Complete firstProcess Mapping
An up-to-date process map with steps, handoffs, and processing or waiting times is available.
Without: Without a process picture, bottleneck analysis becomes a hypothesis and the bottleneck lacks evidence.
RunsheetOperations
Bottleneck Analysis
ComplexityMedium
Time1-3 h
Participants3-8
FormatWorkshop + async
MaturityEstablished
01
Prerequisite
What needs to be finished first
Complete firstValue Stream Mapping
For value-stream-oriented processes, a value stream map with cycle time, lead time, and WIP per step is available.
Without: Without time and WIP data, the bottleneck suspicion stays intuitive and optimization targets the wrong step.
02
Preparation
What needs to be ready before start
Materials
Whiteboard with a process or value-stream overview; data on throughput, WIP, processing time, and waiting time per step for at least 4 weeks; template for bottleneck table with step, cycle time, WIP, queue indicator; Little's Law formula sheet.
People / roles
One analyst or coach with lean knowledge; 3-6 people with process context (operations, engineering, data); one scribe; sponsor with authority to approve measures.
Pre-read
Current throughput data per step; observations about waiting times and queues; demand profile; known complaints; planned process changes.
Time needed
2-4 h for the initial analysis, then iteration depending on complexity
Setup
Make the process or value-stream map visible. Add columns for cycle time, WIP before the step, waiting time, and throughput per step. Prepare the bottleneck table. Note: the bottleneck is the step with the highest WIP before it and throughput below demand.
03
Core question
The one question this method answers
Where does work accumulate in the process, which step limits overall throughput, and which escalation level, utilization, subordination, or elevation, is the next effective move?
04
Flow
Marker: Phase
| Step | Duration | Action | Hint |
|---|---|---|---|
1Phase 1: Consolidate data | 30-45 min | Enter cycle time, WIP before the step, waiting time, and throughput per process step. Document the data source and period. Mark gaps. | If data gaps exceed 30 percent, do data collection first. Estimates are a fallback, not the basis. Define a collection plan with owner for each gap. |
2Phase 2: Identify the bottleneck | 30-45 min | Find the step with the highest WIP before it and throughput below demand. If several candidates exist, use Little's Law as a test. Agree on one bottleneck. | A common mistake is to identify the bottleneck by feeling. Data first. If data suggest multiple bottlenecks, plan two iterations instead of attacking both at once. |
3Phase 3: Clarify bottleneck character | 30-45 min | Classify the bottleneck type: capacity (people, machine), variability (swings break flow), quality (returns), policy (rule that limits throughput). Base the classification on evidence. | The bottleneck type determines the measures. A capacity bottleneck needs elevation, a variability bottleneck needs buffering, and a policy bottleneck needs a rule change. |
4Phase 4: Develop measures | 45-60 min | Create suitable measures for each bottleneck type. Order: utilization (no interruptions), subordination (adapt the rest to the bottleneck), elevation (increase capacity). Assign owner and date to each measure. | Elevation is the most expensive level. Anyone who hires people immediately without checking phases 2 and 3 is optimizing locally and spending budget unnecessarily. Keep the order. |
5Phase 5: Impact review | 30 min, after 2-4 weeks | Measure again: has the bottleneck been relieved, moved, or stayed the same. If it moved, identify the new bottleneck and start the next iteration. If it stayed, review the measures. | Bottleneck movement is success. Anyone who treats the same bottleneck for 3 months has either not implemented the measure or is dealing with a structural, not operational, bottleneck. |
05
Artifact
What comes out at the end
Form
Bottleneck table with data per step, identified bottleneck and bottleneck type, derived measures with owner and date, review result after 2-4 weeks. Trend overview across multiple iterations, showing bottleneck movement.
Versioning / ownership
One entry per iteration with date and data snapshot. Do not overwrite history; keep bottleneck movement visible.
Tool alternatives
- Confluence page with table and plot embed
- Notion database with iteration entries
- Miro board with process map and bottleneck highlight
- Markdown in the repo under docs/operations/bottleneck-<date>.md
bottleneck-analysis-working-template.md
Compact working template for Bottleneck Analysis with context, input, output artifacts, and next step.
Bottleneck Analysis Canvas
Context
What is this method used for?
Core question
Which question should be answered at the end?
Input
Which data, observations, or materials are available?
Working area
- Area 1:
- Area 2:
- Area 3:
- Relationships / patterns:
Output artifacts
- Bottleneck Map:
- Flow Metrics:
- Improvement Options:
- Follow-up Measures:
Open questions
- ...
Next step
Owner, date, and success signal.
06
Example output
Concrete filled scenario, fictional example
bottleneck-analysis-beispiel.md
Concrete filled scenario, fictional example
Bottleneck Analysis — SaaS support pipeline Q2 2026, iteration 2, 12.05.2026
Data: last 4 weeks, source Zendesk and JIRA.
Bottleneck table
| Step | Cycle Time | WIP | Waiting Time | Throughput | Note |
|---|---|---|---|---|---|
| Intake | 5 min | 12 | 0.5 h | 32/day | Auto |
| L1 triage | 20 min | 8 | 1 h | 30/day | OK |
| Tier-2 diagnosis | 90 min | 24 | 11 h | 11/day | Bottleneck |
| Engineering fix | 4 h | 6 | 12 h | 10/day | OK relative to Tier-2 |
| Verification | 30 min | 4 | 2 h | 12/day | OK |
Bottleneck character
Capacity bottleneck at Tier-2 diagnosis: three engineers, 90 min per ticket, no spare capacity. Variability moderate. No policy limit.
Measures
- Utilization: keep diagnosis slots uninterrupted, no meetings from 10-12 and 14-16; pre-triage by L1 with standard checks reduces diagnosis time by an estimated 25%. Owner @ben, from 15.05.
- Subordination: set WIP limit 'in diagnosis' to max 6; tickets wait in triage instead of blocking the queue. Owner @sabine, from 15.05.
- Elevation: if WIP is still above 10 after 2 weeks, add one person to Tier-2. Decision on 29.05. based on data check.
Movement
Iteration 1 (April): bottleneck was L1 triage. After tooling improvement, the bottleneck moved to Tier-2, as expected.
07
Pitfalls
Recognize symptoms and steer against them
Trap
Bottleneck chosen by feeling
Symptom
The team identifies the bottleneck by complaint volume, while data disagree or are missing.
What to do
Use data first, then discuss. Measure bottleneck candidates against data, not against complaints. Complaints are a source of hypotheses, not proof.
Trap
Bottleneck type unclear
Symptom
A bottleneck is identified, but its character is not classified, so measures hit the wrong lever.
What to do
Clarify the bottleneck type before measures: capacity, variability, quality, policy. Each type needs a different family of measures. A wrong diagnosis creates expensive wrong measures.
Trap
Local optimization away from the bottleneck
Symptom
The team optimizes the upstream step, WIP before it drops, and WIP before the bottleneck keeps growing.
What to do
Clear rule: only the bottleneck step is optimized in phase 2. Other steps are subordinated and deliberately not accelerated. Pause local efficiency metrics.
Trap
Elevation too early
Symptom
The team hires people before utilization and subordination have been tried. The investment is costly and the effect is small.
What to do
Keep the order: utilization, subordination, then elevation. Pre-elevation check: is the bottleneck fully utilized, and is subordination in place?
Trap
No review after 2-4 weeks
Symptom
Measures are implemented, but no review follows and nobody knows whether the bottleneck moved.
What to do
Fix the review date during setup. If the bottleneck moved, start a new iteration. If not, review the measures or the diagnosis.
08
Stop criteria
Done signals checkable in under a minute
No reliable data on throughput and waiting time per step, so the bottleneck cannot be identified.
The process is not stable or is new, the bottleneck changes daily, and the method assumes a steady state.
Political conflict over the bottleneck definition causes data to be ignored.
The bottleneck lies outside the steering range, for example with an external supplier or a regulatory constraint, so the method needs other levers.
No sponsor for subordination is available and non-bottleneck steps insist on local efficiency.
No review cadence is established, so the method remains a one-off snapshot.
Finished the runsheet?
Go to the profile for purpose, similar methods, and sources or continue to the next method in the catalog.