← Back to home

The Science Behind the Platform

Your Network Is
Your Fundraise

We didn't build another CRM. We built on decades of network science to turn relationship chaos into your competitive edge.

↓ Explore the science
83%of successful referrals traverse weak tiesGranovetter, 1973
4.83average hops between any two VCsWatts & Strogatz, 1998
lower failure rate with top-networked VCsHochberg et al., 2007

Two Metaphors. Every Design Decision.

Architecture

The CastleFort

Your data stays within walls. Knowledge compounds behind the moat. Security, multi-tenant isolation, and local AI processing are architectural, not afterthoughts.

Continuity

Institutional Memory

Fundraising is episodic — relationships are continuous. Round N+1 starts with the full knowledge of rounds 1 through N.

“The warm introduction you need is closer than you think.”
Built on 9 peer-reviewed papers across network science, VC research, and AI

Network Science

The Research That Drives Our Design

Our approach is grounded in decades of peer-reviewed research. Here are the four ideas that matter most.

83%

The Strength of Weak Ties

Your acquaintances — not your close friends — open access to entirely new investor clusters. We map your entire team's network, not just the close relationships.

Granovetter (1973), Burt (1992)
~5 hops

Small-World Navigation

Any VC is reachable in 2–4 introductions through shortcut edges. Our pathfinding models the distance-reducing potential of each hop, not just abstract shortest paths.

Watts & Strogatz (1998), Kleinberg (2000)
67%

Network Position Predicts Success

An investor's network position accounts for 67% of predictive power in outcome models. Companies backed by top-networked VCs have 2× lower failure rates.

Hochberg (2007), Bubna (2020)
3 layers

Three Networks, One Picture

Co-investment, affiliation, and social networks are distinct layers with independent signals. Most tools only capture one. We model all three.

Shi (2025), Poole (2025)

Want the full research? Dive deeper.

Institutional Memory

Knowledge That Compounds

Fundraising is episodic — relationships are continuous. Your platform remembers what you've learned, strengthens what you use, and lets go of what you don't.

Relationships That Strengthen

Connections follow Hebbian learning. When two entities co-appear in news, co-invest, or are mentioned together, their connection strengthens. Active relationships naturally rise to prominence — round N+1 inherits the full signal of every prior round.

Intelligent Decay

Unused connections decay — but not by simple time. A connection that was exercised weekly and goes silent for a month fades faster than one that was always quarterly. Decay is relative to expected activity, so your institutional memory stays honest.

The Full Picture

Co-investor cliques form communities. Investor communities form clusters. Bridge nodes and weak ties connect the clusters. Your platform maps it all — and between the filaments of activity, it finds the voids where new bridges create the most value.

Community Intelligence

Smarter Together, Private Always

Every founder who uses CastleFort makes the platform smarter for everyone — without ever exposing private deal data.

Architecture

Canonical + Overlay

A shared knowledge commons grows with every user. Your private annotations, notes, and deal context stay entirely yours. The canonical layer improves; your overlay remains invisible.

Contribution

Anonymous by Default

When you research an investor, verify a connection, or correct a data point, that knowledge flows back anonymously. Individual contributions are never attributable; collective intelligence compounds.

Confidence

Multi-Signal Scoring

No single source is trusted alone. Community-contributed data, public records, and AI extraction are weighted independently. Confidence increases with source diversity, not volume.

How We Build

Four Design Principles

Architectural choices that shape what the platform can do and how much you can trust it.

01

Epistemic Humility

Every data point carries a confidence score with explained provenance. The UI never lets you mistake speculation for fact.

02

Absence Taxonomy

Missing data isn't a black box. Every gap is classified: undiscovered, private, nonexistent, or expired.

03

Entity Resolution

The same person across CRM, news, and SEC filings — resolved with quantified confidence.

04

Privacy by Architecture

Multi-tenant isolation, local AI processing, and zero-knowledge defaults. Security is structural, not a policy checkbox.

See the Science in Action

Every theory on this page maps to a real feature. Let's talk.

Get in Touch