Why Rank Categories?
A frequency table can list every category accurately, but it may take effort to spot which categories occur most often, especially when the original list is not ordered. A Pareto chart makes that ranking immediate: its bars are arranged from largest count to smallest. This is useful when deciding which complaint types, defect types, or other categories account for the most recorded cases.
A Pareto chart describes the data that were counted. It can show which category is most common in the recorded set, but the chart alone does not explain why that category occurs or establish that addressing it will cause a particular outcome. As with any display of categorical data, the category labels, units being counted, and total number of observations should be clear.
Definition and Construction
To construct one, first decide what one counted observation represents—for example, one complaint or one defective item. Count observations in each category, then reorder the categories by decreasing count. Draw one bar for each category, using a common vertical scale that starts at zero. Label both axes, identify the units being counted, and make the category labels easy to read. The order of categories in the original data table does not determine their order in the chart.
The bars show counts, not the categories' names or importance. A taller bar means that more observations in the data set were assigned to that category. It does not necessarily mean the category is more serious, costly, or preventable. Those are different questions and require additional information.
Worked Example: Ranking Customer Complaints
Suppose a hypothetical store sorts 40 customer complaints received in one month into four mutually exclusive categories, so that each complaint is counted once. The frequency table, in the order the categories were first listed, is:
| Complaint category | Count |
|---|---|
| Billing issue | 5 |
| Delayed delivery | 18 |
| Damaged item | 10 |
| Wrong item | 7 |
Check the total before charting: \(5+18+10+7=40\), which agrees with the number of complaints. Sorting the counts from largest to smallest gives \(18, 10, 7, 5\). The bars in the Pareto chart should therefore appear in this order:
| Bar order | Category | Bar height (complaints) |
|---|---|---|
| 1 | Delayed delivery | 18 |
| 2 | Damaged item | 10 |
| 3 | Wrong item | 7 |
| 4 | Billing issue | 5 |
A chart with a vertical count scale from 0 to 20 could show bars with heights 18, 10, 7, and 5, respectively. The first bar is the tallest, so delayed delivery was the most frequently recorded complaint among these 40 complaints. The chart does not show that delayed delivery was the most costly or the most serious complaint.
Adding Percentages and a Cumulative Line
A Pareto chart may also include a cumulative percentage line. To calculate it, divide each category's count by the total number of observations to obtain its percentage, then add the percentages in the chart's left-to-right order. The cumulative percentage after a category answers: what percentage of all observations belong to this category or to any category placed before it?
For a category with count \(f_i\), among \(N\) observations, its percentage is \(100f_i/N\). If the categories have been ranked from largest count to smallest, the cumulative percentage through the first \(k\) categories is:
This result follows directly from the definition of a percentage: the first \(k\) categories contain \(f_1+\cdots+f_k\) observations out of \(N\) total observations. Dividing that subtotal by \(N\) gives the fraction of all observations in those categories; multiplying by 100 expresses the fraction as a percentage. If the categories include every observation exactly once, the cumulative percentage at the final category is \(100\%\).
Worked Example: Calculating Cumulative Percentages
Use the 40 hypothetical complaints in the previous example. In descending order, the counts are 18, 10, 7, and 5. Calculate each category's percentage using \(100f_i/N\), with \(N=40\):
The cumulative percentages are found by adding in the same ranked order. After delayed delivery, the cumulative count is 18, so the cumulative percentage is \(100(18/40)=45\%\). After damaged item, it is \(100((18+10)/40)=100(28/40)=70\%\). After wrong item, it is \(100((18+10+7)/40)=100(35/40)=87.5\%\). After billing issue, it is \(100(40/40)=100\%\).
Thus the first two categories together account for \(28\) of the \(40\) complaints, or \(70\%\). A cumulative line, if included, would rise through 45%, 70%, 87.5%, and 100% from left to right. It is a separate summary from the bars: the bars show individual category counts, while the line shows the accumulating share. If both are drawn, the chart needs a clearly labelled percentage scale for the line.
Reading the Ranking in Context
The descending order makes the largest categories easy to locate, but the interpretation should stay close to what was measured. If a Pareto chart of hypothetical repair reports has “battery failure” as its tallest bar, an appropriate conclusion is that battery failure was the most frequently recorded repair category in the data set. It would not be justified to conclude from the chart alone that battery failure was the most expensive problem or that it caused the most customer dissatisfaction.
A cumulative percentage can help describe how many leading categories make up a chosen share of the total. For example, the first two categories in the complaint example account for 70% of recorded complaints. This is a summary of those observations, not a guarantee that the same percentage will occur in future complaints. It also does not determine which category should be addressed first: frequency is only one possible consideration.
Worked Example: A Pareto Chart of Hypothetical Defects
Suppose a hypothetical workshop inspects 60 items and records one defect category for each defective item. The recorded counts are: loose fastener, 4; surface scratch, 12; paint blemish, 5; misaligned part, 9. Assume the categories are mutually exclusive and cover all 30 defects observed; the other 30 inspected items had no defect and are outside this defect-category table.
First check the defect-category total: \(4+12+5+9=30\). Rank the categories from largest to smallest: surface scratch (12), misaligned part (9), paint blemish (5), and loose fastener (4). These are the bar heights in that order. The highest bar is surface scratch, so it is the most common recorded defect category among the 30 defects.
For a cumulative percentage based on the 30 defects, the first two ranked categories contain \(12+9=21\) defects:
Therefore, surface scratches and misaligned parts together account for 70% of the categorized defects. The denominator is 30, because this calculation describes the distribution among defects. If the question instead concerned the fraction of all 60 inspected items that had either of these defects, the calculation would be \(21/60=0.35\), or 35%. Stating the denominator and what is included in the count prevents these two different percentages from being confused.
Counts Need a Fair Comparison
A Pareto chart ranks counts. That is appropriate when the counted observations are comparable and the question concerns the number of recorded cases in each category. However, raw counts can be a poor basis for comparing underlying rates when categories have different opportunities to produce a case. For example, a machine used for 1,000 hours may produce more defects than a machine used for 100 hours simply because it operated longer. A chart of defect counts can identify which machine produced more recorded defects, but it does not by itself compare defects per operating hour.
Similarly, a category with a larger count may have been observed over a longer period, in a larger population, or under more frequent inspection. When those exposures differ, interpret the chart as a ranking of observed counts, not as a ranking of risk or rate. The correct comparison depends on the question and the available denominators.
Worked Example: Counts Versus Rates
Imagine two hypothetical production lines. Line A operates for 100 hours and has 12 recorded defects. Line B operates for 400 hours and has 30 recorded defects. A Pareto-style ranking by counts puts Line B first because \(30>12\). That answers which line had more recorded defects over its stated operating period.
If the question is defects per operating hour, calculate each rate using its own exposure:
Line B has more defects in total, but its recorded defect rate per operating hour is lower. Thus, the count ranking and the rate comparison answer different questions. A Pareto chart of raw counts should not be described as showing which line has the higher defect rate.
Common Mistakes
- Leaving categories in their original order. A bar chart is not a Pareto chart if its categories have not been ranked by count. Reorder the bars from largest to smallest before interpreting the display.
- Reading the first bar as the most important category. The first bar is the most frequent category in the counted data. Importance, cost, severity, and possible causes require other information.
- Mixing up category percentage and cumulative percentage. A category's percentage is its own count divided by the total. A cumulative percentage adds that category's count to the counts of all categories before it in the ranked order.
- Ignoring the denominator. A percentage among recorded defects is not automatically the percentage of all inspected items. State which observations are included in the total.
- Comparing counts despite unequal exposure. More recorded cases do not necessarily mean a higher rate when groups were observed for different durations or had different numbers of opportunities for cases.
A final check is to confirm that the counts sum to the stated total, the bars are ordered correctly, and the tallest bar's interpretation names the counted unit and context. If a cumulative line is used, its last value should be 100% when all observations are represented once.
Check Your Understanding
Answer each question using the definitions and calculations from this tutorial.
- A hypothetical help desk records 14 password-reset requests, 8 login errors, and 11 software-installation requests. In what order should the categories appear in a Pareto chart?
- For counts 20, 12, and 8, what is the cumulative percentage after the first two categories? Show the denominator and calculation.
- In a Pareto chart of complaints, the tallest bar represents “late delivery.” What conclusion is supported, and name one conclusion the chart alone does not support.
- Why should the final cumulative percentage equal 100% if all observations belong to exactly one of the displayed categories?
- Two sites record 16 and 24 incidents, but the second site was monitored for twice as long. What does ranking the counts tell you, and why does it not by itself compare incident rates?