Communication Flows

Communication in an organization has direction. Information flows vertically (up and down hierarchies), horizontally (across peer teams), and diagonally (across hierarchies and functions). Each direction has its own failure modes, its own ideal channels, and its own required infrastructure. Most communication pathologies in mid-to-large organizations are directional — the flow is clogged or missing in a specific direction, and the fix is specific to that direction.

This subtopic is less about a single named framework and more about integrating the organizational communication literature into a practical map of the flows that must work for the organization to function.

Origin

The study of organizational communication direction goes back to Chester Barnard's The Functions of the Executive (1938) and Herbert Simon's work on organizations in the 1940s-50s. Henry Mintzberg's The Structure of Organizations (1979) formalized the distinction between vertical and horizontal information flow in managerial work.

The modern engineering-and-tech treatment draws from:

Ed Catmull's writing about communication at Pixar (Creativity, Inc., 2014)
Ray Dalio's radical transparency at Bridgewater (Principles, 2017)
Amy Edmondson's psychological safety research (The Fearless Organization, 2019)
Netflix and GitLab's public culture docs
The Spotify engineering blog's treatment of information flow across squads/tribes/guilds

The core insight: different information needs different flow paths. A single shared channel for everything — like a single Slack channel for a 500-person company — collapses under its own volume.

The Framework

The Three Directions

VERTICAL DOWN: From leadership to teams
  - Strategy, priorities, decisions, context on the bigger picture
  - "What we're doing and why"

VERTICAL UP: From teams to leadership
  - Status, blockers, ground-truth, early warning signals
  - "What's actually happening"

HORIZONTAL: Between peer teams
  - Coordination, dependencies, shared learnings
  - "How do we work together"

DIAGONAL: Across hierarchy and function
  - Mentorship, cross-functional input, informal networks
  - "What I know that someone outside my reporting line needs"

Each direction fails in different ways; each requires different infrastructure.

Vertical Down

The classic problem: executives communicate strategy, but by the time it reaches the IC level, it has been compressed, reinterpreted, or lost entirely.

Why it fails:
  - Each manager in the chain adds their own interpretation
  - "Strategic context" gets stripped to "tasks"
  - By the fifth layer, ICs don't know why they're doing what
    they're doing
  - Leadership thinks they communicated; ICs heard nothing

What good looks like:
  - Leaders communicate directly to broad audiences (AMAs,
    all-hands, written posts)
  - Strategy documents are readable at any level
  - Middle managers clarify and localize but don't replace
    direct leadership communication
  - ICs can trace any task back to a stated priority

Tools:
  - All-hands meetings (but with real substance, not theatre)
  - Written strategy memos (Bezos's memos, Musk's memos,
    Brian Chesky's letters)
  - Quarterly OKRs or equivalent
  - Leadership AMAs
  - Video recordings for asynchronous access

Vertical Up

The more dangerous problem: leadership doesn't hear what's actually happening. People filter bad news before it reaches the top.

Why it fails:
  - Bad news-bearing is punished (even subtly)
  - Each layer softens the message
  - By the time a message reaches leadership, it's been
    filtered through 3-5 "shouldn't worry them about this"
    judgments
  - Leadership gets a rosy picture that doesn't match reality

What good looks like:
  - Skip-level 1:1s (leaders meet with 2+ levels down)
  - Anonymous feedback channels that are actually read and acted on
  - Engineering leadership reads incident post-mortems and
    engineering RFCs directly
  - Culture of psychological safety (Edmondson) — bad news
    doesn't cost the messenger
  - Leaders actively seek out dissent

Tools:
  - Skip-level 1:1s
  - All-hands Q&A (especially with anonymous submission)
  - Upward surveys (engineering satisfaction, management
    feedback)
  - Open office hours
  - Direct channels leaders monitor (e.g., Slack channels
    leaders actively read)

Horizontal

Between peer teams is where most cross-team coordination lives — and where most of it breaks.

Why it fails:
  - Teams don't know what other teams are working on
  - Shared infrastructure changes without coordination
  - Teams duplicate work
  - Integration points become chronic pain
  - "They never told us" becomes recurring retrospective fuel

What good looks like:
  - Public channels for all team-visible work
  - Regular cross-team demos or showcases
  - Shared roadmaps across related teams
  - Explicit interfaces and contracts between teams
  - Guild / community-of-practice structures for shared concerns
  - Documented escalation paths for cross-team conflicts

Tools:
  - Public Slack channels (not DMs)
  - Engineering guilds (backend guild, platform guild, security
    guild)
  - Cross-team demos / showcases
  - Shared roadmap tools visible to adjacent teams
  - ADRs / RFCs visible to the broader organization
  - Regular architectural reviews

Diagonal

Diagonal flow is often under-appreciated. Mentorship from a senior engineer in another team; feedback from a customer-success manager to an engineer; input from legal into a product decision. These are diagonal flows — they cross both hierarchy and function.

Why it fails:
  - No durable structure to enable it
  - People default to same-team, same-level communication
  - Junior engineers don't know how to reach senior engineers
    in other teams
  - Cross-functional input happens ad hoc, usually too late

What good looks like:
  - Mentorship programs that span teams
  - Cross-functional rituals (e.g., legal review built into
    the RFC process)
  - "Coffee chat" norms where anyone can book time with anyone
  - Internal speaker series that showcase cross-functional work
  - Senior engineers available to newer engineers in other teams

Tools:
  - Formal and informal mentorship pairings
  - Guilds / CoPs (cross-team, single-function)
  - Office hours across functions
  - Lunch-and-learn / tech talk series
  - RFC templates that require cross-functional stakeholders
    by default

How to Use It

Diagnosing Flow Health

For each direction, ask:
  1. Is this flow happening? (Not "do we have the tools" but
     "is information actually moving?")
  2. Is it timely? (Weeks-late information is nearly as bad
     as no information.)
  3. Is it accurate? (Is what reaches the destination what
     was sent?)
  4. Is it complete? (Is bad news being filtered out?)

If any answer is "no," that flow needs intervention.

Prescriptive Use

When designing or improving org communication:

1. Start with vertical up.
   If leaders don't hear ground truth, every other flow is
   compensating for a broken feedback loop.

2. Then vertical down.
   With a feedback loop, leaders can check whether their
   communication is landing.

3. Then horizontal.
   With good vertical loops, horizontal coordination has clear
   escalation paths.

4. Diagonal grows organically on top.
   Once the other three work, diagonal flows develop
   naturally given the right lightweight infrastructure.

Scaling Transitions

Communication patterns must change as companies grow. A practice that works at 30 people fails at 300.

~30 people:
  - Informal works: everyone knows everyone
  - A single Slack channel and weekly all-hands suffices

~100 people:
  - Need structured all-hands, functional sub-meetings
  - Need explicit horizontal coordination (Slack channels
    per team, not just #general)
  - Skip-level 1:1s start mattering

~300 people:
  - Written strategy is essential (memos, not just slide decks)
  - Guilds / CoPs needed for cross-team knowledge flow
  - Anonymous feedback channels start mattering
  - Middle managers' communication skills become critical

~1000+ people:
  - Full written-first culture (GitLab model)
  - Formal RFC processes
  - Multiple levels of summary memos (exec summaries, team
    summaries, IC context)
  - Leadership communication professionalized (videos, all-hands
    production, etc.)

Orgs that don't evolve their communication structures stall or fragment at each transition.

Tech & Company Example

Example 1: The Strategy That Didn't Land

Situation: A 400-person company's CEO writes a thoughtful annual
strategy memo. Reads it at the all-hands. Emails it to everyone.

Three months later, a skip-level 1:1 with an IC reveals the IC
has never read it and doesn't know the top three priorities for
the year.

Diagnosis: Vertical-down flow broke.
  - The all-hands was one-time; no reinforcement.
  - Middle managers didn't integrate it into weekly team contexts.
  - The memo was long; the IC never finished it.
  - Nothing in the IC's daily work referenced it.

Intervention:
  - CEO creates a 1-page summary (TL;DR version).
  - Top 3 priorities posted visibly in every team's Slack.
  - Each team's quarterly goals explicitly map to priorities.
  - Monthly Q&A session on the strategy (with questions solicited
    async beforehand).
  - ADRs and RFCs reference the strategic context they support.

Result: Six months later, ICs can name the top 3 priorities and
trace their current work to them.

Example 2: The Bad News That Didn't Travel

Situation: An engineering team notices that a new service they've
been working on has a fundamental scalability issue. It won't
handle the launch-day load.

Over the next 6 weeks, the issue is discussed internally. The
team's manager mentions it in their skip-level but "doesn't want
to worry leadership yet" — they're trying to solve it.

Two days before launch, the CTO learns about it. Launch is
delayed 3 months, burning $M in planned marketing spend.

Diagnosis: Vertical-up flow broke.
  - Culture punished early bad news (the manager had seen colleagues
    get blamed for raising problems early).
  - No structural requirement to surface risks above a threshold.
  - CTO hadn't cultivated channels where ground truth flowed
    freely.

Intervention:
  - Incident-preemption culture: "near-miss" reports that
    leadership reviews, without punishing teams for raising them.
  - Launch-readiness reviews with explicit "risks to launch"
    required section that CTO reads.
  - Engineering leads given explicit mandate: "Bad news is the
    most valuable information. Delay in surfacing it is the
    problem, not the news itself."
  - Quarterly skip-level 1:1s with specific question: "What's
    the biggest thing I don't know?"

Result: Major issues surface 4-8 weeks earlier on average; launches
rarely surprise anyone.

Example 3: The Horizontal Coordination Breakdown

Situation: Two teams (Team A and Team B) ship related features
independently for a year. They occasionally coordinate via ad hoc
DMs.

One Friday, Team A ships a breaking change to their API. Team B's
service starts failing. A production incident takes 4 hours to
resolve. Team A's engineers didn't know Team B depended on that
API.

Diagnosis: Horizontal flow broke.
  - No shared channel for API changes.
  - No dependency registry.
  - No standard for API versioning or deprecation.
  - Teams communicated via private channels when they communicated
    at all.

Intervention:
  - Public channel per service for API changes.
  - Dependency graph published and kept current.
  - Standard deprecation policy (3-month notice for breaking
    changes).
  - RFCs for breaking changes, visible to all downstream teams
    with comment period.

Result: Breaking-change incidents drop from 3-4/quarter to near-zero.

Example 4: Diagonal Success via Guilds

Situation: A 200-person engineering org wanted to improve
cross-team knowledge sharing on testing practices.

Intervention: Created a "Testing Guild" — weekly 30-min
meeting, membership voluntary, purely informational.

Activities:
  - 15 min: Someone presents a testing practice from their team
  - 15 min: Q&A and discussion

Over 6 months:
  - 30-40 engineers regularly attend across 20+ teams
  - Best practices from Team A spread to 15 other teams
  - A published "testing playbook" emerged organically from
    guild discussions
  - New hires joining any team find a community that can
    answer their testing questions

This is diagonal communication at its best: cross-team, cross-level,
cross-function, enabled by a lightweight structure, producing
compound knowledge value.

When It Works

Any organization with more than 50 people
Growing companies (where the next scale break is approaching)
Organizations that have scaled past the point where everyone knows everyone
Multi-office / distributed organizations
Post-merger / post-acquisition integrations
Organizations that have had a specific communication failure and want to avoid repetition

When It Does Not Work

Very small companies — <20 people usually don't need explicit flows; everyone is in the loop by default.
Top-down cultures that punish upward flow — Without psychological safety, no structural intervention restores vertical-up flow.
Extremely political organizations — Communication flows are shaped by power, not by design, and frameworks are ignored.
Very siloed organizations by strategic design — Some orgs (defense, finance) deliberately compartmentalize information; this framework assumes healthy information sharing.

Common Failure Modes

Vertical-Down

Strategy Without Context — "Here's what to do" without "here's why." ICs execute but don't understand.
Manager-as-Filter — Each management layer adds its own interpretation until the message is unrecognizable.
Announcement ≠ Absorption — Believing that sending a message is the same as having it received and understood.
All-Hands Theatre — All-hands meetings that are performance rather than communication. No Q&A, no follow-up, no substance.

Vertical-Up

Shoot-the-Messenger Dynamics — Bearer of bad news is punished. Everyone learns not to bring bad news.
Sanitized Status — Weekly statuses pretend everything is green. Eventually something surprises leadership.
Skip-Level Theatre — Skip-level 1:1s that don't actually surface anything useful. Reports give the politically safe answers.
Leader-Too-Distant — Leaders so far from ground truth that they can't even evaluate whether the information they're getting is accurate.

Horizontal

Team Silos — Each team optimized locally; cross-team concerns fall through cracks.
Coordination via Managers — All cross-team communication routes through manager-to-manager. Slow, lossy, and leaves engineers without peers to talk to.
Tool-Fragmentation — Different teams use different tools, dashboards, and channels. No one can follow what's happening outside their tool bubble.
Duplication Across Teams — Three teams solve the same problem independently because they didn't know about each other's work.

Diagonal

No Mentorship Structure — Junior engineers don't know how to reach senior engineers outside their team.
Function-Only Socialization — Engineers only talk to engineers; no cross-functional diagonal flow.
Informal-Only — All diagonal flow is informal; introverts and new hires excluded from informal networks.

Scale

Same Practices at 30 and 300 — Informal practices that worked at 30 people get frozen in place at 300 and break.
Over-Structuring Too Early — Imposing heavy communication processes on a small team. Overhead exceeds benefit.
Tooling Without Practice — Buying a tool (Jira, Confluence, Notion) and expecting it to produce flow. Tools don't produce culture.

Cultural

Presence ≠ Substance — Measuring communication by activity (messages sent, meetings held) rather than outcomes (decisions made, context shared).
Transparency Theatre — "Radical transparency" slogan but in practice information is still siloed.
Privacy Mishandled — Legitimate private information (HR, compensation, legal) leaked in service of "transparency." Erodes trust.

Conway's Law — Your architecture mirrors your communication. Flow design shapes system design.
Team Topologies — Formal framework for team interaction (collaboration / as-a-service / facilitating); maps to horizontal flow patterns.
Psychological Safety (Edmondson) — The cultural substrate for working vertical-up flow.
Radical Transparency (Dalio / Bridgewater) — A specific (extreme) approach to vertical flow.
OKRs — A mechanism for vertical-down alignment.
All-Hands / Town Halls — Vertical-down ritual.
Skip-Level 1:1s — Vertical-up mechanism.
Guilds / Communities of Practice — Diagonal flow infrastructure.
Async-First Culture (GitLab, Basecamp) — Enables information flow across timezones without bottlenecks.
Amazon 6-pager / Memo Culture — A vertical-up and horizontal flow mechanism at enterprise scale.