← Back to all heuristics

Avoid Local Optimization

Optimize the system, not the silo.

Flow

Heuristic

Avoid optimizing a single step at the expense of the whole system.

Hook

Local wins can be global losses.

Problem

Teams often optimize their own part of the pipeline without considering the full system. This can increase throughput locally while creating queues, rework, or delays elsewhere.

Why it matters

System throughput is limited by its constraint. The theory of constraints reminds us that the slowest step sets the pace. When you optimize locally, you can shift pain downstream and make overall delivery slower.

Signals you are here

  • Team metrics are green but delivery is slow
  • Upstream output overwhelms downstream capacity
  • Queues grow at a single stage
  • More coordination meetings are needed to move work

Anti-patterns

  • Per-team KPIs that ignore end-to-end flow
  • Maximizing utilization instead of throughput
  • Automating steps that just create more backlog
  • Blaming downstream teams for delays

Try this

  • Align on system-level metrics (lead time, throughput)
  • Map the value stream and fix the constraint first
  • Reduce handoffs and shared queues
  • Design cross-team goals and shared ownership

Example

A build team speeds up builds by 50 percent, but releases slow down because testing becomes the bottleneck. Instead of pushing builds faster, they invest in automated testing and the system throughput improves.

Reflection prompt

Where did a local improvement make someone else's work harder? Fix that connection.

More like this

Heuristic

Optimize for Flow, Not Silos

If downstream is blocked, upstream speed is just inventory.

Flow

Heuristic

Empower Autonomous Teams

Autonomy with guardrails.

CollaborationFlow

Heuristic

Every Output Is Someone Else's Input

Handoff quality sets the pace of flow.

FlowCollaboration

Heuristic

Limit Work in Progress

WIP is invisible debt on your flow.

Flow

Heuristic

Short Feedback Loops

Fast feedback beats perfect plans.

FlowLearning

Heuristic

Work in Small Batches

Small batches make failure cheap and learning fast.

FlowAutomationDelivery