Bottlenecks
A system can contain many problems, but only a few determine how fast the whole system can move. Find the limiting point, protect it and improve it before optimizing everything else.
System performance is often limited by a constraint, so improvement matters most where it can increase the flow of the whole system.
What currently limits the rate at which this system achieves its goal?
Operations, projects, hiring, sales pipelines, content production, approvals, AI workflows and personal productivity.
Fast overview
A bottleneck is a limiting point in a flow. The broader idea is a constraint: the factor that most limits progress toward the system's goal. It may be a machine, person, approval rule, missing information, demand, policy or scarce skill.
Making a nonconstraint faster can create more waiting or unfinished work without increasing total output. Once the current constraint improves, the limiting point can shift.
Mental Model Compass
Work piles up, lead times grow or many improvements fail to improve total output.
Persistent queues, waiting, overloaded approvals, scarce skills and recurring stalls.
If this point gained useful capacity, would the whole system produce more of what matters?
Optimizing every step independently or assuming the busiest resource is the constraint.
Learn it progressively
Define the goal
A constraint only makes sense relative to what the system is trying to achieve.
Map the flow
Trace the main steps, handoffs and dependencies from demand to finished outcome.
Find the limiting point
Look for persistent accumulation, waiting or a step whose effective capacity constrains system output.
Protect and align
Reduce avoidable losses at the constraint and coordinate other work so it supports rather than overwhelms it.
Improve, measure and repeat
Add capacity or redesign when needed. Then measure the whole system again because the constraint may have moved.
Interactive Bottleneck Finder
Enter three major steps. Capacity means useful completed units per period under normal conditions.
Examples
Leadership
A team prepares proposals quickly, but every decision waits for one executive. Better templates may help, but approval capacity or delegation is the likely leverage point.
AI workflow
AI can draft 100 items a day while expert review validates only 12. Generating more drafts increases the queue. Improve review design, triage or quality at source.
Sales
Marketing creates enough leads, but qualified prospects wait days for follow-up. More advertising may add work without increasing closed business.
Worked application
A publishing team keeps missing release dates
Goal: Publish 20 high-quality lessons each month.
Flow: Research can prepare 30, writing can complete 25, expert review can approve 12 and publishing can release 24.
Constraint: Expert review has the lowest effective capacity and a persistent queue.
First move: Protect reviewer time, improve submission quality and route only work ready for expert judgment.
Next move: Standardize routine checks, add qualified review capacity if needed and measure finished lessons rather than local activity.
Recheck: If review rises above 20, another step may become the new constraint.
Common mistakes
Fixing the loudest problem
The most visible queue is evidence, not proof. Confirm that improving it changes system output.
Maximizing every resource
Keeping every step busy can flood the constraint with excess work and increase lead time.
Assuming constraints are physical
Policies, approvals, information, skills and demand can limit a system too.
Forgetting the constraint can move
After improvement, measure again. Yesterday's bottleneck may no longer be today's.
When not to use it
Do not force a single-bottleneck story onto every situation. Some systems have several interacting constraints, variable demand or risks that matter more than throughput. Safety, quality, legal duties and resilience should not be sacrificed merely to increase flow.
Related models
Leverage
A true constraint is often a high-leverage place because a local improvement can change system-level results.
Feedback Loops
Queues and delays provide feedback about where work is accumulating and how the system responds.
Pareto Principle
Both direct attention toward a small number of causes that account for a large share of the result, but a bottleneck is specifically about what limits system flow.
Try It Yourself prompt builder
Save as Skill
Make Bottlenecks reusable in your AI workflow
Use this skill when activity is high but total output, speed or completion is still constrained.