What is a Bottleneck?

How to Identify Bottlenecks

Three signals that a bottleneck exists:

• Work piling up in one stage — a Kanban board where one column has 12 cards while others have 2 shows where work is stuck
• One person always the constraint — if every review goes through the same person and they're always behind, they're the bottleneck
• Upstream idle, downstream starved — when people before the bottleneck have nothing to do because work is backed up, and people after have nothing because work isn't getting through
• Consistent late delivery from one workstream — if the same phase is always late across different projects, look for a structural bottleneck

Bottleneck vs Blocker

These are related but different:

A blocker is a temporary stop — a specific obstacle that's preventing one task from progressing. Fix the blocker and work resumes normally.

A bottleneck is a structural constraint — a capacity or process limitation that consistently limits throughput regardless of which specific tasks are in the queue. Fixing one blocker at a bottleneck doesn't help if the next task immediately backs up behind the same constraint.

Blocblockers are addressed task by task in daily standups. Bottlenecks require systemic intervention: add capacity, redistribute work, or redesign the process.

How to Resolve Bottlenecks

The Theory of Constraints (developed by Eliyahu Goldratt) offers a five-step process:

1. Identify the bottleneck — find the constraint 2. Exploit the bottleneck — maximize its throughput without adding resources (reduce setup time, eliminate wasted capacity, prioritize highest-value work) 3. Subordinate everything else — make all other processes feed the bottleneck optimally; don't push work into the bottleneck faster than it can process 4. Elevate the bottleneck — if the constraint is still limiting overall throughput after exploitation, add capacity (hire, outsource, buy tools) 5. Repeat — once you fix this bottleneck, a new one will emerge elsewhere. Continuous improvement means repeatedly finding and resolving the current constraint.

Bottleneck Analysis for Agency Workflows

Identifying a bottleneck is different from finding where work is slow. The bottleneck is the specific constraint that limits the entire system's throughput. Fix anything else and it won't matter; fix the bottleneck and the whole workflow speeds up.

For agencies, the most common workflow bottlenecks fall into predictable categories. Creative approval is the most frequent: a senior designer or creative director whose approval is required on all work creates a single-point bottleneck. Client approval is another: projects stall while waiting for the client, and no amount of internal process improvement helps.

To find your bottleneck, look at where work accumulates. Cards piling up in the 'Review' column, tasks sitting in 'Awaiting Client' for more than 3 business days on average, or team members consistently waiting for input from one specific person are all bottleneck signals.

In Theory of Constraints, the five focusing steps apply: identify the constraint, exploit it (maximize what it produces), subordinate everything else to it, elevate it (add capacity if needed), and repeat the process because the bottleneck may shift.

• The bottleneck limits the whole system — fixing non-bottleneck steps doesn't increase throughput
• Signals: work accumulating before a step, tasks waiting for one specific person
• Common agency bottlenecks: senior creative approval, client approval, QA, PM sign-off
• Root cause analysis before solution: capacity? skill concentration? process friction?
• After fixing one bottleneck, the constraint shifts — run the analysis again

WIP Limits: The Kanban Solution to Bottlenecks

Work-in-progress limits are Kanban's primary tool for making bottlenecks visible and forcing the team to resolve them. Without WIP limits, work accumulates silently in front of bottlenecks. With WIP limits, the board shows the blockage immediately.

If your 'In Review' column has a WIP limit of 3 and it fills to 3, nobody can push another card into Review. The team is forced to either pull the reviewer away from other work to clear the queue, reassign who can do reviews, or surface that Review is a structural bottleneck.

Setting WIP limits requires judgment. Too low and the limit creates unnecessary friction. Too high and the limit never triggers. A starting rule: WIP limit = 2× the number of people who can work on that stage at once. For a stage where 2 people do reviews, a WIP limit of 4 is a reasonable starting point.

• WIP limits make bottlenecks visible by stopping new work when the queue fills
• When a column hits its WIP limit, the team must clear it before adding more
• Starting formula: WIP limit = 2× the number of people who work that stage
• Adjust WIP limits after observing actual flow — too low creates friction, too high misses signals

Permanent vs Temporary Bottlenecks

Not all bottlenecks require the same response.

A temporary bottleneck exists because of specific, fixable circumstances — a team member is out sick, a new project created a spike in demand for one skill. Temporary bottlenecks resolve themselves or with targeted short-term action.

A permanent bottleneck is structural — it exists because of how the team is organized or how processes are designed. It recurs consistently. Common permanent bottlenecks in agencies: only one senior person can do QA, only one account manager handles all client communications, all design requires one creative director's approval.

The diagnosis matters because the responses differ. Temporary: add short-term capacity, communicate the delay, track the resolution date. Permanent: hire, cross-train, delegate, or redesign the process.

A useful diagnostic question: if this constraint disappeared tomorrow, would the bottleneck reappear within 3 months? If yes, it's permanent and needs structural attention.

• Temporary bottleneck: caused by specific circumstances; resolves with short-term action
• Permanent bottleneck: structural; recurs consistently; requires hiring, training, or process redesign
• Diagnostic: if the constraint disappeared tomorrow, would it reappear within 3 months?
• Growing teams: temporary bottlenecks become permanent at scale — address them proactively