The network you can’t see is the one running your company

Most executives can sketch their org chart from memory. Almost none of them can accurately describe how work actually moves through their organization.

Rob Cross has spent two decades mapping these invisible networks across more than 300 organizations using Organizational Network Analysis. His consistent finding is that leaders are only about 50% accurate at identifying their central connectors and roughly 30% accurate at identifying boundary-spanning brokers who link disconnected groups. That gap between what leaders think they see and what’s actually there has real financial consequences. In one engineering company, annual network mapping over five years drove a reorganization that cut IT costs from 5.2% to 3.6% of gross revenue. The org chart couldn’t have told them where the bottlenecks were. The network could.

We wrote two weeks ago about how innovation depends on collaborative relationships, not lone geniuses. The patent data from Singh and Fleming, the team disruption research from Wu and colleagues, and Uzzi’s work on diverse teams producing more novel outcomes all point in the same direction. But confirming that relationships drive innovation doesn’t tell a mid-market executive what to do on Monday morning. The question that matters is practical: how do we build and maintain the specific kinds of internal connections that the research says drive performance?

The answer starts with understanding what’s actually valuable in a network. Ronald Burt’s research on structural holes, published in the American Journal of Sociology (2004), studied 673 managers in a large electronics company and found that people whose networks span gaps between groups generate better ideas, receive higher performance evaluations, and get promoted faster. The finding that should keep executives up at night: only 3% of connections in that organization crossed business unit boundaries. The structural holes between divisions were enormous, and the few people who bridged them captured disproportionate value.

Cross’s work identifies who fills these bridging roles. He distinguishes central connectors (well-connected within a group), brokers (linking disconnected clusters), and energizers (people who generate enthusiasm and action in others). Being an energizer, Cross found, is four times more predictive of high performance than any other network factor he measured. Most organizations have no idea who these people are, and their reward systems certainly don’t recognize them.

Granovetter’s 1973 research on weak ties, casual acquaintances rather than close colleagues, established why bridging connections matter so much. Close contacts tend to circulate the same information. Weak ties across group boundaries deliver non-redundant knowledge, which is exactly what recombinant innovation requires. Uzzi and Spiro confirmed this by studying Broadway musicals over four decades: creative networks perform best at a specific balance point where distinct clusters exist with different ideas, and enough connections for those ideas to travel between clusters. Too insular and you stagnate. Too interconnected, and everyone thinks alike.

The problem is that these bridging connections are extraordinarily fragile. Burt’s longitudinal tracking found that 9 out of 10 bridge ties disappeared within a year without active maintenance. The Microsoft study by Yang and colleagues (2022), covering 61,000 employees during the shift to remote work, showed that collaboration networks became more static and siloed. Workers formed fewer new ties and spent more time with existing strong ties at the expense of weaker cross-boundary connections. Networks froze in place.

So we have a situation where the most valuable connections for innovation are also the most fragile, and most organizational design decisions are quietly destroying them. Expanding the manager’s span to 12 direct reports eliminates discretionary time for relationship-building. Incentive structures that reward only within-team performance punish people for investing in cross-boundary connections. Efficiency mandates compress the very kind of unstructured time in which weak ties form.

Knowing the problem, though, is the easy part. Making it stick requires specific organizational design choices.

See the network before redesigning it. Cross’s ONA methodology, which surveys employees about who they turn to for information, decision-making, and energy, reveals patterns invisible to leadership. This isn’t a one-time exercise. Annual mapping tracks whether interventions are actually changing connection patterns or whether the network is reverting to its default shape. The investment is modest relative to the decisions it informs.

Design spaces for functional proximity, not just co-location. Thomas Allen’s research at MIT established that communication frequency drops exponentially with physical distance. Engineers are four times more likely to communicate regularly with someone two meters away than with someone twenty meters away. But the nuance matters more than the headline. Kabo and colleagues (2014) found that shared walking paths predict collaboration formation independently of raw distance. Placing shared resources — coffee, printing, mail — between groups rather than within them creates encounter opportunities that straight-line proximity cannot. One caution from Bernstein and Turban’s 2018 research: transitioning to open-plan offices actually reduced face-to-face interaction by roughly 70% as workers retreated into headphones. Removing walls doesn’t create connections. Creating reasons to cross paths does.

Structure time for cross-boundary work. Ben Waber’s sociometric research found that synchronized breaks, allowing teams to take breaks together rather than on staggered individual schedules, produced a 15-20% increase in productivity and a 19% drop in stress levels in a Bank of America call center. That’s a zero-cost intervention. Morten Hansen’s research across 5,000 individuals in 15 industries argues for what he calls disciplined collaboration: not more collaboration, but collaboration structured around goals that genuinely require cross-unit work. His findings at HP across 140 projects were that organizations pursuing disciplined collaboration outperformed those pursuing either isolation or indiscriminate collaboration. Cross-functional problem-solving sessions work when the problem is framed as shared, and the solution requires co-production, which Kerrissey, Mayo, and Edmondson (2021) call a joint problem-solving orientation.

Hire and promote for brokerage, not just expertise. Hansen’s T-shaped management framework, developed from BP case studies, makes a straightforward argument: organizations need people who perform deeply in their domain and share knowledge freely across boundaries. Evaluate for both. Promote for both. The horizontal bar of the T doesn’t develop accidentally — it develops when organizations make boundary-spanning an explicit expectation rather than an extracurricular activity.

Treat bridge maintenance as infrastructure, not goodwill. Given Burt’s finding that 90% of bridge ties decay within a year, cross-boundary relationships require continuous, deliberate investment. Rotating cross-functional task force membership, periodic network audits, and making connection-building part of performance expectations rather than optional. These are structural decisions, not personality traits. Yang and colleagues’ recommendation from the Microsoft study was direct: make cross-boundary connecting part of employees’ jobs, not an additional task for people with extra time.

None of this is complicated in concept. All of it requires sustained attention in execution, which is where most organizations fall short. We default to optimizing what we can see: the org chart, the budget, the headcount. Meanwhile, the invisible network that actually determines how work flows, how ideas combine, and how people decide whether to stay or leave operates entirely outside our field of vision.

Mervin Kelly designed Bell Labs’ Murray Hill campus to foster cross-disciplinary encounters because he intuitively understood what Burt, Cross, and Granovetter have since quantified. The relationship infrastructure within an organization isn’t a nice-to-have that leaders address after operational priorities are met. It is the operational priority. Every other capability the organization needs: innovation, execution, retention, adaptation, all of it; flows through it.

This work synthesizes research from organizational network analysis (Cross, Borgatti), structural holes theory (Burt), weak ties and collaborative creativity (Granovetter, Uzzi), proximity and workspace design (Allen, Kabo, Bernstein), disciplined collaboration and boundary spanning (Hansen, Ancona, Edmondson), and large-scale remote work studies (Yang et al./Microsoft), spanning publications in Management Science, American Journal of Sociology, Nature Human Behaviour, Administrative Science Quarterly, Research Policy, and Academy of Management Discoveries.

Contact us for a complete list of works cited.