Uncovering hidden signals of value: the power of a portfolio relationship intelligence system

Every venture fund claims to have a strong network, however very few can describe its boundaries, density, or blind spots. Over the past 12 months, we mapped every business relationship connected to the portfolio companies across our three funds, turning a common claim into something we can audit. The premise is simple: a network that cannot be measured is an asset that cannot be managed, and gaps that cannot be seen are the ones that will cost you.

Building out the map

The work didn’t start with a graph, but a taxonomy. Before mapping anything, we defined the nodes and edges of our investment landscape, including the distinct kinds of relationships connecting them. Large language models excel at extracting entities from text, but they struggle to determine which distinctions matter to a specialist fund. That judgement must come from us. For example, a relationship labelled ‘customer’ carries different strategic weight than one labelled ‘distribution partner.’ Collapsing these distinctions, as most CRM systems do by default, destroys the signal that makes a network map valuable.

The taxonomy runs across numerous axes, including categorical (sector and industry hierarchy), institutional, geographic, and relational. Each axis answers a different strategic question, which can be interrogated directly. Because the map is fed by the fund’s regular information flow, it remains up to date without the need for manual censuses, unlike every network-mapping effort that came before it.

What the data revealed

The graph now tracks over 3,800 organisations and 5,300 relationships across our portfolio, with these figures continuing to grow. Nearly half (47%) are in healthcare, which confirms a deliberate concentration in our area of focus. The rest are in adjacent sectors, such as Cloud, data infrastructure, and finance, where portfolio companies actually do business. Approximately 30% of the tracked relationships are revenue-generating customer connections, proving that this network is not social capital in the abstract, but commercially load-bearing.

Density also reveals where the map runs thin, and the useful distinction is between thin-by-design and thin-by-accident. Light coverage of consumer-branded partnerships is thin by design and sits outside of our thesis. However, thin coverage of a geography where portfolio companies are actively trying to sell is a genuine gap that we can now measure and deliberately close, instead of learning about after it costs us.

With more than fifty portfolio companies across three funds, it is impossible for any one partner to know all the relationships. The map revealed portfolio companies that shared a customer or distributor, and cases where a Fund I relationship turned out to be the exact introduction a Fund III company needed. Most recently, the map opened a route into a new long-term care market for one company, and has matched numerous others with advisors, fractional CFOs, and Series A investors that they would not have otherwise thought to reach.

The results of increased visibility

This level of visibility has improved three areas of our operations. In sourcing, for example, we can identify genuinely differentiated positioning earlier because the map shows which competitive spaces are sparse rather than crowded. In market intelligence, shifts in relationship formation have become measurable leading signals, rather than trends we only notice when they surface in lagging survey data. In resource allocation, we now distribute partner time to the most impactful introductions.

Today’s map will become tomorrow’s training data, and we are building a forecasting layer on top of the graph, paired deliberately with model surveillance, interpretability, and scoring governance. In healthcare, where the cost of an opaque model extends beyond money, interpretability is not a feature but a necessity. None of this replaces human judgement and the map only records what the fund can observe, so will always underestimate informal ties. Relationship quality remains difficult to showcase, meaning that a network map should be used as a decision support tool, not a decision maker. It does not replace the specialisation and service that build real relationships, it simply gives us a new way to see what was always there.

See Also

Webinar: women in AI – lessons from female founders

One final observation is that this instrument rewards specialisation and would help a generalist firm far less. Our taxonomy only works because we know which distinctions matter in healthcare, such as which relationship types predict commercial traction and which regulatory bodies function as gatekeepers rather than counterparties. The graph is legible only because our portfolio is vertically concentrated. Fifty companies in one domain share customers, distributors, and counterparties, producing a network dense enough to reveal structure and gaps. If you run the same exercise across a generalist portfolio, you will have breadth without density – a map of everywhere that shows you nothing.

As it turns out, network intelligence compounds the same way expertise does: only when it is focused.