-US markets, giving a triangulated population instead of a single analyst's guess.

Incidence vs Prevalence — know when to use each

Incidence = new cases diagnosed in a period — use it for forecasting new patient starts or short-duration/acute therapies. Prevalence = total existing cases at a point in time (new + previously diagnosed still living with disease) — use it for chronic, long-duration therapies, since your addressable pool includes people diagnosed years ago who are still eligible today. As a rule of thumb, Prevalence ≈ Incidence × Average Disease Duration. Example: A curative hepatitis C drug should be sized on incidence only — cured patients exit the pool. A chronic hypertension drug should be sized on prevalence — every diagnosed hypertensive patient is a potential user each year. Using incidence for a chronic condition can undercount the market by 4× or more.

Patient funnel modelling — diagnosed to drug-eligible

Structure the disease population as a sequential funnel: Total/At-Risk Population → Diagnosed → Treated (on any therapy) → Eligible for your specific drug (label-matched) → Actually prescribed. Each stage has its own real-world drop-off rate, sourced separately — diagnosis rates from registries, treatment rates from claims data, eligibility from label/guideline criteria. Example: 100,000 prevalent patients → 70% diagnosed (70,000) → 65% of diagnosed treated (45,500) → 55% of treated patients meet your label criteria, e.g. failed first-line (25,025). That 25,025 — not the initial 100,000 — is your true addressable pool.

Eligible population sizing — diagnosis, treatment, and uptake logic

The final, most granular layer — diagnosis rate × treatment rate × label/guideline eligibility × any further exclusions (contraindications, prior-therapy requirements, biomarker status) — to compute patients who could actually receive a prescription for YOUR drug, not just any drug in the category. Example: A biomarker-driven oncology drug (e.g. EGFR-mutant NSCLC) — of all diagnosed NSCLC patients, only ~15% test positive for the mutation, and only 60–80% get tested at all depending on the market. Both cuts apply multiplicatively before you reach the eligible population.

🧬 How-To: Patient-Based Epidemiology Model

What it is: a bottom-up forecast that starts from the total population and narrows it down, filter by filter, to the number of patients who will actually be on your drug — instead of assuming a market share of someone else's number.

Eligible Patients = Population × Incidence/Prevalence Rate × Diagnosis Rate × Treatment Rate × Addressable %
Revenue(t) = Eligible Patients × Brand Share(t) × Price × Persistency

Worked example: Population 10,000,000 → Prevalence 0.8% = 80,000 patients → Diagnosis rate 65% = 52,000 diagnosed → Treatment rate 70% = 36,400 treated → Addressable to your line of therapy 60% = 21,840 eligible patients. At a 15% peak share, $20,000/year price, and 70% persistency: 21,840 × 15% × $20,000 × 70% ≈ $45.9M peak-year revenue.

WK2

Competitive Intelligence

Landscape mapping · Pipeline analysis · MoA-based share logic

▼

Competitor landscape mapping

A living, continuously-updated tracker of every relevant competing asset — approved products, Phase III (near-term threats), Phase II/I (directional, longer-term), and discontinued programmes (so you stop tracking false threats). Structure it with columns: asset, company, MoA, phase, expected approval date, target-population overlap %, and threat level. Example: Forecasting a Phase III psoriasis asset — track not just the 3 approved biologics but the 4 Phase III JAK inhibitors in the pipeline, since one could launch mid-way through your 5-year window and materially change your peak-share assumption.

Pipeline analysis

Read clinical trial data through a commercial lens, not just a scientific one: primary/secondary endpoints (do they match what payers and guidelines actually reward), patient population/inclusion criteria (does it overlap your label), and realistic timelines (trial completion + regulatory review time, not the sponsor's optimistic press-release date). Example: A competitor's Phase III primary endpoint is PFS, but payers in your market reimburse based on OS data — a real access-delay risk you'd only catch by reading the trial design, not the topline headline.

MoA-based share logic

How mechanism of action, safety profile, and dosing convenience drive switching behaviour between competing products — see the full How-To box below for the weighted displacement formula and a worked example.

Trial reading for forecasters

Translate clinical endpoints into commercial differentiation language: NNT (Number Needed to Treat — lower is better, drives prescriber preference), ORR (Objective Response Rate — magnitude of efficacy, common in oncology), and PFS/OS deltas (the actual months of benefit, which becomes your pricing and value story). These numbers feed directly into your MoA-based share logic. Example: Drug A has an NNT of 4 vs Drug B's NNT of 12 for the same endpoint — Drug A requires a third as many patients treated to prevent one additional bad outcome, a powerful differentiator for guideline committees that should translate into a higher assumed peak share vs Drug B.

🎯 How-To: MoA-Based Share Displacement Model

What it is: estimates how much share a new entrant takes from each existing competitor — not evenly, but weighted by how mechanistically similar/substitutable the two products are. A same-MoA competitor loses share fastest; a different-MoA, different-line competitor barely moves.

Share Taken From Competitor X = New Entrant Peak Share × Substitution Weight(X)
Substitution Weight = f(same MoA, same line of therapy, same administration route, overlapping label) — score 0–1 per competitor, normalise so weights sum to 1 across the competitive set

Worked example: New entrant reaches 18% peak share. Three competitors: A (same MoA, same line) weight 0.55; B (different MoA, same line) weight 0.30; C (same MoA, later line) weight 0.15. Share erosion: A loses 9.9pts, B loses 5.4pts, C loses 2.7pts of the new entrant's 18% — guiding which competitor's forecast to revise down hardest.

WK3

Revenue Forecast Models

Patient-based & market-based models · Pricing · Gross-to-Net logic

▼

Patient-based models

The bottom-up approach detailed in Week 1 — repeated here to compare against market-based models. Preferred when your drug is first-in-class or first-in-indication (no true analogue exists), local epidemiology data is strong, and stakeholders need full assumption-level auditability, defensible back to a named source at every step.

Market-based models

The analogue-driven top-down approach — see the How-To box below for the formula and worked example. Preferred when a genuinely comparable product exists in the same class or mechanism, epidemiology data is unreliable or unavailable (common in ex-US emerging markets), or leadership needs a fast sanity-check forecast without a full funnel build.

Pricing / dosing / duration

Three variables that determine per-patient revenue: Average Daily Dose (from the trial's dosing regimen), Treatment Duration (chronic vs acute — how long a typical patient stays on therapy), and Pack Size (units per prescription, affecting days-of-therapy per fill). WAC (Wholesale Acquisition Cost) is list price before discounts; NAS/Net Price is WAC minus all rebates — model both, since WAC drives gross revenue but NAS drives what you actually collect. Example: A once-daily oral drug, 30-day pack at $3,000 WAC → Annual Gross Revenue per Patient = (365/30) × $3,000 ≈ $36,500. At a 35% gross-to-net deduction, Net Revenue per Patient ≈ $23,725.

Gross-to-Net logic

The bridge from list-price gross revenue to actual net revenue collected — driven by rebates (paid for formulary placement), chargebacks (wholesaler-to-provider adjustments), co-pay support (reduces what the patient pays, but rarely reduces the payer-negotiated net), and mandatory Medicaid Best Price/Unit Rebate adjustments, often the deepest discount tier. Example: $500M gross (WAC) revenue with a blended 28% gross-to-net deduction (typical for a competitive specialty category) → Net Revenue = $500M × 0.72 = $360M — the number that actually hits the P&L, and the one your forecast should headline.

📈 How-To: Analogue / Market-Share Curve Method

What it is: instead of building a forecast from scratch, you find a real product that launched into a comparable class/access environment and scale its actual adoption curve to your market. Use this when epidemiology data is thin or the launch is too novel for a clean bottom-up build.

Normalised Curve(t) = Analogue Sales(t) ÷ Analogue Peak Sales [gives % of eventual peak reached by year t]
Your Revenue(t) = Your Total Market Size × Adjustment Factor × Analogue Peak Share × Normalised Curve(t)

Worked example: Your market size is $800M. The analogue reached 20% peak share in Year 5. Your access/competitive position is 80% as favourable as the analogue's (adjustment factor). By Year 3, the analogue had reached 51% of its eventual peak. Your Year 3 revenue ≈ $800M × 80% × 20% × 51% ≈ $65.3M.

∫ How-To: Logistic Growth Curve (S-Curve)

What it is: the mathematical shape most product adoption actually follows — slow start, fast middle, plateau at a ceiling — rather than a straight line. The same curve, mirrored, models LOE/biosimilar erosion (slow initial decline, steep mid-period loss, then plateau).

Value(t) = L / (1 + e^(−k×(t − t0)))
L = ceiling (peak value/share) · k = growth rate (steepness) · t0 = inflection year (where you cross 50% of L)

Worked example: L = $600M ceiling, k = 0.8, t0 = Year 3 (inflection). At t=3, Value = L/2 = $300M exactly. At t=5: 600M / (1+e^(−0.8×2)) = 600M / (1+0.20) ≈ $499M — showing how fast the curve approaches its ceiling after the inflection point.

WK4

Advanced Skills

Sensitivity analysis · Scenario planning · Generic / biosimilar erosion

▼

Sensitivity analysis

Tests how much the forecast output moves when you change ONE input at a time, holding others at base case — usually visualised as a tornado diagram (bars sorted widest-to-narrowest, showing which variable swings the output most). One-way tables vary a single variable across a range; two-way tables show two variables interacting in a grid. Example: Varying Peak Share 10%→20% swings Year 5 revenue by ±$45M; varying Persistency 60%→80% only swings it ±$12M — Peak Share is the variable worth the most validation effort, not persistency.

Scenario planning

Instead of moving one variable at a time, bundle a coherent SET of assumptions into named cases: Base (most likely), Bull (access, share and pricing all land favourably), Bear (competitive entry, access delay, price erosion). Weight each by probability — the resulting probability-weighted average, not the Base Case alone, is what should inform a major LCM investment decision. Example: Base $400M (55% prob.), Bull $580M (20%), Bear $250M (25%) → Weighted Forecast = 0.55×400 + 0.20×580 + 0.25×250 = $398.5M.

Generic / biosimilar erosion

Erosion curves model post-LOE decline speed, calibrated by therapeutic class (small-molecule generics erode fast — 70–90% volume loss within 12 months; biosimilars erode more gradually — 20–40% in Year 1 — due to manufacturing complexity and prescriber inertia), market (the US erodes faster than the EU due to automatic generic substitution), and payer pressure (mandatory step-edits accelerate erosion). This calibrates the k and t0 parameters of the Logistic Erosion Curve from Week 3. Example: A biologic's first biosimilar entrant typically leaves the originator with 60–70% of volume by end of Year 1; a small-molecule facing generics typically retains only 10–20% by the same point. Using the wrong benchmark class understates remaining revenue by 3–4×.

Forecast storytelling

The substance of how you walk a stakeholder through the forecast — distinct from the slide-ordering discipline in Week 5's presentation how-to. Cover: what you're confident about and why (grounded in named sources), what you're uncertain about and why (data gaps, novel mechanism, thin analogue), and what single piece of evidence would change your mind. A forecast with no clear "what would change this number" is one nobody should trust blindly.

🎲 How-To: Monte Carlo Simulation

What it is: instead of forecasting a single "base case" number, you acknowledge that your key drivers (peak share, price, persistency) are uncertain — so you simulate thousands of plausible combinations and report the resulting range, not a false-precision point estimate.

1. Identify uncertain variables (e.g. Peak Share, Price, Persistency)
2. Assign each a distribution — Triangular(Low, Base, High) is standard for expert-elicited ranges
3. Randomly sample all variables together, compute Revenue for that one scenario
4. Repeat 2,000–10,000 times
5. Sort the results; report P10 (pessimistic), P50 (median/base), P90 (optimistic)

Worked example: Share sampled as Triangular(5%, 12%, 22%); Price as Triangular($15K, $20K, $24K). Across 2,000 runs, Year 5 revenue might land: P10 = $38M, P50 = $84M, P90 = $156M. You present all three — the P10–P90 band is your defensible range, not the P50 alone.

🌊 How-To: Bass Diffusion Model

What it is: separates adoption into two forces — innovators who adopt independently of others, and imitators who adopt because they see peers using it. Built for genuinely novel launches (new mechanism, new category) where word-of-mouth materially drives uptake — common in specialty and orphan disease.

F(t) = (1 − e^(−(p+q)t)) / (1 + (q/p)×e^(−(p+q)t))
p = coefficient of innovation (typ. 0.01–0.03) · q = coefficient of imitation (typ. 0.3–0.5) · m = market potential
Cumulative(t) = m × F(t)

Worked example: m = $700M, p = 0.02, q = 0.38. At t=3: F(3) ≈ 0.42, so Cumulative(3) ≈ $294M captured. At t=5: F(5) ≈ 0.71, Cumulative(5) ≈ $497M — showing the characteristic slow-start, fast-middle diffusion shape driven by imitation once early adopters establish credibility.

WK5

Capstone Case Study

Solve a real business case · Build your portfolio-ready model · Present like a pro

▼

Solve a real business case

Applies everything from Weeks 1–4 to one deliberately messy, realistic scenario — incomplete data (mirroring real pharma forecasting, where you rarely have perfect epidemiology AND a clean analogue AND full competitive clarity at once), competitive uncertainty (a pipeline asset that may or may not launch inside your window), and payer ambiguity (unclear formulary placement). You make and defend judgment calls, not just execute a formula. Example case: "Forecast a Phase III oral GLP-1 launching in 3 years, in a market where 2 injectable competitors are already approved and a rival Phase III oral agent could launch 6 months before or after yours — build the range, not a point estimate."

Build portfolio-ready final model

Integrates everything into one professional, formula-linked Excel workbook: an Epidemiology/Patient Flow sheet, a Competitive sheet (landscape + share logic), and a Revenue sheet (pricing, gross-to-net, scenario outputs) — linked so changing one epidemiology assumption flows through to the final revenue number automatically, the way a real commercial forecasting team's model must work for audit and governance. Tip: Use this platform's Forecast Builder output as your Revenue sheet, and layer in Epidemiology and Competitive tabs manually to complete the portfolio piece.

Present like industry professionals

Structure your assumptions, defend your numbers, and handle pushback confidently — see the How-To box below for the exact slide order and structure a CFO or investment committee expects.

🎓 Full Worked Capstone: Forecasting "Zenbrolix" — an Oral JAK Inhibitor for Moderate-to-Severe Atopic Dermatitis

Illustrative teaching case — Zenbrolix is a fictional asset built to walk through every method in this course end-to-end. The numbers are constructed to be realistic in scale, not published real-world figures. US market, 5-year post-launch forecast.

Step 1 — Epidemiology & Patient Funnel (Week 1)

Build the eligible pool from population down to the patients who could actually receive Zenbrolix, using prevalence (chronic condition) rather than incidence. US adults 260M → AD prevalence 7% → 18.2M → Moderate-to-severe subset 20% → 3.64M → Diagnosed & under specialist care 60% → 2.18M → Systemic-therapy eligible (failed topical) 55% → 1.20M → Failed/contraindicated to biologic OR oral-preferring 35% → 420,000 eligible patients

Step 2 — Competitive Landscape (Week 2)

Map the field: 2 approved injectable biologics (established, different MoA/route — low substitution weight), 1 approved topical JAK (different route — low weight), and 1 approved oral JAK already on market 18 months (same MoA and route — high substitution weight, and critically, a real analogue for Step 3). Zenbrolix differentiates on a cleaner safety label (no boxed-warning age restriction) but launches 2nd-to-market in its sub-class.

Step 3 — Technique Selection & Adoption Curve (Week 3)

A genuine analogue exists (the first oral JAK), so the Analogue Method is the right primary technique, cross-checked against the Step 1 epidemiology funnel as a ceiling sanity-check. Analogue reached 15% peak share of its eligible pool by Year 4 Zenbrolix adjustment factor: 85% (2nd-to-market discount, partly offset by better safety label) Zenbrolix peak share = 15% × 85% ≈ 13% of the 420,000 eligible pool → 54,600 patients at peak

Step 4 — Pricing & Gross-to-Net (Week 3)

WAC: $5,800/month → $69,600/year Gross-to-net deduction: 35% (competitive immunology category, PBM rebates) Net price per patient/year ≈ $45,240 Persistency: 65% (oral convenience offset by modest GI tolerability dropout)

Step 5 — Revenue Build (Weeks 1–3 combined)

Apply the analogue's normalised adoption curve (15% / 35% / 60% / 85% / 100% of peak by year) to the peak-year revenue. Peak-year (Y5) Revenue = 54,600 patients × 65% persistency × $45,240 net price ≈ $1.61B

Year	% of Peak	Patients on Therapy	Net Revenue
Y1	15%	5,320	$241M
Y2	35%	12,425	$563M
Y3	60%	21,300	$966M
Y4	85%	30,175	$1.37B
Y5	100%	35,490	$1.61B

Step 6 — Scenario Planning (Week 4)

The Base Case above is one path — bundle it with Bull and Bear cases for the LCM investment decision. Base (55% prob.): $1.61B peak — analogue curve holds Bull (20% prob.): $2.00B peak — faster formulary wins + no new entrant Bear (25% prob.): $1.00B peak — a 3rd oral JAK enters Year 3, price erosion accelerates Probability-weighted peak = 0.55×1.61 + 0.20×2.00 + 0.25×1.00 ≈ $1.29B

Step 7 — Sensitivity Analysis (Week 4)

Tornado-rank the Base Case drivers to find where validation effort should go first. Peak Share ±3pts → Revenue ±$370M (widest bar — #1 priority) Net Price ±15% → Revenue ±$240M Persistency ±10pts → Revenue ±$246M Peak share is the single biggest swing factor — the pre-launch validation ask should be a dermatologist ATU (Attitude & Usage) study to narrow that range before this number goes into planning.

Step 8 — Present It (Week 5 how-to below)

Headline slide: "$1.29B probability-weighted Year 5 revenue (Base $1.61B, range $1.00B–$2.00B)." Method slide: Analogue Method anchored to the first oral JAK's real launch curve, cross-validated against the epidemiology funnel. Assumptions table: the 5 drivers above with sources. Sensitivity: peak share is the #1 risk. Validation ask: a dermatologist ATU study, plus the analogue's real-world persistency data once its post-launch adherence study publishes.

📋 How-To: Structuring Your Forecast Presentation

What it is: the order a CFO or investment committee actually wants to hear a forecast in — not the order you built it. Lead with the number and range, not the methodology.

1. Headline number + P10–P90 range (or scenario range) — one slide, no build-up
2. Method used and why it fits this situation (1 slide)
3. Key assumptions table — the 4–5 drivers that matter most, with source citations
4. Sensitivity — which single assumption moves the forecast most if wrong
5. Validation ask — the one data point or study that would most increase your confidence before this goes into planning

Rule of thumb: if a reviewer can't restate your headline number and its single biggest risk after your first two slides, the structure has failed — regardless of how sound the underlying model is.

6 Real Models You Will Build

✅

Oncology

Epidemiology-driven patient funnel · PFS/OS endpoints

✅

Immunology

Biologic share curves · Biosimilar erosion modelling

✅

Dermatology

Prevalence-based sizing · JAKi vs biologic split

✅

MedTech

Procedure-volume model · Capital vs disposables mix

✅

Consumer Health

Retail offtake model · Rx-to-OTC switch scenarios

✅

General Pharma

Integrated brand P&L · Launch curve · LOE scenario

Best For

🔎

CI Professionals

Add quantitative rigour to your intelligence outputs

📊

Analysts

Move from data to decision-grade financial models

🏥

Market Access Teams

Translate HTA evidence into payer-relevant revenue scenarios

🌱

Pharma Freshers

Build an industry-ready skill set from day one

🚀

Professionals Targeting a Salary Jump

Forecasting fluency is among the highest-valued skills in pharma commercial — use it to negotiate and level up

Ready to Build Your First Forecast Model?

Pick your forecast mode and technique, enter your assumptions, and the Forecast Builder will run the real calculation and write up the rationale — end to end, no spreadsheet required.