Document intelligence with a modeling engine and memory

Read any private-capital document.
Model it.
Find the leverage.

Reading a document is the commodity. Modeling it on a deterministic engine, then benchmarking it against a growing memory to model and find the leverage — what’s market, what’s unusual, what’s missing — is the hard part. That’s what we built.

Before · Unstructured prosePPM · p.47 of 82

The Management Fee shall be payable quarterly in advance and shall be equal to one and one half percent (1.50%) per annum1 of aggregate Commitments during the Investment Period… and shall be further offset by seventy-five percent (75%)2 of transaction, break-up, monitoring, directors’ and consulting fees received by the General Partner…

…second, to the Limited Partners until each has received a Preferred Return of eight percent (8%) per annum, compounded annually3; third, a catch-up to the General Partner; and thereafter, eighty percent (80%) to the Limited Partners and twenty percent (20%)4 to the General Partner.

After · Structured variablesProvenance 100%
1mgmt_fee.rate1.50% p.a.p47 ¶2
2mgmt_fee.offset75% of feesp47 ¶3
3waterfall.pref8% compoundedp48 ¶1
4waterfall.carry80 / 20p48 ¶1

Every term in every document — fees, waterfalls, covenants, gates, lock-ups, triggers, side-letter rights, missing protections — becomes a structured variable, traced to its exact page and paragraph. We read the deal’s logic, not just its numbers, and stand it up as a digital twin you can model, benchmark and run forward — so strain on cash or covenants is visible before it bites. One document or a hundred, a fund or a direct deal.

The Problem

The economics are in the document. They just aren’t usable yet.

A private-capital document hides its economics in prose. Fees, waterfall, commitments, the conditions that change all three — written to be read by a lawyer, not modeled by an analyst. So three things go wrong before anyone has done any analysis.

Numbers you can’t trust

A figure you can’t trace back to a term is a figure you can’t stand behind. Until it’s modeled, the economics stay locked in the prose.

Benchmarks locked in heads

“Is this market?” gets answered from memory and the last few deals someone happened to see — not from evidence.

What’s missing never shows up

The protection you didn’t get, the term a peer secured and you didn’t, the clause that should be there and isn’t.

That’s the gap between reading a deal and being able to act on it.

Named by academia. Recognized by ILPA. Priced in by the secondaries market. Four limitations, one root cause: nobody models the documents. That is the layer we built.

The limitationsDealAnalysis
No standardized termsevery PPM uses different language for the same economics
Structured variablesevery term mapped to one schema
No comparabilityno benchmark exists across funds, vintages, strategies
Growing memorypercentile positioning on every term
Sparse, lagging dataprivate companies report quarterly, partially, and late
Modeled reconstructionpartial inputs completed to a working model, each estimate labeled
Opaque attributionGPs claim alpha; LPs can’t verify skill vs market
PME attributionskill, timing, and market decomposed

Virtually everything sold as a fact was not quite so.

Ludovic Phalippou · Oxford Saïd Business SchoolMost-cited academic in PE performance and fee structures · author of “Private Equity Laid Bare”
What it does

Read the logic. Model it. Benchmark it. Find the leverage.

Drop in a document — a PPM, an LPA, a model — and get the structured, modeled, benchmarked, flagged read back in minutes.

01

Read the logic

We read the conditions, rights and obligations a deal encodes — the structure of the economics, not only the figures on the page. The same discipline reads the logic of a model, not just its cells.

02

Model it

Provable figures, computed and traceable to the term they came from. Same inputs, same outputs, every time — reproducible, auditable, built to be certified. No model guesses where the math should be.

03

Benchmark it

Every term and projection located against a growing memory of the market: what’s standard, what’s off-market, what’s missing. Anomaly and missing-term detection — and it sharpens with every deal it sees.

04

Find the leverage

Forecasts bounded against the empirical record and shown as ranges with the confidence attached — never a single figure pretending to certainty. The off-market terms, the gaps and the negotiating room, surfaced.

Audit trail · Reproducible insight

Every number can be auditably tracked back to source. Every insight is reproducible.

Who it’s for

One engine, both sides of the table.

GP-side

Managers and their deal teams

  • Read an incoming deal in full and model the economics before you’re at the table.
  • See where the negotiating leverage is — what’s off-market in your favor, what you’re missing, what to ask for.
  • Carry the same model through the portfolio.

LP-side

Allocators · family offices · foundations · their advisors

  • Diligence a fund or a direct commitment on provable numbers, not the GP’s summary.
  • Benchmark the terms against what your peers are getting.
  • Monitor what you hold on a model that updates — not a PDF that ages.
The Scope

One engine. The full private-capital instrument set.

Twelve deal types across three analytical tiers, built on one engine and one set of primitives — different instruments, the same provable spine. This is the analytical scope the engine is designed around.

Tier 1Modelable outcomes
Direct Dealasset cashflows, scenario-testable inputs
Private Creditcovenant stress-tests, position in credit stack, priority waterfall
GP StakesAUM trajectory, fee & carry income DCF
Co-Investmentasset economics + thin cost layer
Tier 2Engineered payoffs
Structured Productconditional waterfalls, trigger scenarios
Secondary (LP-led)NAV discount → IRR, fee drag, breakeven
Continuation / Striprollover-vs-exit, carry crystallization
Tier 3Manager-mediated
SMAbespoke bilateral constraints
Primary Fund (PFC)terms modeling, J-curve, unfunded exposure
Fund-of-Fundsfee stack, look-through, vintage pacing
Platform Wrapperwrapper drag, look-through, jurisdictional routing
OverlayNot standalone
Side Letterstrue terms = base + overlay · fee discounts, enhanced liquidity, MFN, excuse rights, carve-outs. Modeled as base terms plus the modification layer — never in isolation.
Why DealAnalysis

The reader is the commodity. The engine and the memory are the company.

The moat

Anyone can pull a term out of a document now — the reader is the commodity. What can’t be copied is an engine that turns those terms into provable cash you can stand behind, and a memory that already knows what market looks like. That pairing is the company.

And it compounds. Every deal the platform reads makes the growing memory a sharper benchmark for the next one — the product gets better with use, and the advantage widens with it. New corners of the market are captured as data, not rebuilt as new logic, so coverage grows the way data grows.

The Team

Domain, technical, and institutional credibility.

A financier who builds technology, rare. A technologist who understands finance, rare. Together: unicorn rare.

CEO · Co-Founder

Stuart Fotheringham

Twenty years public and private equity capital. Previous business exit to Rothschild & Co., where he served as Managing Director. Former VC. Co-inventor of Citi’s financial forecasting patent.

A dealmaker who architects technology — designed the methodology, not just identified the problem.

CTO · Co-Founder

Kevin Fletcher

20+ years enterprise fintech at tier-1 financial institutions. Architected systems processing billions daily — including HFT platforms for hedge funds. Expert in building platforms for regulated environments.

An engineer who thinks in financial models — built for fiduciary-grade, not just enterprise-grade.

Programs & Recognition

Validation

  • Fintech Sandbox residency
  • Venture Studio partnership
Trust & control

Built to be certified. Built to keep you in control.

Architectural commitments — not policy promises. The platform proves what’s provable, calibrates what’s estimable, and leaves the deciding to you.

Built to be certified

The figures the engine computes are traceable to the term they came from and reproducible on demand — an engine built to be certified, not asserted to be trustworthy.

  • Built for SOC 2 Type II certification — architectural foundation in place
  • ISO 27001 · ISO 27701 — security and privacy, on the same path
  • Cyber Essentials Plus — targeted; more as we enter new markets
  • Every change traces to a ticket; every deploy reproducible

Every number is labeled

You always know what you’re standing on.

  • Each figure is observed, estimated with a confidence range, or computed — and says so
  • Provable numbers never wear forecast clothing
  • Forecasts never borrow provable language

The human stays in control

The platform surfaces; it doesn’t decide.

  • Probable outcomes across scenarios, laid out for comparison
  • A person — not the platform — chooses and acts, and can audit every provable number behind the choice
  • We don’t interpret genuine ambiguity; that stays with you and your counsel

Security & infrastructure

Private-tenant by design — NDA-safe by architecture.

  • Microsoft Azure · private tenancy · regional data residency
  • Tenant isolation · client data segregated; dedicated tenancy where required
  • AES-256 at rest, TLS 1.3 in transit, SSO + MFA + RBAC
Planned

A live layer that watches a portfolio and raises an alert when covenant, cash or revaluation risk moves — automatically, as new information arrives, with the human still deciding what to do about it. Shown as planned, because it is.

FAQ

Questions, answered.

What is DealAnalysis?

DealAnalysis is document intelligence with a modeling engine and memory, built for private capital. It reads any private-capital document — a PPM, an LPA, a credit agreement, a model — turns its terms into provable figures on a deterministic engine, and benchmarks them against a growing memory of the market, so you can model and find the leverage: what is market, what is unusual, and what is missing.

How is DealAnalysis different from ChatGPT, a large language model, or a RAG tool?

Reading a document is the commodity; DealAnalysis is the part that comes after. A language model can summarize a contract, but it cannot model the economics deterministically or tell you how a term compares to the market. DealAnalysis stands the deal up as a model whose figures are traceable to the term they came from, then benchmarks every term against a growing memory of the market. The reader is the commodity; the engine and the memory are the company.

What does DealAnalysis do?

It reads the logic of a deal, models it, benchmarks it, and surfaces the leverage. Drop in a document and get a structured, modeled, benchmarked and flagged read back in minutes: what is market, what is off-market, what is missing, and where the negotiating room is. Every figure is labeled as observed, estimated with a confidence range, or computed.

Who is DealAnalysis for?

Both sides of the private-capital table. On the GP-side, managers and deal teams read incoming deals, model the economics, and find the leverage before the negotiation. On the LP-side, allocators, family offices, foundations and their advisors diligence funds and direct commitments on provable numbers, and monitor what they hold on a model that updates rather than a PDF that ages.

What deal types and documents does DealAnalysis cover?

One engine spans the private-capital instrument set across three analytical tiers: direct deals, private credit, GP stakes and co-investments; structured products, secondaries and continuation vehicles; SMAs, primary funds, fund-of-funds and platform wrappers — with side letters modeled as base terms plus their overlay. This is the analytical scope the engine is designed around.

Is my data kept confidential and secure?

Yes — it is built for institutional procurement. Each client runs in a private, tenant-isolated environment on Microsoft Azure with regional data residency, AES-256 encryption at rest and TLS 1.3 in transit, and single sign-on, multi-factor authentication and role-based access control. The platform is built for SOC 2 Type II certification, with ISO 27001, ISO 27701 and Cyber Essentials Plus on the same path.

Does DealAnalysis use AI, and can I trust the numbers?

Every number is labeled, and a person stays in control. The platform tells you which figures are observed, which are estimated with a confidence range, and which are computed; provable numbers never wear forecast clothing. It lays out probable outcomes across scenarios, but a person — not the platform — chooses and acts, and can audit every provable number behind the decision.

How can I see DealAnalysis?

Through an NDA-gated demo on your own documents. Bring a PPM, an LPA or a model, and we will read it, model it, and benchmark it back to you. You can request a demo or the investor deck, both under NDA, from the contact form.

Who is behind DealAnalysis?

DealAnalysis was co-founded by Stuart Fotheringham, Chief Executive Officer — a former Managing Director at Rothschild & Co and co-inventor of a Citi financial-forecasting patent — and Kevin Fletcher, Chief Technology Officer, an enterprise fintech engineer who has architected trading and processing systems for tier-1 financial institutions.

See it

The interesting part, we show you under NDA.

You’ve seen what it does. How the engine stays provable, reproducible and certifiable — the part that’s genuinely hard, and genuinely ours — is what we walk you through under NDA, on your own documents. Bring a PPM, an LPA or a model. We’ll read it, model it, and benchmark it back to you.

A demo

Request an NDA-gated demo

A working session on your own documents — what the engine reads, models and benchmarks on your actual deal types. We respond within one business day.

The deck

Request the investor deck

The full pre-seed materials, shared under NDA.

Your information is never shared. Confidential materials are released under NDA only.