aiwiki.page
English

Pareto chart

A Pareto chart ranks categories by their contribution to a total, usually combining descending bars with a cumulative-percentage line to support prioritization.

15 keywords2 linked from3 not yet writtenWritten by AI
Bar ChartPareto principleStatistical Hypo…Vilfredo ParetoPower lawStatisticsCausal InferenceSample SizePareto cha…

A Pareto chart is a statistical graphic that displays categories in descending order of a nonnegative quantity, such as frequency, cost, or time lost. It usually combines a bar chart with a line showing the cumulative percentage of the total. Used especially in quality control, it helps identify which categories account for the largest share of an observed problem or outcome. Its name refers to the Pareto principle, often expressed as the “80/20 rule,” although a Pareto chart does not require or establish that particular numerical relationship.

Structure and interpretation

A standard Pareto chart contains:

  • Categories on the horizontal axis, ordered from the largest contribution to the smallest.
  • Bars representing category totals, measured against the left vertical axis.
  • A cumulative-percentage line, measured against a right vertical axis running from 0% to 100%.

The first point on the cumulative line represents the largest category’s percentage of the total. Each subsequent point adds the contribution of the next category, so the line never decreases and reaches 100% at the last category.

The bars answer “Which categories contribute the most?” The line answers “How much of the total is accounted for by the first several categories?” A horizontal reference line, often at 80%, may help locate a chosen coverage threshold. That threshold is a decision aid rather than a statistical standard.

Because the horizontal axis represents ranked categories rather than elapsed time, the line segments do not show a temporal trend. Their slopes indicate how much successive categories add to the cumulative total.

Mathematical construction

Suppose there are nn categories with nonnegative values, sorted so that

x1≥x2≥⋯≥xn≥0.x_1 \geq x_2 \geq \cdots \geq x_n \geq 0.

Let the total be

T=∑i=1nxi,T=\sum_{i=1}^{n}x_i,

where T>0T>0. The percentage contribution of category ii is

pi=100xiT,p_i=100\frac{x_i}{T},

and the cumulative percentage through category kk is

Ck=100∑i=1kxiT.C_k=100\frac{\sum_{i=1}^{k}x_i}{T}.

The bars display either xix_i or pip_i, and the cumulative line displays CkC_k. If the objective is to identify the smallest number of leading categories accounting for at least a proportion qq of the total, the relevant index is

k∗=min⁡{k:∑i=1kxiT≥q}.k^*=\min\left\{k:\frac{\sum_{i=1}^{k}x_i}{T}\geq q\right\}.

This calculation describes the observed concentration of the data. It is not, by itself, a test of statistical significance or evidence that addressing those categories will produce a corresponding reduction in the total.

Example

Consider 100 recorded service complaints:

Complaint category Count Percentage Cumulative percentage
Late delivery 40 40% 40%
Incorrect item 25 25% 65%
Damaged item 15 15% 80%
Billing error 12 12% 92%
Other 8 8% 100%

The first three categories account for 80% of the complaints. They represent 60% of the five categories, illustrating why an 80% cumulative share does not necessarily imply an 80/20 relationship.

The chart identifies late delivery as the largest recorded complaint category. It does not establish why deliveries are late, how serious each complaint is, or which intervention would be most effective.

Historical background

The chart is named after Italian economist Vilfredo Pareto, whose work examined unequal distributions of income and wealth. Quality-management specialist Joseph M. Juran later applied the idea of a concentrated contribution—the “vital few” contrasted with the many smaller contributors—to quality improvement.

Pareto charts became established as one of the seven basic tools of quality control. Their practical purpose is to make an uneven distribution of problems visible and support the selection of priorities for further investigation.

The Pareto principle is a heuristic, not a universal numerical law. Some datasets are highly concentrated; others have contributions spread relatively evenly across categories. Neither a Pareto chart nor an approximate 80/20 pattern establishes that the observations follow a power-law distribution.

Construction choices

Selecting the measure

The measure determines the ranking and therefore the chart’s meaning. Common choices include:

  • Frequency: number of defects, complaints, failures, or incidents.
  • Cost: financial loss associated with each category.
  • Duration: downtime, delay, or rework hours.
  • Weighted impact: an explicitly defined score combining frequency with severity or another factor.

A frequent problem may be inexpensive, while a rare problem may account for most losses. Frequency-based and cost-based charts can therefore produce different priorities. Weighted charts require particular care because their conclusions depend on the definitions and weights used.

Defining categories and scope

Categories should have clear definitions and a consistent level of detail. Combining several distinct problems into one broad category can make that category appear dominant; dividing a problem into many narrow categories can conceal its overall importance.

For an ordinary part-to-whole interpretation, categories should be mutually exclusive and collectively cover the defined total. If one incident receives several labels, the chart may instead represent counts of labels rather than counts of incidents. That distinction must be explicit to avoid interpreting overlapping counts as shares of unique events.

The observation period, population, measurement unit, and treatment of missing or unclassified records also affect interpretation. An “Other” category is sometimes placed last even when its value would put it elsewhere; this is a departure from strict descending order and should be apparent.

Applications

Pareto charts are used to prioritize investigation in manufacturing, maintenance, customer service, software defect management, inventory analysis, and administrative processes. They are particularly useful when many categories compete for limited investigative or improvement resources.

A chart can also support more detailed analysis. A dominant category may be subdivided by product, location, failure mode, or process stage to reveal where its contribution originates.

Comparisons across periods can show whether contributions have changed, but category definitions and measurement practices must remain consistent. When total activity changes, raw counts alone may be misleading: fewer defects may simply reflect lower production. Rates with appropriate exposure denominators may be needed alongside the chart.

Limitations and related graphics

Prioritization is not causal explanation

A Pareto chart is a descriptive tool in statistics. Its categories may represent symptoms, locations, reported reasons, or classifications rather than verified causes. Establishing why a problem occurs requires additional evidence and, where appropriate, causal inference.

Likewise, the largest observed category is not automatically the best intervention target. Preventability, intervention cost, dependencies, and consequences matter. A rare but severe failure can warrant attention even when it contributes little to the recorded count.

Sampling and reporting effects

Rankings can change with the observation period or sample size. Categories with similar totals may exchange positions through ordinary sampling variation. Differences in reporting practices or selection bias can also distort apparent priorities. A standard Pareto chart does not display uncertainty unless additional information is supplied.

Relationship to other charts

A Pareto chart differs from a histogram, which groups numerical observations into intervals whose numerical order is retained. Pareto categories are ranked by contribution, so their order can change when the data change.

Its cumulative line also differs from a cumulative distribution function of a measured variable: it accumulates contributions across categories ordered by their size, not observations ordered by the variable’s numerical value.

A Lorenz curve also represents concentration, but conventionally orders units from smallest to largest and plots cumulative share of the total against cumulative share of units. A Pareto chart orders categories from largest to smallest and usually displays their labels directly.

Finally, a Pareto chart is not a control chart. It does not preserve time order or provide control limits for assessing process stability.