Institutional Diligence Infrastructure

The Proof-of-Record Workspace for Biotech Investment Decisions.

Preserve deal context, asset scenarios, material assumptions, evidence, decisions, audit history, and signed export receipts in one governed workspace — callable by analysts and AI agents.

12,728+

clinical transitions

264

benchmark cells

3

validation studies

Signed Dossier Receipt

NSCLC small-molecule valuation

verified

Content hash

7f4c...91ad

Engine

2.8.0

Methodology

methodology@2026-06-23

Dataset

pos-benchmarks@2021

42

Sources cited

111

Drugs backtested

11

Indications benchmarked

Agent-Callable Substrate: verify_export resolves the same receipt over JSON and MCP before a downstream model relies on the number.

The Platform

A system of record across a deal’s life.

From inbound triage to a signed, verifiable dossier — one governed workspace where every assumption keeps its provenance and every action lands in the audit trail.

  1. Intake & Triage

    Paste or drop a deck; the screener extracts and cites company, deal, and asset facts, then returns a worth-the-look verdict.

  2. Asset Workspace

    The deal becomes a workspace — scenarios, contributors, notes, and an Evidence Register with cell-level provenance.

  3. Valuation

    Deterministic, version-stamped rNPV under the published methodology. No language model in the math path.

  4. Signed Dossier & Verify

    Export an IC-ready, content-hashed PDF or Excel; any reviewer verifies it at the public endpoint — no account required.

  5. Monitoring

    Watchlisted assets re-score when a competitor reads out, a milestone lands, or a comparable transacts.

    In development

Triage Co-Pilot

Five-minute inbound triage: a grounded, cited advance/watch/kill verdict over a deck.

Asset Workspace

The deal record — scenarios, contributors, notes, and the evidence behind each number.

Evidence Register

Cell-level provenance for every load-bearing assumption: source, date, excerpt, confidence.

Valuation Engine

Version-stamped rNPV and Monte Carlo, computed deterministically on the backend.

Signed Exports & Verify

Content-hashed PDF and Excel dossiers, verifiable at the public endpoint with no account.

Audit Trail

Material actions — created, changed, exported, denied — recorded and asset-scoped.

Asset Landscape

Comparable trials, competing programs, and sponsor priors for the specific asset under review.

MCP Access

Query the engine, benchmarks, and verification from an AI agent over the public MCP surface.

PoS priors span 11 indications, 8 benchmark modality baselines, and 3 biomarker strategies, with expanded modern modality labels normalized before lookup. Source: BIO/QLS 2021 clinical-development transition data.

Built for institutions

Diligence is more than the methodology.

Methodology matters, but diligence also depends on governed access, isolated tenancy, and a record of who changed what.

Multi-Tenant Workspace

Every organization's deals are isolated; pilot data lives in its own org and never shares a demo workspace.

Role-Governed Access

Organization roles gate who can read, write, and administer. Missing or unknown roles fail closed to viewer.

Audit Trail

Material actions — created, changed, exported, or denied — are recorded against the asset they touch.

Independently Verifiable

Every export is content-hashed and checkable at the public verify endpoint — no recipient account required.

Security Posture

Security Posture For Private Diligence.

PhaseFolio separates implemented tenant controls from planned external attestations, so investors can see what protects confidential deal work today and what is moving into formal review.

Read the security posture →

Role-Based Access

Live App Layer

Organization roles govern read, write, and admin access. Unknown or missing roles fail closed to viewer-level permissions.

Tenant Isolation

RLS Enforced

Product app traffic uses a non-bypass PostgreSQL runtime role and row-level security policies keyed to the active organization.

Audit And Verification

Live Trust Layer

Material actions are recorded against assets, and signed exports resolve through public verification so recipients can inspect provenance.

Compliance Roadmap

Planned Attestation

SOC 2 Type I readiness is the first external-attestation target, followed by a SOC 2 Type II evidence window. ISO 27001 and HIPAA readiness will be evaluated as customer and data scope require.

Machine-readable substrate

Built for human and agent workflows.

PhaseFolio is a proof-of-record diligence workspace for biotech investment decisions — run by the analyst on the deal, and callable by an AI agent over the same durable record.

An MCP server, a public verify endpoint, an llms.txt index, and structured public data let an agent confirm a number’s provenance before relying on it — so the GP, analyst, or CMO reviewing its work can stand behind the result. No account required.

Source Register

Evidence you can trace.

The register below indexes core public sources behind PhaseFolio’s base rates, multipliers, and backtest library. Signed dossiers carry their full source receipts; every row here links to evidence a reviewer can inspect.

Transition Base Rates

2 sources

01
Thomas et al. (2021)

Clinical Development Success Rates 2011-2020 · BIO · QLS Advisors · Informa · N=12,728 transitions

02
Wong, Siah & Lo (2019)

Estimation of clinical trial success rates · Biostatistics 20(2)

Signal & Multiplier Evidence

6 sources

03
Minikel et al. (2024)

Genetic evidence and clinical success · Nature 629

04
Zhou et al. (2025)

Dynamic clinical trial success rates for drugs in the 21st century · Nature Communications 16

05
Schwaederle et al. (2016)

Biomarker-based treatment strategies · JAMA Oncology 2(11)

06
Parker et al. (2021)

Biomarker use and clinical-trial failure risk in oncology · Cancer Medicine 10(6)

07
Haslam et al. (2022)

Predictors of success for oncology drugs tested in phase 1 trials · Int J Cancer

08
Vreman et al. (2019)

Evidence weighting for conditionally approved drugs · Clin Pharmacol Ther 105(3)

Validation

1 source

09
PhaseFolio Back-Test Library (2026)

3 public validation studies with per-drug outcome labels and backtest methods

Full methodology, worked examples, and source citations are available in the methodology paper.

Who Is This For

Public methodology, governed access.

Methodology, backtests, and verification remain public. Engine access and signed-export issuance are governed through research and early-access partnerships.

Public Methodology

Open research surfaces

Methodology pages, backtest reports, public benchmarks, and the verify endpoint are available without authentication.

View Methodology

Early Access

VCs, family offices, biotech BD

Full engine access, signed-export issuance, evidence register workflows, asset-anchored landscape, and co-development on indication-specific cohorts.

Request Access
Methodology LedgerRelease 64816fd · deployed 2026-06-24

A public version trail.

PhaseFolio is working with an initial cohort of design partners. Substantive methodology changes are published with version history, every issued export carries a signed hash, and each public methodology page preserves its version trail.

Recent Changes

06/23Expanded modality normalization methodology — PoS Calibration now discloses how RNA therapeutics, protein degraders, radioligands, oncolytic viruses, and cancer vaccines map to existing PoS benchmark baselines; IRA Framework separately discloses Year 9/Year 13 cliff defaults for those labels. rNPV wizard and client-side IRA helper now share the expanded vocabulary; no PoS dataset, matrix cell, or engine version changed (methodology@2026-06-23)
06/21Audit Trail Phase 3 integrity — public.audit_events is DB append-only via UPDATE/DELETE/TRUNCATE trigger + runtime-role revokes, audit writes now carry per-org SHA-256 hash-chain fields with chain_format_version v1, and admin chain verification detects tampering while tolerating legacy unchained rows
06/19Clinical Stage Normalization surfaced publicly and in-product — wizard/results disclosure for exact labels → canonical PoS gates, /methodology/stage-normalization, REST/MCP methodology discovery, sitemap, llms.txt, OKF bundle, version history, and changelog all wired to stage-normalization@2026-06-19. Label-only PoS effect remains none; explicit cost/timing can still move rNPV.
06/19rNPV engine 2.8.0 — clinical stage normalization compute contract: optional development_stage_plan separates cost/timing work packages from canonical PoS gates, with legacy stages[] scenarios pinned bit-identical and methodology token stage-normalization@2026-06-19
06/19rNPV Engine methodology — the core risk-adjusted-NPV calculation (cumulative PoS, midpoint-discounted risk-adjusted costs, S-curve ramp/plateau/LOE revenue net of COGS/SG&A/tax, WACC discounting, Monte Carlo / expected NPV, tornado sensitivity) now has a dedicated, citable section at /methodology/rnpv; the methodology hub leads with the engine and consolidates the two JHTV cards. Engine math unchanged (methodology@2026-06-19)
06/17Development-path methodology — path labels now carry explicit methodology tier and automatic-effect policy (baseline / structural template / analyst-confirmed / display-only); no path-based PoS, revenue, cost, duration, or launch-timing default by label alone. Engine math unchanged (methodology@2026-06-17)
06/16rNPV engine 2.7.0 — Phase B deep-triage provenance threading: reviewed biomarker_quality sources now travel through scenario recompute into signed-export signal receipts; attribution-only engine bump, no deterministic math change
06/15Enterprise-platform site: /diligence renamed to /platform as the product surface (deal-workflow spine, named module grid, built-for-institutions governance band), homepage restructured around the platform story + human/agent-workflow positioning, Security added to nav; IA/copy/presentation only, no engine/methodology/dataset version change
06/12Public substrate apex migration + research IA restructure — methodology/research/benchmarks moved to the phasefolio.com apex (trust surfaces stay on app), /evaluations → /research/evaluations, back-tests consolidated onto /research/backtests/[cohort]; URL/IA only, no engine/methodology/dataset version change
06/08Triage deck-intelligence capture — deck-asserted economics, deal-sanity check, rNPV pre-fill, clickable history
06/07Triage Co-Pilot v1 — unstructured text clinical context extraction, 3D PoS lookup calibration, indication landscape aggregation, and provenance-linked conditional reads via synthesis_prompt@2026-06-07
06/07Backtest methodology — engine-provenance correction: the published RA / NSCLC / antimicrobial backtests are now labeled by the engine version that computed them (1.0.0 base BIO/QLS path; antimicrobial adds the Sprint-1 M3 multiplier), with an invariance note that the AUCs are unchanged through the current engine 2.6.0. Research-report footers no longer auto-drift to the live engine version. No engine math changed (methodology@2026-06-07)
05/30Asset-landscape comparator relevance — match scoring narrows to the project's sub-indication (e.g. cUTI, not the broad parent), so off-target trials score low instead of crowding the pool; candidate pool ranked by relevance + recency before the cutoff. Not an engine bump — a landscape display score (landscape-match-quality@2026-05-30)
05/29Citation integrity pass — ~80 references across methodology, benchmarks, and research surfaces re-verified and made clickable (DOI / official source); COGS, SG&A, and IRA citations re-sourced after an audit caught misattributions. Engine math unchanged (methodology@2026-05-29)
05/28Evaluations framework v1 — public /evaluations surface publishing signal verdicts: biomarker_quality (validated, shipped) and Phase 1 ORR (not predictive, never shipped). Machine-readable manifest at /evaluations/verdicts.json
05/28rNPV engine 2.6.0 — drug-specific clinical signal layer: biomarker_quality now scores PoS in oncology solid tumor at Phase II/III (+5.2pp held-out AUC). Phase 1 ORR is extracted and stamped on exports but does not score the engine after Phase 0 found it double-counts biomarker quality. methodology@2026-05-28
05/26rNPV engine 2.4.0 — SEC EDGAR cohort classifier: per-row cohort_indication populated at ingestion (deterministic rule + LLM verdict; disagreement or ambiguous → human-review gate). Comparator pool surfaces only CMO-defensible deals; methodology@2026-05-26-v2
05/26rNPV engine 2.3.2 — SEC EDGAR wizard suggest: deal-structure pre-fill moved from engine to wizard with gold-banner consent + 'clear suggestions' action. Full Ownership is now a hard 'no deal' signal (no silent SEC overlay)
05/25Results-page substrate-trust hero — P10–P90 confidence cone under rNPV (gated on fresh Monte Carlo + not deal-stale), validation strip linking to the indication's cohort backtest, engine + methodology version chip in header
05/25SEC EDGAR deal-terms substrate v1 — public 8-K Item 1.01 comparator pool with accession-number provenance; live at /methodology/sec-deal-terms (sec_deal_terms@2026-Q3)
05/21SG&A + Tax benchmarks — sga_pct keyed to commercial model (orphan vs primary-care vs specialty), 21% statutory tax auto-fills; methodology@2026-05-21 with overfitting-governance gate
05/20Modality COGS benchmarks — wizard cogs_pct anchored to sourced cost÷reference-price ratios per modality (gene therapy ~13% bimodal); methodology@2026-05-20
05/19Antimicrobial backtest published — AUC 0.629 across 36 Phase 3 entrants (2004–2019); failure-only candidate signals demoted out of the score with full ablation made public
05/18rNPV engine 2.1.0 — deal-adjusted Pro-Forma Cash Flows: royalty/milestone/profit-split assets now show the partnered position (terminal pinned to the deal-adjusted Unadjusted NPV) across the results page, share view, and signed PDF/Excel exports
05/17rNPV engine 2.0.0 — adversarial audit pass: deal-aware Unadjusted NPV, corrected patent-cliff & market-exclusivity modeling, robust milestone-gate resolution, and an engine-authoritative value-progression curve
05/06JHTV public-data case study — ~602 inventions scored, top 10 deep-dives with rNPV envelopes (jhtv-portfolio@2026-Q2)
05/05Claude Code plugin shipped — install via TamalAdebisi/phasefolio-claude-plugin (5th distribution surface, HTTP transport direct to /api/mcp)
05/03MCP server polished to production-grade — annotations, parameter descriptions, output schemas (Smithery quality 73 → ~95)
05/03MCP distribution — npm wrapper, Official Registry + Smithery submissions, install snippets on /methodology + /diligence
05/03MCP server protocol handshake (initialize/notifications/ping) + JSON Schema serialization for tools/list
05/03Two-product framing — Research / Fund Tier rebrand + /contact + /diligence pages
05/02Smoke-test sweep — landscape filter wiring, Excel IB green xrefs, /api/v1 discovery, NDA cross-surface fix
05/01Substrate MVP — REST API + 9-tool MCP server + signed exports + /verify + methodology hub
04/30Asset-anchored Landscape view (Screen 3 of the 4-screen VC experience)
04/29IRA terminal-value modeling — Y9/Y13 MFP cliffs + closed-form FE slider
04/28Evidence Register MVP — cell-level provenance for load-bearing assumptions
04/27IC Dossier export — 6-section PDF + 5-tab IB-grade Excel
04/27Wilson 95% CI on RA back-test — both threshold cuts on the public report
04/21Pre-launch hardening — RLS policies, schema canonicalization, MC per-iter PV, bcrypt, CSP
04/20Per-sub-indication backtest report page (Rheumatoid Arthritis cohort published)
04/19Intelligence flagship dashboard — disease-specific deep dives
04/17Security audit — 4 criticals + 5 highs hardened (RLS, plan gates, webhook idempotency)