org-design-principles

# Org Design Principles ## Core Principles 1. **Purpose, then structure**: design the org to enable the work 2. **Conway's Law is real**: whatever you ship reflects the org chart 3. **Cognitive load is a constraint**: teams have finite capacity 4. **Subtractive design beats additive**: remove a layer before adding one 5. **Decision rights matter more than reporting lines**: who decides what is the structure 6. **Stage matters**: a 30-person org doesn't need product management; a 500-person org without it is suffering ## Mintzberg's Configurations (Brief) Five organizational configurations (Mintzberg, *Structure in Fives*, 1983): - **Simple Structure**: founder + a few employees; direct supervision - **Machine Bureaucracy**: standardized processes; large operational orgs - **Professional Bureaucracy**: standardized skills; expertise-driven (hospitals, law firms) - **Divisional**: independent business units; HQ light coordination - **Adhocracy**: project-based; flexible; innovation-driven Most tech companies are some hybrid. The config evolves with stage. ## Functional vs. Divisional vs. Matrix | Structure | When It Works | Risk | |-----------|---------------|------| | **Functional** (eng, product, design) | Small to mid-scale; coordination via shared rituals | Hard to coordinate cross-function for customer outcome | | **Divisional** (GMs per business unit) | Multiple distinct business lines | Duplication; weak functional craft | | **Matrix** (dual reporting) | Truly cross-cutting work | Coordination cost; needs explicit operating norms | Default for most growth-stage tech: functional with strong cross-functional rituals. ## Team Topologies (Skelton & Pais, 2019) Four team types: - **Stream-aligned**: deliver value end-to-end on a slice of business — the default team type - **Platform**: provide internal services to make stream-aligned teams faster - **Enabling**: temporary; help stream-aligned teams adopt new capabilities; dissolve - **Complicated-subsystem**: deep expertise required (ML, payments, compliance) Four interaction modes: - **Collaboration**: high coupling; short-lived - **X-as-a-service**: low coupling; long-lived - **Facilitating**: enabling team to stream team - **Compliance / governance**: when audit/regulation requires gating ## Spans of Control | Manager type | Healthy span | |--------------|--------------| | First-line manager (manages ICs) | 6–12 | | Manager of managers | 5–9 | | Functional VPs | 5–8 | | CEO directs | 7–10 (12+ is a sign) | | Founder directs | Audit if > 9; founders often over-collect | Wider spans require more autonomous teams; narrower spans mean more management overhead per IC. ## Layers From CEO to most junior employee: | Headcount | Healthy layers | |-----------|----------------| | <100 | 3–5 | | 100–500 | 4–6 | | 500–2000 | 5–7 | | 2000+ | 7–9 | More layers = slower decisions, more political work, more "translation." When the founder adds an extra layer to manage trust issues, the layer rarely comes out. ## Decision Rights Name the decisions, then assign authority: - Hiring approval - Comp approval - Headcount allocation - Product decisions - Customer commitments - Vendor selection - Brand and messaging Pick a framework (RACI, RAPID, DACI) and use it consistently. Bias toward decentralization: senior leaders typically under-delegate. The test: "if this decision was wrong, would the consequences be reversible?" If yes, delegate. ## Conway's Law Conway, 1968: "Organizations design systems that mirror their communication structures." Practical implication: the products you ship reflect how the org is organized. Want a customer-coherent product? Organize around the customer journey, not around technical components. Inverse Conway: design the org for the systems / customer experience you want. ## Common Failure Modes - **Premature specialization**: Chief Strategy Officer at 80 people; "Center of Excellence" at 200 - **Hidden bottlenecks**: one person on every decision; team won't admit it - **Founder span explosion**: 14 directs because no one can be promoted - **Unclear ownership at seams**: marketing-product handoff, sales-CS handoff - **Matrix without operating norms**: dual reporting with no priority resolution - **Reorg as solution to performance problem**: shuffling boxes to avoid firing - **Importing structure**: copying Stripe at Series E into your seed-stage company ## Signs of Org Distress - Decisions take more than 2 weeks - Same conflict between two teams every quarter - "We need to align" appears in 5+ Slack messages a day - Senior leaders attending meetings to "represent their function" and adding nothing - Roles created to retain people, not to do work - "Shadow leaders" with informal control ## Reorgs Reorgs are expensive. Only do them when the cost of not is higher. When you do: - Strategy first; structure second - Communicate generously and early - Be specific about who's affected and how - Address comp and reporting changes clearly - Plan the transition — not just the announcement ## Cross-References - `org-designer` agent - `spans-and-layers` skill - `workforce-planning` skill - `change-management` skill ## Key References - Mintzberg, H. (1983). *Structure in Fives*. - Skelton, M., & Pais, M. (2019). *Team Topologies*. - Stanford, N. (2018). *Guide to Organisation Design*. - Conway, M. E. (1968). "How do committees invent?" - Lencioni, P. (2012). *The Advantage*.