Is ONA employee surveillance?

It doesn't have to be — and the distinction matters. Passive ONA mines email metadata, calendar data, and chat logs, which raises real consent and works-council issues. Active ONA asks people directly ("who do you go to for advice?") and analyzes the declared ties. OrgTool implements the active model only: ties are entered deliberately, the graph math runs locally on the planner's machine, and no communications data is ever collected.

What do the five network dimensions mean?

Advice (who do you consult to get work done), trust (who would you share a half-formed idea or concern with), energy (who leaves you more energized after an interaction), information (who keeps you informed about what matters), and influence (whose opinion moves decisions). The same pair of people can be connected in one dimension and strangers in another — which is exactly why one aggregated "collaboration score" hides more than it reveals.

Is network centrality an AI black box?

No — it's deterministic graph mathematics, decades older than modern machine learning. Betweenness centrality (Freeman 1977, computed efficiently via Brandes 2001) counts how often a person sits on the shortest paths between others. Same ties in, same numbers out, every time — fully explainable to a works council, an executive committee, or a court.

Network analysis

ONA: seeing the company behind the org chart

OrgTool Learn · 10 min read · Updated July 2026

Every organization is two organizations. There's the formal one — the reporting lines on the chart — and the informal one: who actually asks whom for advice, who is trusted with the unpolished truth, whose desk every decision quietly crosses. Krackhardt and Hanson named this thirty years ago in the Harvard Business Review: the informal network is "the company behind the chart." Organizational network analysis (ONA) is simply the discipline of making that second company visible before you make decisions that depend on it.

The research canon, in four ideas

Weak ties carry the news

Stanford sociologist Mark Granovetter's "The Strength of Weak Ties" (1973) — among the most cited papers in all of social science — showed that novel information travels disproportionately through weak connections: the acquaintance in another department, not the teammate you talk to daily. Strong ties cluster; weak ties bridge. For an organization, this means the people connecting distant clusters are carrying information no process document accounts for.

Brokers span the structural holes

Ronald Burt's "Structural Holes" (1992) formalized the value of those bridges: people who span gaps between otherwise disconnected groups — brokers — consistently show higher performance, better ideas, and faster promotion, because they see combinations others can't. Structurally, they are also single points of failure: remove the broker and the two groups stop hearing each other.

The informal org does the work

Krackhardt & Hanson (HBR, 1993) and later Rob Cross & Andrew Parker's "The Hidden Power of Social Networks" (2004) brought this into management practice: advice networks predict how problems actually get solved, trust networks predict how change actually gets adopted, and both routinely contradict the formal chart. Cross's collaborative-overload research adds the warning label: the most connected people are often the most overloaded, and organizations systematically under-recognize them.

Energy is a network, too

Later ONA practice added dimensions beyond information flow — notably energy (who leaves others more energized) — after research found energizing relationships predict performance and retention in ways raw communication volume doesn't.

The five dimensions worth mapping

Modern ONA practice, which OrgTool implements directly, maps relationships across five distinct questions: advice, trust, energy, information, and influence. (Advice and information are on every plan; energy, trust and influence are on Team.) The separation is the point. A person can be the advice hub of engineering and invisible in the trust network; an influence broker can sit three layers below the executives whose decisions they shape. Aggregating these into one score erases exactly the signal you mapped them to find.

The math, without mystery

The workhorse metric is betweenness centrality (Freeman, 1977): for every pair of people in the network, find the shortest paths between them, and count how often each person sits on those paths. A high score means the network routes through you — the signature of brokers and gatekeepers. Computing it efficiently uses Brandes' algorithm (2001). Two properties matter for organizational use: it is deterministic (same declared ties in, same scores out, every run) and explainable (you can show precisely why a score is what it is). No training data, no model drift, no black box — which is why it can be defended in front of a works council or a court in a way an opaque ML "collaboration score" cannot.

ONA in restructuring: the two-lens rule

In a reorganization, ONA has one job above all: find the people the chart undervalues before the plan makes them redundant. The classic restructuring failure is cutting a mid-level role that looked duplicative on paper and turned out to be the only bridge between two functions. In AI-era restructuring the stakes compound — a person can score high on AI exposure (their tasks are automatable) and high on centrality (the organization routes through them) at the same time. Exposure says the role changes; centrality says the person is hard to replace. Plans that read only the first lens optimize the org chart and break the organization.

Doing ONA ethically

The method has a fork in the road. Passive ONA harvests email metadata, calendars, and chat logs — powerful, but consent-fraught, regulated in much of Europe, and corrosive to trust if discovered. Active ONA asks people directly and analyzes what they chose to declare. Active data is sparser but consented, explainable, and — run locally — never becomes a vendor-held surveillance dataset. If the analysis is feeding restructuring decisions, the active, local, deterministic route isn't just the ethical choice; it's the defensible one.

Key takeaways

The informal network does the work; the chart only approximates it.
Map five dimensions separately — advice, trust, energy, information, influence.
Betweenness centrality finds brokers deterministically and explainably.
Never cut on exposure alone: cross it with centrality first.
Prefer active, declared-tie ONA over passive metadata mining — especially when the output feeds decisions about people.