Descriptive Statistics and Data Visualization

Introduction

International business decisions are made from data generated across countries, currencies, markets, customers, suppliers, employees, and digital channels. Descriptive statistics provides a disciplined way to organize, summarize, and communicate those data before managers attempt prediction or causal explanation. A regional sales director may need to know not only the average order value, but also whether sales are concentrated in a few markets, whether delivery times are unusually variable, and whether advertising expenditure is associated with revenue.

Descriptive analysis does not establish that one variable causes another. Its purpose is to reveal the main features of observed data accurately and transparently. Effective visualization then converts those features into accessible managerial evidence. Poorly designed tables or charts can, however, exaggerate differences, conceal uncertainty, or encourage inappropriate comparisons.

This chapter develops methods for frequency distributions, central tendency, dispersion, position, shape, association, tables, charts, dashboards, and managerial narratives.

Learning Objectives

By the end of this chapter, you should be able to:

1. Construct and interpret frequency and relative-frequency distributions.
2. Select appropriate measures of central tendency and dispersion.
3. Calculate and interpret percentiles, quartiles, z-scores, skewness, and correlation.
4. Choose charts that match the structure and purpose of a dataset.
5. Design concise, truthful dashboards and managerial narratives.
6. Recognize misleading visualizations, data-quality problems, and ethical risks.
7. Use descriptive evidence appropriately in international business decisions.

1. Data, Variables, and the Descriptive-Statistics Workflow

A population is the complete set of units of interest, such as all customers who purchased from a company during a year. A sample is a subset observed from that population. An observation is one recorded unit, and a variable is a characteristic measured for each observation.

Variables may be:

• Categorical, such as country, payment method, or customer segment.
• Quantitative, such as revenue, delivery time, or number of units.
• Discrete, taking countable values, such as number of orders.
• Continuous, capable of taking values within an interval, such as weight or temperature.

Measurement scales affect the statistics that are meaningful. Nominal categories have labels only; ordinal categories have an order; interval variables have meaningful differences but no meaningful zero; ratio variables have both meaningful differences and a meaningful zero. Revenue and quantity are ratio variables, whereas country is nominal.

A sound descriptive workflow is:

1. Define the managerial question and unit of analysis.
2. Inspect data types, missing values, duplicates, and outliers.
3. Summarize frequencies and key numerical measures.
4. Visualize distributions and relationships.
5. Compare results across relevant groups.
6. Communicate findings with limitations and context.

The unit of analysis must be explicit. For example, “average sales per customer” differs from “average sales per transaction.” Combining them can produce a misleading conclusion.

2. Frequency Distributions

A frequency distribution records how often values or categories occur. For category i, the relative frequency is:

where fᵢ is the frequency of category i and n is the total number of observations. The percentage frequency is 100pᵢ%. Frequencies must satisfy ∑fᵢ = n, while relative frequencies must sum to 1, apart from rounding.

For quantitative data with many distinct values, observations can be grouped into class intervals. A useful approximate class width is:

where k is the selected number of classes. This is a planning guideline, not a universal rule. Class intervals should be mutually exclusive and collectively exhaustive. Boundary conventions are essential: intervals such as 0–9 and 10–19 avoid overlap, whereas 0–10 and 10–20 do not unless a precise inclusion rule is given.

Worked Example: Delivery-Time Frequencies

Suppose 20 international deliveries have times, in days, grouped as follows:

The table shows that 60% of deliveries arrived within six days. It does not show the exact values within each interval, so calculations based on grouped data are approximations.

For ordered data, a cumulative frequency counts observations at or below a specified value or class. A cumulative relative frequency expresses that count as a proportion of n. Cumulative distributions are useful for service-level questions, such as “What percentage of orders arrived within seven days?”

3. Measures of Central Tendency

Central tendency describes a typical or central value. The appropriate measure depends on the scale, distribution, and decision context.

3.1 Mean

For a sample of n quantitative observations x₁, x₂, …, xₙ, the arithmetic mean is:

For a population, the corresponding notation is μ = (∑Xᵢ)/N, where N is the population size. The mean uses every observation and is appropriate for many approximately symmetric quantitative distributions. It is sensitive to extreme values.

For observations with frequencies fᵢ and values xᵢ, the weighted mean is:

A weighted mean is appropriate when observations contribute unequally, such as country sales weighted by the number of transactions. The weights should be clearly defined and should not be used merely to obtain a preferred result.

3.2 Median

The median is the middle ordered observation. If n is odd, it is the observation at position (n + 1)/2. If n is even, it is commonly the average of the observations at positions n/2 and n/2 + 1. Different software may use different percentile interpolation conventions, so the method should be documented.

The median is resistant to extreme values. For income, order value, and delivery time—often right-skewed—the median may represent a typical case better than the mean.

3.3 Mode

The mode is the most frequently occurring value or category. It is the only central-tendency measure directly suitable for nominal data. A distribution may have one mode, several modes, or no repeated value. In business, the modal payment method or most common product category can be operationally important even when a mean is not meaningful.

Example

For order values of 80, 90, 100, 110, and 620 currency units:

• Mean = 1,000 ÷ 5 = 200.
• Median = 100.

The mean is pulled upward by the 620-unit order. Reporting only the mean would give an incomplete picture.

4. Measures of Dispersion

Two datasets can have the same mean but very different consistency. Dispersion measures quantify variability.

4.1 Range and Interquartile Range

The range is:

It is simple but depends entirely on two observations. The interquartile range (IQR) is:

where Q₁ is the first quartile and Q₃ is the third quartile. The IQR describes the spread of the middle 50% and is resistant to extreme values.

A common exploratory rule flags observations below Q₁ − 1.5(IQR) or above Q₃ + 1.5(IQR) as potential outliers. This is a screening rule, not proof that an observation is erroneous. A high-value international order may be genuine and commercially important.

4.2 Variance and Standard Deviation

For a sample, variance is:

and sample standard deviation is:

For a population, divide by N rather than N − 1 and use σ² and σ. The sample denominator n − 1, called Bessel’s correction, provides an unbiased estimator of population variance under the usual random-sampling assumptions.

Variance is measured in squared units; standard deviation has the same units as the original variable and is therefore easier to interpret. Standard deviation is most informative when the distribution is reasonably symmetric and the mean is meaningful.

Worked Calculation

For delivery times 2, 3, and 7 days:

• Mean: x̄ = (2 + 3 + 7)/3 = 4 days.
• Deviations: −2, −1, and 3.
• Squared deviations: 4, 1, and 9; sum = 14.
• Sample variance: s² = 14/(3 − 1) = 7 days².
• Sample standard deviation: s = √7 ≈ 2.65 days.

The standard deviation indicates that delivery times typically vary by several days around the sample mean, but it should not be interpreted as a guarantee that every delivery lies within 2.65 days of the mean.

4.3 Coefficient of Variation

For a positive mean, the coefficient of variation is:

Often it is reported as 100s/x̄%. It is dimensionless and useful for comparing relative variability across variables with different units or substantially different means. It is inappropriate when the mean is zero, close to zero, or not substantively meaningful.

5. Measures of Position and Standardization

Quartiles divide ordered data into four parts. Percentiles divide data into one hundred parts. The 90th percentile is a value at or below which approximately 90% of observations fall, subject to the chosen percentile convention.

A z-score standardizes an observation relative to a mean and standard deviation:

For a population, μ and σ replace x̄ and s. The z-score is dimensionless. A z-score of 2 means the observation is two standard deviations above the reference mean, not that it is necessarily abnormal. Interpretation depends on the distribution and business context.

Standardization supports comparisons across countries when variables use different currencies or scales, but currency conversion and purchasing-power considerations must be addressed before comparing monetary data.

6. Shape of a Distribution

Shape concerns symmetry, concentration, tails, and modality.

• A symmetric distribution has broadly balanced sides around its center.
• A right-skewed distribution has a long upper tail; income and order values commonly show this pattern.
• A left-skewed distribution has a long lower tail.
• A unimodal distribution has one prominent peak; a bimodal distribution has two, possibly indicating two customer segments or market regimes.

In a clearly right-skewed distribution, the mean is often greater than the median because large values pull the mean upward. This relationship is descriptive, not a universal law.

Histograms display quantitative distributions by placing observations into adjacent intervals. Bin width affects the apparent shape: very narrow bins can create noise, while very wide bins can hide structure. Box plots show the median, quartiles, IQR, and potential outliers compactly and are particularly useful for comparing countries or products.

7. Association Between Quantitative Variables

A scatterplot displays paired observations and can reveal direction, form, clusters, and unusual points. Association may be positive, negative, nonlinear, weak, or strong.

The Pearson sample correlation coefficient is:

Here xᵢ and yᵢ are paired observations, and x̄ and ȳ are their sample means. When both variables have nonzero variation, −1 ≤ r ≤ 1. A positive value indicates that larger values of one variable tend to accompany larger values of the other; a negative value indicates the opposite linear tendency.

Correlation is unit-free and measures linear association. It can be distorted by outliers, does not adequately describe nonlinear relationships, and does not establish causation. For example, advertising and sales may correlate because both rise during holiday periods. A third variable, seasonality, may influence both.

For ordinal data or strongly non-normal data, rank-based measures such as Spearman’s rank correlation may be more suitable. The choice should reflect measurement level and assumptions rather than convenience.

8. Effective Tables and Visualizations

A visualization should answer a defined question. Chart selection should follow the analytical task.

Tables

A good table has a descriptive title, named variables, units, time period, denominators, and a source. Percentages should identify whether they are calculated by row, column, or total. Totals should reconcile, and rounding differences should be explained where necessary.

Charts

Axes must be labeled, scales must be appropriate, and colors should carry meaning rather than decoration. For bar charts, a zero baseline is normally required because bar length encodes magnitude. A line chart may use a nonzero axis when the purpose is to show variation around a meaningful benchmark, but the truncated scale must be conspicuous. Three-dimensional effects, excessive colors, dual axes, and decorative images often reduce accuracy.

Maps require special care. A country with a large geographic area may appear visually dominant even when its sales are modest. Per-capita, per-customer, or per-unit measures may be more appropriate than total values. Choropleth maps should generally use rates or standardized values, not raw counts, when population or market size differs substantially.

9. Dashboards and Managerial Narratives

A dashboard is a coordinated display of metrics intended to support recurring decisions. It should begin with decisions, not available data. A useful dashboard normally includes:

1. A defined audience and purpose.
2. A small number of prioritized indicators.
3. Current value, target, prior-period comparison, and time frame.
4. Consistent definitions and filters.
5. Drill-down capability where operational action is required.
6. Visible data freshness, source, and ownership.
7. Accessibility for color-blind and screen-reader users where relevant.

A managerial narrative should connect evidence to action without overstating certainty. A practical structure is:

• What happened? State the level, change, and relevant distribution.
• Why might it matter? Explain operational or strategic significance.
• What should be investigated or done? Recommend an action.
• What remains uncertain? Identify data limitations and alternative explanations.

For example: “Median delivery time fell from 6 to 5 days, while the 90th percentile remained at 11 days. Typical service improved, but late deliveries remain concentrated in two customs corridors. Operations should investigate those corridors before claiming a general service improvement.” This is stronger than saying simply, “Delivery performance improved.”

10. Illustrative Case: Global E-Commerce Fulfilment Dashboard

Background

The fictional company Meridian Commerce sells consumer products in Europe, Asia, and Latin America. Managers report average delivery time, total sales, and customer complaints each month. Executives believe that the region with the highest average delivery time has the worst service.

Problem

The average is calculated from transaction-level data, while order volumes and product mixes differ by region. A few exceptionally delayed shipments may also affect the mean. The dashboard uses a truncated bar chart that makes a difference of 0.4 days appear visually large.

Analysis

The analytics team validates the unit of analysis and calculates mean, median, IQR, and the 90th percentile by region. It also reports the percentage delivered within the seven-day service promise.

Interpretation and solution

Latin America has the lowest median and the highest percentage within seven days, but also the largest IQR and 90th percentile. Most deliveries are relatively fast, while a smaller group experiences severe delays. Asia has a similar mean but a higher median and a lower service rate than Latin America. Therefore, the mean alone conceals materially different service distributions.

The redesigned dashboard uses:

• KPI cards for median delivery time and percentage within promise.
• Box plots by region to show spread and outliers.
• A line chart for monthly service-rate trends.
• A table showing the 90th percentile and shipment volume.
• A filter for customs corridor and product category.

Outcome

Managers direct process improvement toward the small number of Latin American customs corridors associated with extreme delays rather than imposing a broad intervention on the entire region. They also avoid ranking regions solely by average delivery time.

Learning points

1. Central tendency must be paired with dispersion and service thresholds.
2. A dashboard metric should match the managerial decision.
3. Visual scale can change perceived importance without changing data.
4. Segmentation can reveal operationally distinct distributions.
5. Descriptive statistics identify where to investigate; they do not by themselves prove the cause of delay.

11. Misleading Visualizations, Data Quality, and Ethics

Misleading communication can result from deliberate manipulation or from poor design. Common problems include:

• Truncated bar-chart axes that exaggerate small differences.
• Unequal time intervals or missing periods in a trend line.
• Cherry-picked dates, categories, or favorable denominators.
• Percentages without counts, especially for small samples.
• Mixing currencies, inflation levels, or exchange-rate bases.
• Suppressing missing data or treating it as zero.
• Double-counting customers or transactions.
• Using averages when distributions are highly skewed.
• Comparing rates with different populations at risk.
• Inferring causation from correlation.

Ethical descriptive analytics requires transparency about definitions, exclusions, transformations, uncertainty, and data provenance. Personal data should be minimized, protected, and aggregated when individual identification is unnecessary. International businesses must consider applicable privacy and data-protection requirements, cross-border transfers, access controls, and retention policies. Dashboard access should follow role-based permissions because even aggregated data can reveal sensitive commercial information in small markets.

AI-enabled tools can automatically generate charts or narratives, but they may select inappropriate scales, omit caveats, hallucinate explanations, or reproduce biased historical patterns. Human review should verify calculations, labels, definitions, subgroup effects, and source data. Automation should assist interpretation, not replace accountability.

12. Best-Practice Checklist

Before publishing a descriptive analysis, ask:

• Is the business question and unit of analysis explicit?
• Are variables measured on appropriate scales?
• Do frequencies and percentages reconcile?
• Have missing values, duplicates, and outliers been investigated?
• Does the selected center match the distribution?
• Is dispersion reported with central tendency?
• Are comparisons based on comparable denominators, periods, and currencies?
• Does the chart type match the analytical purpose?
• Are axes, units, sources, and time frames visible?
• Could a reasonable reader draw a causal conclusion that the data do not support?
• Are privacy, fairness, accessibility, and governance requirements satisfied?

Chapter Summary

Descriptive statistics organizes observed data into interpretable evidence. Frequency distributions show how observations are allocated across categories or intervals. The mean, median, and mode summarize central tendency, while range, IQR, variance, standard deviation, and coefficient of variation describe dispersion. Percentiles and z-scores express position, and distributional shape reveals skewness, modality, concentration, and tails. Scatterplots and correlation summarize association but do not establish causation.

Tables and visualizations should be designed around managerial questions. Bar charts compare categories, line charts show time trends, histograms and box plots describe distributions, scatterplots examine paired quantitative data, and maps require careful attention to denominators and geographic representation. Dashboards should combine relevant indicators, targets, trends, definitions, and data-quality information. Ethical practice requires transparency, appropriate privacy protection, accessible design, and human review of automated analytics.

Key Terms

Population: The complete set of units of interest.

Sample: A subset of a population used for observation or analysis.

Variable: A characteristic measured for each observation.

Frequency: The number of observations in a category or interval.

Relative frequency: A frequency divided by the total number of observations.

Mean: The arithmetic average of quantitative observations.

Median: The middle value after observations are ordered.

Mode: The most frequently occurring value or category.

Range: The maximum value minus the minimum value.

Quartile: A positional value dividing ordered data into four parts.

Percentile: A positional value below which a specified percentage of observations fall.

Interquartile range: The difference between the third and first quartiles.

Variance: The average squared deviation, using an appropriate population or sample denominator.

Standard deviation: The square root of variance, expressed in original units.

Coefficient of variation: Standard deviation divided by a positive mean.

Z-score: A standardized distance from a reference mean in standard-deviation units.

Skewness: Asymmetry in a distribution.

Correlation: A standardized measure of linear association between two quantitative variables.

Dashboard: A coordinated display of metrics supporting recurring decisions.

Outlier: An observation unusually distant from the main body of data.

Review Questions

Short Answer Questions

1. Distinguish a population from a sample and explain why the distinction matters.
2. When is the median preferable to the mean?
3. What does the interquartile range measure?
4. Why does correlation not establish causation?
5. State two requirements of an effective business table.

Long Answer Questions

1. Explain the differences among mean, median, mode, standard deviation, and interquartile range, including situations in which each is useful.
2. Describe how you would construct and interpret a frequency distribution for international delivery times.
3. Explain how skewness, outliers, and sample size affect descriptive summaries.
4. Compare bar charts, line charts, histograms, box plots, scatterplots, and maps by purpose and limitation.
5. Discuss the design principles of an effective international-business dashboard.

Analytical / Application Questions

1. A company reports average order value of 120 currency units, but the median is 72 and the 90th percentile is 260. What does this suggest about the distribution, and which additional measure would you report?
2. Region A has a mean delivery time of 6 days and 80% delivered within seven days. Region B has a mean of 5.5 days and 68% within seven days. Which region appears better, and what further statistics would you request before deciding?
3. A manager claims that a new advertising campaign caused sales to rise because advertising expenditure and sales have a correlation of 0.82. Identify at least three alternative explanations or analyses needed before accepting the claim.

Multiple Choice Questions with Answers

Question 1

Which measure is generally most resistant to an extreme value?

A. Mean

B. Median

C. Variance

D. Range

Correct Answer: B. Median

Question 2

A dataset contains 40 observations, of which 10 belong to a particular category. What is the category’s relative frequency?

A. 4%

B. 10%

C. 25%

D. 40%

Correct Answer: C. 25%

Question 3

For the sample values 2, 3, and 7, what is the sample standard deviation approximately?

A. 1.53

B. 2.00

C. 2.65

D. 7.00

Correct Answer: C. 2.65

Question 4

Which chart is most appropriate for examining the relationship between advertising expenditure and sales revenue?

A. Scatterplot

B. Pie chart

C. Histogram

D. Single-series bar chart

Correct Answer: A. Scatterplot

Question 5

The interquartile range is calculated as:

A. Maximum minus minimum

B. Q₃ minus Q₁

C. Mean divided by standard deviation

D. Median divided by sample size

Correct Answer: B. Q₃ minus Q₁

Question 6

A z-score of −2 indicates that an observation is:

A. Two units below zero

B. Two standard deviations below the reference mean

C. Two percent below the mean

D. At the second percentile in every distribution

Correct Answer: B. Two standard deviations below the reference mean

Question 7

Which statement about Pearson correlation is correct?

A. It proves that one variable causes another.

B. It can only take positive values.

C. It measures linear association and lies between −1 and 1.

D. It is measured in the units of both variables.

Correct Answer: C. It measures linear association and lies between −1 and 1.

Question 8

Why can a map of total sales be misleading when comparing countries?

A. Maps cannot display quantitative data.

B. Large geographic areas may appear important even when sales are low relative to population or customers.

C. Countries cannot be compared using totals.

D. Maps always exaggerate small countries.

Correct Answer: B. Large geographic areas may appear important even when sales are low relative to population or customers.

Question 9

Which practice most directly improves the honesty of a bar chart?

A. Adding three-dimensional effects

B. Removing axis labels

C. Using a clear baseline and labeling units

D. Using as many colors as possible

Correct Answer: C. Using a clear baseline and labeling units

Question 10

A coefficient of variation is most useful for comparing:

A. Nominal categories only

B. Relative variability across positive-mean variables with different scales

C. Causal effects across experiments

D. The number of missing observations

Correct Answer: B. Relative variability across positive-mean variables with different scales