Information flow within teams

The invisible infrastructure of team intelligence

Thomas Allen, an MIT researcher who spent decades studying communication patterns in R&D organizations, discovered a finding so consistent it became known as the "Allen Curve": the probability of two engineers communicating regularly drops off dramatically with physical distance. Engineers sitting six feet apart communicate four times more frequently than engineers sitting sixty feet apart. Beyond about one hundred feet, the communication frequency is essentially zero — indistinguishable from engineers in different buildings or different countries (Allen, 1977).

Allen's finding was about physical proximity, but its underlying principle is about information flow architecture. Communication does not happen automatically between people who need each other's knowledge. It happens along paths that are either designed (through channels, protocols, and routing mechanisms) or that emerge accidentally (through proximity, friendship, and habit). When the paths are well-designed, the right information reaches the right people and the team's collective intelligence is activated. When the paths are poorly designed or undesigned, critical information sits in one person's mind while another person struggles without it — and neither person knows about the gap.

Melvin Conway formalized a related insight in 1968: "Any organization that designs a system will produce a design whose structure is a copy of the organization's communication structure." Conway's Law, as it came to be known, means that the architecture of the team's information flow directly shapes the architecture of its products. A team with poor cross-functional information flow will build a product with poor cross-functional integration — not because the engineers are unskilled but because the information needed for integration never reaches the people doing the work (Conway, 1968).

The anatomy of information flow

Team information flow has four structural properties that determine its effectiveness.

Reach is the extent to which information can travel from any point in the team to any other point. In a team with high reach, the database engineer's discovery about a performance anomaly can reach the frontend engineer who is unknowingly exacerbating it. In a team with low reach, information is trapped in subgroups — the backend team knows about the API instability, but the client team does not, and they discover it only when their integration fails. Reach is determined by the connectivity of the team's communication channels and the norms about who shares information with whom.

Velocity is the speed at which information moves from source to destination. In an incident, velocity is critical — the difference between a five-minute detection-to-response time and a five-hour one is often the difference between a minor disruption and a major outage. Velocity depends on the latency at each handoff point in the information's path: automated monitoring triggers instant alerts (high velocity); information shared in a weekly report reaches its audience days after the event (low velocity).

Fidelity is the accuracy with which information arrives at its destination. The original signal — "the checkout service is experiencing intermittent 500 errors under load" — may degrade through each handoff into "there might be a checkout issue" and eventually "someone mentioned a bug." Each translation step strips context, nuance, and specificity. High-fidelity information flow preserves the original signal's precision. Low-fidelity flow produces a game of telephone where the message that arrives is barely recognizable as the message that was sent.

Signal-to-noise ratio is the proportion of information that is relevant to the recipient versus total information received. A Slack channel where every message matters has a high signal-to-noise ratio. A Slack channel where important announcements compete with casual conversations, automated notifications, and off-topic threads has a low ratio — and the cognitive cost of filtering signal from noise falls on every reader, consuming attention that should be directed at the work itself.

Designing information routing

Rob Cross and Andrew Parker's research on organizational network analysis demonstrated that most teams have informal information networks that differ dramatically from the formal organization chart. The actual flow of critical information often depends on a few "information brokers" — individuals who bridge between subgroups and route knowledge that would otherwise be siloed. When these brokers leave, are reassigned, or become bottlenecked, information flow degrades across the entire team (Cross & Parker, 2004).

The dependency on individual brokers is a design vulnerability. Effective information routing should be structural — built into the team's processes and tools — rather than dependent on specific people. Several design patterns address this:

Broadcast with filtering. Important information is shared to a broad channel, but with metadata that allows recipients to filter efficiently. A production alert tagged with "service: checkout, severity: high" reaches everyone who has opted into checkout or high-severity alerts, and no one else. The tagging system replaces human judgment about "who needs to know" with a structural mechanism that routes based on declared interests and responsibilities.

Push for urgency, pull for reference. Urgent information (incidents, blocking issues, customer escalations) is pushed to the relevant people through notifications, pages, or direct messages. Reference information (documentation, decisions, status updates) is made available in well-organized locations for people to pull when they need it. The failure of most teams is treating all information as push — creating a flood of notifications that teaches people to ignore notifications, including the urgent ones.

Redundant paths. Critical information should have more than one path from source to destination. If the stand-up is the only channel for surfacing blockers, then anyone who misses the stand-up misses the blockers. A redundant path — a written stand-up summary posted in a dedicated channel, a bot that collects and distributes blocking items — ensures that the information reaches its destination even if one path fails.

Handoff reduction. Every handoff point in an information path introduces latency and fidelity loss. Reducing the number of handoffs between information source and destination improves both speed and accuracy. The ideal is zero handoffs: the person who discovers the information is the person who can act on it. When this is not possible, the minimum is one handoff: the discoverer routes directly to the actor, without intermediaries.

Morten Hansen's research on information sharing in organizations found that the biggest barrier to effective flow is not the absence of information but the costs of transferring it. Hansen identified "search costs" (finding who has the information), "transfer costs" (getting the information from them in usable form), and "integration costs" (incorporating the information into the recipient's work). Reducing these costs — through knowledge maps (The team's knowledge graph), structured formats, and clear routing — improves information flow more than simply encouraging people to "share more" (Hansen, 1999).

The Third Brain

Your AI system can serve as an information flow optimizer. Describe your team's communication channels, tools, and norms to the AI and ask: "Where are the most likely information bottlenecks in our setup? What critical paths have single points of failure? Where is information likely to be lost between source and destination?" The AI can analyze the structural properties of your information architecture and identify vulnerabilities that are invisible from inside the system.

The AI can also serve as an information router. When team members share updates, discoveries, or decisions, the AI can analyze the content and suggest who else on the team should be informed: "This API change affects the mobile team's upcoming release. Consider notifying Chen and Priya." The routing suggestion replaces the unreliable human judgment of "who needs to know?" with a structural mechanism that cross-references information content against team members' responsibilities and current projects.

For ongoing monitoring, the AI can analyze the team's communication patterns over time: "Which team members are communicating frequently and which are isolated? Which types of information are flowing well and which are getting stuck? Are there pairs of people who should be communicating based on their work but are not?" These patterns reveal the actual information flow network — which may differ significantly from the designed one — and highlight gaps where information is being lost.

From flow to focus

Information flow ensures that knowledge reaches the people who need it. But reaching someone is not the same as getting their attention. In a world of abundant information and scarce attention, the critical question is not just "Does the information arrive?" but "Does it get attended to?"

The next lesson, Team attention management, examines team attention management — the collective process of directing the team's finite cognitive resources toward the problems and opportunities that matter most.

---

Sources:

1. Allen, T. J. (1977). Managing the Flow of Technology: Technology Transfer and the Dissemination of Technological Information Within the R&D Organization. MIT Press.
2. Conway, M. E. (1968). "How Do Committees Invent?" Datamation, 14(4), 28-31.
3. Cross, R., & Parker, A. (2004). The Hidden Power of Social Networks: Understanding How Work Really Gets Done in Organizations. Harvard Business School Press.
4. Hansen, M. T. (1999). "The Search-Transfer Problem: The Role of Weak Ties in Sharing Knowledge Across Organization Subunits." Administrative Science Quarterly, 44(1), 82-111.