Local vs Global Optimization
The principle that optimizing individual parts of a system often degrades overall system performance because local efficiency can conflict with global effectiveness.
Also known as: Local Optimization Trap, Sub-optimization, Global vs Local Optima
Category: Principles
Tags: systems-thinking, management, mental-models, strategies, productivity
Explanation
Local vs Global Optimization is a fundamental systems thinking principle: making each part of a system individually efficient can make the whole system worse. This counterintuitive insight appears across management, software engineering, personal productivity, and economics.
## The core problem
When each department, team, or process step optimizes for its own metrics, the system as a whole suffers. This happens because:
- **Parts compete for shared resources**: Two teams both optimizing their output may overload a shared service
- **Local buffers create global waste**: Each step adding safety margin compounds into massive overall lead time
- **Handoff inefficiencies multiply**: Optimizing within silos ignores the cost of transitions between them
- **Metrics misalign**: What's 'good' for a part may be 'bad' for the whole
## Classic examples
### Manufacturing (Theory of Constraints)
Keeping every machine running at 100% utilization seems efficient, but it builds mountains of work-in-progress inventory before bottlenecks. The system produces more by deliberately idling non-constraint resources.
### Software development
Optimizing each developer's utilization to 100% creates a traffic jam. Code reviews pile up, context switching increases, and throughput drops. Teams with slack capacity often deliver faster.
### Organizations
Each department optimizing its budget independently leads to duplicated tools, incompatible systems, and missed collaboration opportunities. The sum of locally optimal departments is not a globally optimal organization.
### Personal productivity
Optimizing every hour for 'productivity' eliminates the slack needed for creative thinking, relationship building, and recovery. Peak individual output requires deliberate inefficiency.
## Why it happens
- **Measurement**: We measure what's easy (local metrics), not what matters (system outcomes)
- **Incentives**: People are rewarded for local performance, not system performance
- **Visibility**: Local effects are visible; system effects are diffuse and delayed
- **Control**: People can control their part but not the whole system
- **Intuition**: 'More efficiency everywhere = more efficiency overall' feels true but isn't
## The solution
1. **Define the global goal**: What is the system trying to achieve?
2. **Identify the constraint**: What single point limits system performance?
3. **Subordinate**: Let local efficiency suffer where it serves global throughput
4. **Measure system outcomes**: Track end-to-end metrics, not just local ones
5. **Accept local 'waste'**: Some slack, idle time, and redundancy is necessary for system health
## Key insight
The goal is not to make every part efficient — it's to make the whole system effective. The best-performing systems have deliberately inefficient parts that serve the global optimum.
Related Concepts
← Back to all concepts