Systems Theory
Cynefin Framework: How to Identify the System's Domain
Managers often apply the same tools to different problems. Analysis everywhere. Processes everywhere. Experiments everywhere. But **fixing a car** and **raising a child** are fundamentally different tasks requiring DIFFERENT approaches. **Cynefin Framework** teaches the most important leadership skill: **understand where the system is before deciding what to do**.
- **9/11 (Chaotic):** Rick Rescorla didn't analyze - he commanded the evacuation. Saved 2,687 people
- **Facebook 2004 (Complex):** Zuckerberg didn't know the best practice - he experimented college by college
- **Nokia 2007 (Clear→Chaotic):** 'We've always done it this way' → ignored the iPhone → collapse
- **COVID-19:** January 2020 (Chaotic) → Spring (Complex) → 2021 (Complicated) → 2023 (Clear)
Cynefin: Five Decision Domains
**2001, 9/11.** Leaders of companies in the Twin Towers made decisions in seconds - evacuation, calling families, helping colleagues. Nobody called a meeting. Nobody waited for analysis. Why?
**2008, financial crisis.** Lehman Brothers went bankrupt in 48 hours, even though problems had been known for months. Experts analyzed, consultants built models, CEOs held meetings. Why didn't it help?
The answer: **wrong domain = wrong strategy = catastrophe**.
**Cynefin Framework** (pronounced 'kuh-NEV-in,' a Welsh word meaning 'habitat') - a framework for understanding context and choosing the right decision-making strategy. Created by Dave Snowden at IBM in 1999.
Cynefin divides all situations into **5 domains**, each requiring a DIFFERENT strategy:
**The critical mistake:** Managers apply their familiar tools to every problem. Best practices for Complicated. Analysis for Chaotic. Experiments for Clear. Cynefin teaches: **first identify the domain, then choose the tools**.
**The key idea of Cynefin:** Not all problems are the same. Fixing a car and raising a child are fundamentally different tasks requiring different approaches.
There is a universal decision-making strategy (e.g., 'always analyze before acting')
The strategy depends on the domain. In Clear - follow the procedure; in Complicated - analyze; in Complex - experiment; in Chaotic - act immediately
Context determines the right approach. A 'good leader' in a crisis (commands) differs from a 'good leader' in innovation (creates conditions for experiments).
A company always solves problems through analysis and best practices. But when a crisis hit (a hacking attack), nothing works. What's the problem?
Ordered Domains: Clear and Complicated
Clear and Complicated are **ordered domains**. In them, cause-and-effect relationships EXIST and are KNOWABLE (though in Complicated they require analysis). A correct answer exists.
**Clear (Simple) domain:** Cause-and-effect relationships are obvious to everyone. The right answer is known. 'Best practices' work.
**Clear: Sense → Categorize → Respond**
- **Sense** - assess the situation
- **Categorize** - assign it to a known category
- **Respond** - apply the standard solution
**The Clear trap:** Complacency. 'We've always done it this way and it worked.' Ignoring changes - sudden fall into Chaotic. This is the most dangerous transition in Cynefin.
**Complicated domain:** Cause-and-effect relationships exist but require analysis. Multiple right answers. Experts needed. 'Good practices' (not best!).
**Complicated: Sense → Analyze → Respond**
- **Sense** - gather data
- **Analyze** - bring in experts for analysis
- **Respond** - choose from among several good solutions
| Clear | Complicated | |
|---|---|---|
| Relationships | Obvious | Require analysis |
| Solution | One best | Several good ones |
| Strategy | Sense-Categorize-Respond | Sense-Analyze-Respond |
| Practices | Best practices | Good practices |
| Who decides | Anyone / the system | Experts |
| Time | Fast | Takes time |
| Predictable? | Yes | Yes, after analysis |
**The Complicated trap:** Analysis paralysis. 'We need more data, more analysis.' Experts argue, consultants build models. While they analyze - the world changes, the problem shifts to a different domain.
Clear and Complicated are about problem size (small vs. large)
Clear and Complicated are about the type of relationships (obvious vs. requiring analysis)
Refueling a supertanker (a big task) - Clear. Diagnosing a rare illness in an infant (a small task) - Complicated. It's not about size, but about the nature of cause-and-effect relationships.
Choosing a programming language for a new project. There are several good options (Python, Go, Rust), each with trade-offs. The requirements need analysis. This is...
Complex: Emergent Practices
**2004, Facebook.** Zuckerberg didn't know if it would work. Consultants couldn't predict. Experts were skeptical. But through experiments (college by college, feature by feature) an **emergent practice** appeared - a new type of social network.
**Complex domain:** Cause-and-effect relationships are only visible **in retrospect** (after the fact). There is no correct answer in advance. Emergent patterns. 'Emergent practices.'
**The key sign of Complex:** The correct answer **DOES NOT EXIST** in advance. It can only be discovered **AFTERWARD**, through experiments.
**Complex: Probe → Sense → Respond**
- **Probe** - run safe-to-fail experiments (in parallel!)
- **Sense** - observe what works
- **Respond** - amplify what works, dampen what doesn't
**Critical difference from Complicated:** In Complicated, the correct answer EXISTS and CAN BE FOUND through analysis. In Complex, the correct answer DOESN'T EXIST in advance - it EMERGES through interactions.
**Complex examples:**
| Area | Why Complex | How to work |
|---|---|---|
| Raising a child | Every child is unique; no best practice | Try different approaches, watch what works |
| Startup (product-market fit) | Unknown what the market will respond to | MVP, experiments, pivot based on results |
| Agile development | Requirements change, tech landscape shifts | Sprints = safe-to-fail experiments |
| Organizational change | People react unpredictably | Pilots in different teams, observe patterns |
| Markets, ecosystems | Emergent behavior of many agents | Observe signals, adapt |
**Agile as a strategy for Complex:**
**The Complex trap:** Applying a Complicated strategy (analysis, planning). 'Let's analyze all the risks and write a plan.' No. In Complex the plan will be obsolete. Experiments are needed.
In complex situations, more analysis and better planning are needed
In Complex situations, analysis won't give an answer. Safe-to-fail experiments are needed.
Complicated ≠ Complex. In Complicated it's possible to analyze and find the answer. In Complex the answer emerges through interactions - it can't be calculated in advance.
A company is rolling out a new culture. Consultants propose: '6 months of analysis → detailed plan → implement by the plan.' What's wrong?
Chaotic: Act First, Think Later
**September 11, 2001, 8:46 AM.** The first plane hits the tower. Leaders at Morgan Stanley (offices on floors 44-74) have seconds to decide. Rick Rescorla, head of security, doesn't analyze, doesn't call a meeting. He COMMANDS: 'Evacuation. NOW.' Result: 2,687 of 2,700 people are saved.
**Chaotic domain:** No visible cause-and-effect relationships. No time for analysis. Action is needed IMMEDIATELY. 'Novel practices' - new approaches emerge right now.
**Chaotic: Act → Sense → Respond**
- **Act** - act IMMEDIATELY (stabilize the situation)
- **Sense** - observe the effect
- **Respond** - move the system into Complex or Complicated
**Critically important:** Chaotic is NOT a normal state. It is a crisis. The goal is to move the system into Complex or Complicated as quickly as possible, where there is time to think.
**Chaotic examples:**
| Situation | Why Chaotic | Right action |
|---|---|---|
| Building fire | No time, chaos | Evacuate immediately - analyze after |
| DDoS attack on a service | Service going down right now | Enable protection - understand source - plan |
| Market panic (2008) | Cascading crash | Halt trading - inject liquidity |
| First days of COVID | Unknown virus, panic | Quarantines - research - protocols |
| Industrial accident | Threat to life | Stop the line - assist - investigate |
**Leadership style in Chaotic: Command & Control**
In a crisis, democracy doesn't work. A leader is needed who COMMANDS. Not discussing, not consulting - ORDERING. This is the only correct strategy for Chaotic.
**The Chaotic trap:** Staying in crisis mode too long. Command & Control exhausts the team. As soon as the situation is stabilized - move to Probe-Sense-Respond (Complex) or analysis (Complicated).
A good leader always involves the team in decisions
In Chaotic, a good leader COMMANDS. In Complex - creates conditions for experiments. Style depends on the domain.
Democracy during a fire kills people. Directives during innovation kill creativity. Context matters.
Company servers are under attack, customers can't log in, losing $10K/minute. CTO says: 'Let's hold a meeting and analyze all the options.' What's wrong?
Transitions Between Domains and Disorder
Systems are not static - they **move** between domains. Understanding the dynamics of transitions is critically important.
**Disorder - the fifth domain at the center of Cynefin:** A state in which it is unknown which domain the system is in. The most dangerous zone - people apply whatever methods they're used to.
**Complacency Cliff:** The most dangerous transition - Clear → Chaotic. When procedures have worked for a long time, the illusion arises that 'it will always be this way.' Weak signals of change are ignored. And one day - a fall into chaos.
**Complacency Cliff examples:**
- **Nokia (2007):** 'We're market leaders, our processes work' → iPhone → collapse in 4 years
- **Kodak:** 'Film has always sold' → digital cameras → bankruptcy
- **Lehman Brothers:** 'Our risk model has been tested' → subprime mortgages → collapse in 48 hours
- **Blockbuster:** 'People will always rent discs' → Netflix → closure
**Disorder: when it is unknown where the system is**
**Example: Breaking into domains**
A startup launch is not a single domain!
| Aspect | Domain | Strategy |
|---|---|---|
| Product-market fit | Complex | Probe-Sense-Respond (MVP, experiments) |
| Technical architecture | Complicated | Sense-Analyze-Respond (experts) |
| Accounting, compliance | Clear | Sense-Categorize-Respond (procedures) |
| Launch day | Chaotic | Act-Sense-Respond (bugs, load spikes) |
**Typical Disorder mistake:** Applying ONE strategy to the entire project. 'We're Agile - sprints everywhere!' Or 'We're enterprise - processes everywhere!' No. Different parts of a system = different domains.
**How to identify the domain:**
If an approach worked before, just apply it better
If an approach worked before but the situation has changed - the system may now be in a different domain
Best practices (Clear) become an anchor when the world shifts into Complex. 'Doing the same thing better' won't help if something different is needed.
A company has been selling through retail for 15 years (Clear). Sales are declining. CEO: 'Let's optimize our processes!' But the problem isn't processes - the market has moved online. What's happening?
Cynefin: Key Ideas
- **5 domains:** Clear, Complicated, Complex, Chaotic, Disorder - each requires a DIFFERENT strategy
- **Clear:** Obvious relationships → Sense-Categorize-Respond → Best practices → Delegate
- **Complicated:** Analysis needed → Sense-Analyze-Respond → Good practices → Experts
- **Complex:** Emergence → Probe-Sense-Respond → Emergent practices → Experiments
- **Chaotic:** Crisis → Act-Sense-Respond → Novel practices → Command and stabilize
- **Disorder:** Unknown which domain → Break into parts → Assess each separately
- **Complacency Cliff:** The most dangerous transition Clear→Chaotic ('we've always done it this way' → the fall)
- **The main rule:** Wrong domain = wrong strategy = catastrophe
Where Next?
Cynefin is a meta-framework for understanding systems. Connections:
- Emergence — Why in Complex there is no correct answer in advance
- Feedback Loops — Mechanisms that create dynamics in all domains
- Resilience — How systems transition between domains
- Leverage Points — Where to intervene in each domain
Вопросы для размышления
- Recall a project that failed. Which domain was it perceived as being in? Which domain was it really in? What strategy was used?
- Find parts of current work that belong to different domains. Are different strategies being applied to each?
- Has there been a 'Complacency Cliff' - when 'what always worked' suddenly stopped? What were the early warning signals?
- How are decisions made in a crisis? Does the tendency arise to analyze (Complicated strategy) instead of acting (Chaotic)?
- In which situations are best practices applied when experiments are called for (Clear instead of Complex)?
Связанные уроки
- st-13-mental-models — Switching between models in different Cynefin domains
- st-23-wicked — Wicked problems live in the Complex domain of Cynefin
- st-04-leverage — Leverage points depend on the system's Cynefin domain
- st-17-chaos — Chaotic domain of Cynefin and chaos dynamics
- ct-02-biases — Cognitive biases when misidentifying the domain
- ct-04