Why Seven Tools?
In the early 1950s, Kaoru Ishikawa — one of Japan's founding quality engineers — observed that factory workers were overwhelmed by complex statistics. He selected seven graphical tools that required no advanced mathematics, could be taught in a few days, and together could analyse virtually any quality problem. The selection was deliberate: these seven tools work on numerical data that workers already collected.
The 7 QC Tools are not the only tools, and they are not always the best tools. But they are the right starting point — and knowing which one fits your current question will save you hours of wasted analysis.
Tool 1 — Check Sheet (チェックシート)
A structured form for recording data in real time, at the location where events occur. Marks are tallied by category so patterns become visible as data is collected — no separate analysis step required.
Classic use: Recording defect types and their frequency on a production shift. After 500 parts, you can see immediately that scratch defects outpace all other types.
Design principle: The form should be completable with a single mark (✓, ////, etc.). If a worker must write words, the form is too complex and data quality will suffer.
Tool 2 — Histogram (ヒストグラム)
A bar chart showing the frequency distribution of a continuous variable. Unlike a time-series chart, a histogram collapses time and shows you what values occur and how often.
Classic use: Plotting 200 shaft diameter measurements to see whether the process is centred, spread too wide, or producing a double-peaked (bimodal) distribution that suggests two different machine setups in the data.
Warning sign: A histogram that is truncated sharply at one side indicates 100% inspection and sorting — the process is out of control and someone is manually removing failures before they reach the chart.
Tool 3 — Pareto Chart (パレート図)
A bar chart sorted from most-frequent to least-frequent, with a cumulative percentage line overlaid. Embodies the Pareto principle: approximately 80% of problems come from 20% of causes.
Classic use: After one month of defect tracking, building a Pareto chart that shows scratch defects represent 62% of all rejects — making the priority decision obvious rather than political.
Critical rule: Sort by cost or impact, not just frequency. A defect that occurs 5 times and causes a line stoppage may rank above one that occurs 50 times but causes only a cosmetic rework.
Tool 4 — Cause-and-Effect Diagram (特性要因図)
Also called the fishbone diagram or Ishikawa diagram. A structured brainstorming framework that organises potential causes into categories — traditionally the 4M: Man, Machine, Material, Method. Sometimes extended to 6M by adding Measurement and Mother Nature (Environment).
Classic use: A team meeting to investigate why the scratch defect rate doubled. The fishbone ensures the team considers all categories rather than immediately blaming the newest machine or newest operator.
The diagram is not the answer. Every bone on the fishbone is a hypothesis. The bones that survive investigation — verified with data — become the real root causes. A fishbone without follow-up data collection is just a wall decoration.
Tool 5 — Scatter Diagram (散布図)
A plot of two variables against each other, one per axis. The shape of the cloud reveals whether a relationship exists: positive correlation, negative correlation, no correlation, or a non-linear relationship that a correlation coefficient would hide.
Classic use: Plotting cutting speed against surface roughness for 60 machined parts. If the fishbone pointed to cutting speed as a cause, the scatter diagram either confirms or refutes it visually before any expensive experiment is run.
Caution: Correlation is not causation. A scatter diagram showing strong correlation is evidence that supports a hypothesis — it is not proof. Always ask: is there a third variable that explains both?
Tool 6 — Control Chart (管理図)
A time-series chart with statistically calculated control limits (UCL and LCL). Points within limits indicate a stable, predictable process. Points outside limits or non-random patterns within limits signal that something has changed and investigation is required.
Classic use: Running an X-bar/R chart on a machining cell. When a point exceeds the upper control limit, the operator stops and investigates the tool — before parts become defective, not after.
Control limits are not specification limits. This is the most common misuse. Control limits are calculated from the process itself. Specification limits come from the drawing. A process can be stable (all points within control limits) and still produce defects (outside spec). These are separate questions.
Tool 7 — Stratification (層別)
Not a chart type, but a data collection and analysis strategy: separating data by source before analysing it. Machines, shifts, operators, raw material lots, and time periods are common stratification factors.
Classic use: A histogram of shaft diameters shows a bimodal (double-hump) distribution. Stratifying by machine reveals that Machine A produces parts centred at 10.02mm and Machine B centres at 9.97mm. Without stratification, both problems are invisible in the combined data.
The first question to ask whenever data looks strange: "Are we mixing data from different sources?" Stratification should be built into every check sheet before data collection starts — not applied as a fix after the data has been collected without source information.
The Quick-Reference Decision Guide
| Question you are asking | Reach for this tool first |
|---|---|
| What is happening and how often? | Check Sheet |
| What is the shape of my process output? | Histogram |
| Which defect costs us the most? Where do I start? | Pareto Chart |
| What could be causing this defect? | Cause-and-Effect Diagram |
| Does this factor actually affect that output? | Scatter Diagram |
| Is my process stable over time? | Control Chart |
| Why does my data look strange / bimodal? | Stratification |
品質管理の基本となる7つの手法：チェックシート、ヒストグラム、パレート図、特性要因図（魚の骨）、散布図、管理図、層別。石川馨博士が1950年代に体系化。統計の専門知識がない現場作業者でも使えるよう設計されており、製造現場の問題のおよそ95%はこれらの手法で解決できるとされる。
One Piece of Advice
Most quality teams I have seen reach for the fishbone diagram first, for every problem. The fishbone is excellent — but it is a brainstorming tool, not a data tool. If you have not yet built a check sheet and drawn a Pareto chart, you do not know which problem to put at the head of the fishbone.
Always start by counting. Count what goes wrong, how often, and at which step. The data will tell you where to focus. Then use the fishbone to generate hypotheses, and the scatter diagram to test them. That sequence — data first, hypotheses second, verification third — is what separates systematic quality improvement from repeated fire-fighting.