szoloth/systems-thinking-coach

Systems Thinking Coach

Overview

Analyze complex problems through a systems lens using Donella Meadows' frameworks. Identify feedback loops, delays, leverage points, and systemic structures that drive behavior. Design interventions that address root causes rather than symptoms.

When to Use Systems Thinking

Strong signals:

• Same problem keeps recurring despite multiple "fixes"
• Solutions create new problems (unintended consequences)
• Different stakeholders see completely different problems
• Everyone agrees on symptoms but disagrees on causes
• Simple interventions produce unexpected resistance
• Problem involves multiple reinforcing factors

Examples:

• "We keep missing deadlines despite adding more people"
• "Every process improvement creates new bottlenecks"
• "Users keep requesting features we've already built"

Core Systems Concepts

1. Feedback loops

Reinforcing loops (amplification):

• Success breeds success, failure breeds failure
• Virtuous cycles and vicious cycles
• Exponential growth or collapse

Example: Good tool adoption → More users → More feature requests → Better tool → More adoption

Balancing loops (stabilization):

• System seeks equilibrium
• Goal-seeking behavior
• Resistance to change

Example: Add developers → More code → More bugs → More time fixing → Less new development

2. Delays

Time gaps between action and consequence cause oscillation and overshoot.

Example: Hire more people (action) → 3 months onboarding (delay) → Productivity dip before improvement → Management panics and hires even more → Chaos.

3. System structure determines behavior

People respond rationally to system incentives. "Bad" outcomes usually stem from system design, not bad actors.

Example: Artists duplicate work not because they're careless, but because the search system makes discovery harder than recreation.

Donella Meadows' Leverage Points

Ranked from least to most effective (counterintuitively, high-leverage points are often subtle):

12. Parameters (Weakest)

• Tweaking numbers (budgets, timelines, targets)
• Common but rarely transformative

11. Buffer sizes

• Stabilizers that prevent overshooting
• Example: Caching to smooth load spikes

10. System structure

• Physical infrastructure and flows
• Example: Network bandwidth, storage capacity

9. Delays

• Reduce time between action and feedback
• Example: Ship smaller increments for faster learning

8. Balancing loops

• Negative feedback that prevents runaway
• Example: Automated tests catch bugs before shipping

7. Reinforcing loops

• Positive feedback that amplifies change
• Example: Network effects, word-of-mouth growth

6. Information flows

• Make consequences visible to decision-makers
• Example: Show developers production errors in real-time

5. Rules

• Incentives, constraints, permissions
• Example: Change promotion criteria to value quality over features

4. Self-organization

• System's ability to evolve new structures
• Example: Let teams form around problems vs assigned

3. Goals

• System purpose drives all behavior
• Example: Shift from "ship features" to "solve user problems"

2. Paradigm

• Mental models and assumptions
• Example: From "users are the problem" to "system design is the problem"

1. Transcending paradigms (Strongest)

• Ability to shift between multiple paradigms
• Recognizing all models are incomplete

See references/leverage-points-detailed.md for deep dive on each level.

Analysis Workflow

1. Identify the problem behavior

What pattern keeps recurring? Describe observable symptoms without diagnosis.

Good: "Sprint velocity drops every time we onboard new developers" Avoid: "We need better onboarding" (already jumped to solution)

2. Map the system

Identify key elements:

• Variables: What changes over time? (team size, bug count, velocity)
• Connections: What affects what? (more code → more bugs)
• Feedback loops: What reinforces or balances?
• Delays: Where are time gaps between cause and effect?

Use causal loop diagrams (see assets/causal-loop-template.md).

3. Find the structure

What systemic patterns produce this behavior?

Common archetypes:

• Fixes that fail: Solution has unintended consequences
• Shifting the burden: Symptomatic fix prevents addressing root cause
• Tragedy of the commons: Individual optimization degrades shared resource
• Success to the successful: Winner gets advantages that ensure continued winning

Reference references/system-archetypes.md for pattern library.

4. Identify leverage points

Where can small changes produce disproportionate impact?

Questions:

• What information is invisible to decision-makers?
• What delays prevent fast feedback?
• What rules create perverse incentives?
• What goals are driving unwanted behavior?

5. Design intervention

High-leverage approaches:

• Make consequences visible (information flows)
• Reduce delays (faster feedback)
• Change incentives (rules)
• Shift goals or paradigms

Test for:

• Does this address structure or just symptoms?
• Will this create new problems elsewhere?
• Can system self-correct once started?

Example Analysis

Problem: Artists spend too much time searching for assets

Symptom-level solution: Build better search UI

Systems analysis:

Variables: Time spent searching, asset count, search quality, duplication

Reinforcing loop (vicious cycle):

Poor search → Can't find assets → Recreate from scratch → More assets → Harder to search → Worse
 search

Balancing loop (resistance):

• Better search tools require time to build
• Building takes time away from creating assets
• Leadership sees "low productivity" and demands less tool-building

Leverage points:

1. Information flow (high leverage): Show leadership total hours lost to search vs hours to build tool
2. Rules (high leverage): Make asset reuse a team metric, not just individual output
3. Parameters (low leverage): Add more search filters (doesn't address duplication cycle)

Recommended intervention:

• Measure and visualize cost of poor search (information flow)
• Change incentives to reward asset reuse over creation (rules)
• This breaks reinforcing loop without requiring immediate tool perfection

Resources

references/

• leverage-points-detailed.md - Complete guide to Meadows' 12 leverage points with examples
• system-archetypes.md - Common patterns and how to recognize them
• causal-loop-guide.md - How to draw and interpret feedback loop diagrams

assets/

• causal-loop-template.md - Structure for creating system maps
• analysis-worksheet.md - Step-by-step problem analysis template

Quality Checks

Before finalizing systems analysis:

• [ ] Identified at least one reinforcing and one balancing loop
• [ ] Mapped key delays between cause and effect
• [ ] Intervention addresses structure not just symptoms
• [ ] Considered unintended consequences in other parts of system
• [ ] Identified metrics to track whether intervention works
• [ ] Solutions target leverage points above "parameters" (more than tweaking numbers)