Mean, Median, Variance & SD — Compute and interpret mean, median, variance, and standard deviation - Business Analytics

SECTION 1

Historical Context & Motivation

Long before spreadsheets and dashboards, merchants, astronomers, and actuaries needed ways to distill large collections of numbers into a single representative value. The challenge was universal: given a set of observations — crop yields across provinces, navigational readings at sea, or premium payments for life insurance — how does one produce a summary statistic that faithfully captures what is typical and how spread out the data are? The pursuit of that answer drove centuries of mathematical innovation, ultimately producing the four measures that every business analyst reaches for today: the mean, median, variance, and standard deviation.

1669 • Christiaan Huygens & the Expected Value

Huygens formalized the idea of an expected value in probability, laying the intellectual groundwork for the arithmetic mean as a summary of repeated observations.

1756 • Roger Joseph Boscovich & Median-like Estimates

Boscovich proposed using a middle value to reduce the influence of outlying astronomical observations, anticipating the modern median.

1809 • Gauss & Least Squares

Carl Friedrich Gauss demonstrated that minimizing the sum of squared deviations yields the arithmetic mean, introducing the least-squares principle that underpins variance and standard deviation.

1893 • Karl Pearson Coins 'Standard Deviation'

Pearson introduced the term standard deviation and promoted its use in biometrics, giving analysts a dispersion measure expressed in the same units as the data.

1920s– • Statistical Quality Control in Business

Walter Shewhart at Bell Labs applied the mean and standard deviation to manufacturing via control charts, embedding these measures permanently into business operations and analytics.

Today, whether you are evaluating quarterly sales performance, assessing portfolio risk, or monitoring customer satisfaction scores, you need measures that answer two distinct questions: Where is the center of this data? and How spread out are the values around that center? The mean and median address the first question; variance and standard deviation address the second. Understanding when and why to use each measure is fundamental to descriptive analytics in any business context.

SECTION 2

Core Principles & Definitions

Descriptive statistics can be organized around two complementary goals: identifying a measure of central tendency (a single value that best represents the data set) and quantifying dispersion (the degree to which individual observations deviate from that center). Together, center and spread provide a compact numerical portrait of any business data set — one that guides forecasting, benchmarking, and risk assessment.

Arithmetic Mean (x̄)

The sum of all observations divided by the count. Sensitive to every data point, including outliers. Best used when data are roughly symmetric.

Median

The middle value when observations are sorted. Exactly half the data lie above and half below. Robust to outliers, making it ideal for skewed distributions like income or housing prices.

Variance (σ² or s²)

The average of squared deviations from the mean. Captures total dispersion but is expressed in squared units, which complicates direct interpretation.

Standard Deviation (σ or s)

The square root of variance. Returns the spread to the original units of measurement, making it the most widely reported dispersion statistic in business reports.

✦

KEY TAKEAWAY

Think of central tendency and dispersion as two readings on a GPS. The mean or median tells you where the pin drops — the location. The standard deviation tells you the radius of the blue confidence circle around that pin — how much uncertainty surrounds the location. A business decision based on the pin alone, without knowing the circle's size, is a decision made with incomplete information.

SECTION 3

Visual Explanation — Center & Spread on a Distribution

In a symmetric distribution the mean and median coincide at the center (purple dashed line). The standard deviation (gold brackets) marks the interval containing roughly 68 % of observations. A wider bracket signals greater variability in the underlying business data.

The diagram above depicts an idealized symmetric (bell-shaped) distribution. Notice that the mean and median sit at exactly the same point along the horizontal axis. In practice, business data — revenue distributions, response times, defect rates — are often skewed, meaning the mean gets pulled toward extreme values while the median remains anchored at the 50th percentile. This divergence between the two measures is itself informative: a large gap between the mean and median signals the presence of outliers or asymmetry in your data set, a critical diagnostic when preparing a dashboard or executive summary.

Dispersion adds the second dimension. Two product lines might each average $500 in weekly sales, yet one fluctuates between $490 and $510 while the other swings from $200 to $800. The standard deviation captures this difference quantitatively. When a manager asks, "How consistent are our results?" they are asking about the standard deviation — even if they do not use the term.

SECTION 4

Mathematical Framework

The formulas below apply to a sample of n observations, which is the typical scenario in business analytics (you rarely observe the entire population). Where relevant, the population counterpart is noted.

SAMPLE MEAN

x̄ = (1/n) × Σ xᵢ (i = 1, 2, …, n)

x̄ = sample mean; xᵢ = the i-th observation; n = number of observations. The mean minimizes the sum of squared deviations, making it the least-squares estimator of central location.

MEDIAN

Median = x₍₍ₙ₊₁₎/₂₎ if n is odd; (x₍ₙ/₂₎ + x₍ₙ/₂₊₁₎) / 2 if n is even

Sort the data in ascending order. If n is odd, the median is the single middle value. If n is even, it is the average of the two middle values. The median minimizes the sum of absolute deviations.

SAMPLE VARIANCE

s² = (1/(n − 1)) × Σ (xᵢ − x̄)²

We divide by (n − 1) rather than n to obtain an unbiased estimator of the population variance σ². The quantity (n − 1) is known as the degrees of freedom. Squaring the deviations ensures negative and positive departures do not cancel each other out.

SAMPLE STANDARD DEVIATION

s = √s² = √[(1/(n − 1)) × Σ (xᵢ − x̄)²]

Taking the square root converts variance back into the original units of measurement. If sales are measured in dollars, s is also in dollars, making it directly interpretable on the same scale.

ℹ

Population vs. Sample

When you have the entire population (e.g., every transaction in your database for a closed fiscal year), use N in the denominator for variance: σ² = (1/N) × Σ(xᵢ − μ)². When working with a sample drawn from a larger population, use (n − 1) to correct for the downward bias that arises because the sample mean is estimated from the same data.

SECTION 5

Detailed Breakdown — How Skewness Affects Your Choice of Measure

In business analytics, the shape of a distribution determines which summary measure is most informative. Salary data, home prices, and customer lifetime values tend to be right-skewed — a handful of extreme values on the high end pull the mean above the median. Conversely, metrics like test scores or process cycle times can be left-skewed when most values cluster near a ceiling. Recognizing skewness before selecting a summary statistic prevents misleading conclusions.

In a left-skewed distribution, the mean is pulled below the median. In a right-skewed distribution, the mean is pulled above the median. A symmetric distribution keeps the two measures equal. Business analysts use this relationship as a quick diagnostic for skewness.

Distribution Shape	Preferred Center Measure	Business Example
Symmetric / roughly normal	Mean	Standardized test scores; daily output in a stable process
Right-skewed (long right tail)	Median	Employee salaries; home sale prices; customer lifetime value
Left-skewed (long left tail)	Median	Time-to-failure for reliable components; exam scores near a ceiling
Contains significant outliers	Median	Revenue per customer (few "whale" accounts); insurance claims

Choosing between mean and median based on distribution shape

SECTION 6

Worked Example — Quarterly Sales Analysis

A regional manager collects monthly revenue (in $thousands) from seven retail locations for Q3: 42, 38, 47, 35, 90, 44, 40. Compute the mean, median, variance, and standard deviation, and interpret the results.

Computing Mean, Median, Variance & SD

1
Step 1 — Compute the Mean

Sum all values: 42 + 38 + 47 + 35 + 90 + 44 + 40 = 336. Divide by n = 7: x̄ = 336 / 7 = 48.0. The average revenue across the seven locations is $48.0 K.

x̄ = 48.0 ($K)
2
Step 2 — Find the Median

Sort the data in ascending order: 35, 38, 40, 42, 44, 47, 90. With n = 7 (odd), the median is the 4th value.

Median = 42.0 ($K)
3
Step 3 — Compute Deviations and Squared Deviations

Subtract the mean from each observation and square: (42 − 48)² = 36; (38 − 48)² = 100; (47 − 48)² = 1; (35 − 48)² = 169; (90 − 48)² = 1764; (44 − 48)² = 16; (40 − 48)² = 64. Sum of squared deviations = 36 + 100 + 1 + 169 + 1764 + 16 + 64 = 2150.

Σ(xᵢ − x̄)² = 2150
4
Step 4 — Compute Sample Variance

Divide the sum of squared deviations by (n − 1) = 6: s² = 2150 / 6 ≈ 358.33. The variance is in squared units ($K²), which is difficult to interpret directly.

s² ≈ 358.33 ($K²)
5
Step 5 — Compute Standard Deviation

Take the square root: s = √358.33 ≈ 18.93. Revenue across locations typically deviates about $18.93 K from the mean.

s ≈ 18.93 ($K)
6
Step 6 — Interpret the Results

The mean ($48.0 K) exceeds the median ($42.0 K) by $6 K, signaling a right-skewed distribution — the $90 K location is pulling the mean upward. For a managerial summary, the median ($42 K) more accurately represents the typical store's performance. The standard deviation of ≈ $18.93 K is large relative to the mean (a coefficient of variation of about 39 %), indicating high variability; the manager may want to investigate why one store is such an outlier.

Median preferred; high dispersion driven by one outlier

SECTION 7

Strengths, Limitations & Comparisons

No single statistic is universally superior. Each measure of center and spread carries trade-offs that business analysts should weigh when choosing how to summarize and present data. The table below contrasts the four measures across several practical dimensions.

Measure	Strengths	Limitations
Mean	Uses every data point; algebraically tractable; foundation for regression, ANOVA, and most inferential tests.	Highly sensitive to outliers and skew; can misrepresent the typical case when the distribution is asymmetric.
Median	Robust to outliers and skew; always represents the 50th percentile; easy to explain to non-technical stakeholders.	Ignores the magnitude of extreme values; harder to use in algebraic derivations; less efficient for symmetric data.
Variance	Mathematically convenient (additive for independent variables); central to portfolio theory and ANOVA decomposition.	Expressed in squared units, making direct interpretation unintuitive; amplifies the influence of outliers even more than the mean.
Standard Deviation	Same units as the data; interpretable via the empirical rule (68-95-99.7 %); widely understood in business reports.	Still sensitive to outliers; can be misleading for non-normal data where the empirical rule does not apply.

Comparison of mean, median, variance, and standard deviation

✦

KEY TAKEAWAY

Think of the mean and variance as a matched set designed for well-behaved (roughly symmetric, outlier-free) data — they power most classical statistical tests. The median and interquartile range (IQR) are the robust counterparts, suitable when data are messy or skewed. Skilled analysts often report both pairs and let the gap between the mean and median itself serve as a skewness diagnostic.

SECTION 8

Connection to Advanced Analytics

The four descriptive measures introduced in this lesson are not endpoints — they are building blocks for more sophisticated analyses. Understanding how they link to advanced topics helps you appreciate why mastering them first is essential.

Descriptive Concept	Advanced Extension	Business Application
Sample mean (x̄)	Confidence intervals & hypothesis tests — the sampling distribution of x̄ is approximately normal (Central Limit Theorem).	A/B testing to compare two campaign conversion rates.
Median	Non-parametric tests (Mann–Whitney U, Kruskal–Wallis) that do not assume normality.	Comparing customer satisfaction across regions when data are ordinal or skewed.
Variance (s²)	Portfolio variance & covariance — Markowitz mean-variance optimization relies on the variance-covariance matrix.	Constructing the efficient frontier for an investment portfolio.
Standard deviation (s)	Six Sigma & process capability — specification limits are defined in multiples of σ.	Determining whether a manufacturing process meets quality standards (Cpk index).

How descriptive measures feed into advanced analytics

In subsequent courses you will encounter the coefficient of variation (CV) — defined as s / x̄ — which normalizes the standard deviation to a percentage scale, enabling comparison of variability across data sets with different units or magnitudes. You will also learn how the standard deviation anchors the z-score transformation (z = (x − x̄) / s), a tool that converts any observation into a standardized metric indicating how many standard deviations it lies from the mean. These extensions depend entirely on a solid conceptual and computational command of the measures covered in this lesson.

SECTION 9

Practice Problems

Question 1: CONCEPTUAL

A company reports that the mean salary of its 200 employees is $72,000 while the median salary is $55,000. What does the gap between these two figures tell you about the shape of the salary distribution? Which measure would you recommend a job candidate use to gauge a "typical" salary, and why?

Show answer

The mean ($72,000) is substantially higher than the median ($55,000), which indicates the salary distribution is right-skewed. A small number of highly paid executives are pulling the mean upward. A job candidate should rely on the median ($55,000) as the better representation of a typical salary, because it is not inflated by extreme values. The median guarantees that half of employees earn more and half earn less, providing a realistic expectation for a new hire who is unlikely to be in the executive tier.

Question 2: BASIC CALCULATION

A startup tracks daily website visits over five days: 120, 135, 128, 140, 117. Compute the sample mean and sample standard deviation. Show your work.

Show answer

Sum = 120 + 135 + 128 + 140 + 117 = 640. Mean x̄ = 640 / 5 = 128.0. Squared deviations: (120 − 128)² = 64, (135 − 128)² = 49, (128 − 128)² = 0, (140 − 128)² = 144, (117 − 128)² = 121. Sum = 378. Sample variance s² = 378 / (5 − 1) = 94.5. Standard deviation s = √94.5 ≈ 9.72 visits. The site averages 128 daily visits with a typical day-to-day fluctuation of about ±9.7 visits.

Question 3: INTERMEDIATE

Two investment funds report the following annual returns (%) over four years. Fund A: 8, 10, 12, 6. Fund B: 2, 18, 3, 13. Both have the same mean return. Compute the sample standard deviation for each fund and explain which fund carries more risk.

Show answer

Both funds have a mean of (8 + 10 + 12 + 6)/4 = 9.0 % and (2 + 18 + 3 + 13)/4 = 9.0 %. For Fund A: squared deviations = 1, 1, 9, 9 → sum = 20, s² = 20/3 ≈ 6.67, s ≈ 2.58 %. For Fund B: squared deviations = 49, 81, 36, 16 → sum = 182, s² = 182/3 ≈ 60.67, s ≈ 7.79 %. Fund B's standard deviation is roughly three times that of Fund A, meaning its returns are far more volatile. A risk-averse investor would prefer Fund A because it delivers the same average return with significantly less variability — a direct application of the mean-variance framework used in portfolio theory.

Question 4: APPLIED

A retail chain measures checkout wait times (in minutes) at 10 stores: 3.2, 4.1, 3.8, 15.6, 3.5, 4.0, 3.9, 3.7, 4.2, 3.6. Management wants a single number for the weekly report. Compute the mean and median. Which should the operations VP present to the board, and what action does the data suggest?

Show answer

Sorted data: 3.2, 3.5, 3.6, 3.7, 3.8, 3.9, 4.0, 4.1, 4.2, 15.6. Mean = 49.6 / 10 = 4.96 min. Median = (3.8 + 3.9) / 2 = 3.85 min. The mean (4.96 min) is inflated by the single outlier at 15.6 minutes, while the median (3.85 min) reflects the experience of a typical customer at a typical store. The VP should present the median of 3.85 minutes as the representative wait time, but flag the outlier store explicitly. This store likely has an operational issue — perhaps understaffing or a POS system bottleneck — that warrants targeted investigation. Reporting the mean alone would overstate the chain-wide problem and obscure the fact that nine out of ten stores are performing consistently well.

Question 5: CRITICAL THINKING

A financial analyst calculates that Portfolio X has a mean monthly return of 1.2 % with a standard deviation of 4.0 %, while Portfolio Y has a mean monthly return of 1.0 % with a standard deviation of 1.5 %. A colleague argues that Portfolio X is strictly better because its mean return is higher. Construct a rigorous counter-argument using the concepts from this lesson, and propose a single metric that integrates both center and spread to make a fairer comparison.

Show answer

The colleague's argument ignores dispersion entirely. Portfolio X's standard deviation (4.0 %) is more than 2.5 times that of Portfolio Y (1.5 %), meaning its returns fluctuate far more dramatically from month to month. In a bad month, X could easily produce a negative return of −2.8 % (one SD below the mean), while Y's comparable downside is only −0.5 %. To integrate both dimensions, we can compute the coefficient of variation (CV = s / x̄): CV(X) = 4.0 / 1.2 ≈ 3.33; CV(Y) = 1.5 / 1.0 = 1.50. A lower CV indicates a better return-to-risk trade-off. Alternatively, one could compute the Sharpe ratio (excess return per unit of risk), which formalizes this comparison. By either metric, Portfolio Y delivers more consistent value per unit of risk, making it the superior choice for a risk-conscious investor. This problem illustrates why reporting a mean without its accompanying standard deviation can lead to fundamentally flawed business decisions.

SUMMARY

Lesson Summary

Summary

This lesson introduced the four foundational descriptive statistics. The arithmetic mean (x̄ = Σxᵢ / n) provides the least-squares center of a data set and is ideal for symmetric distributions, whereas the median — the middle value of sorted data — is robust to outliers and better represents the typical case in skewed distributions such as salaries or home prices. The gap between the mean and median serves as a practical skewness diagnostic that every analyst should check before choosing which center measure to report.

For dispersion, variance (s² = Σ(xᵢ − x̄)² / (n − 1)) quantifies total spread in squared units and feeds directly into portfolio theory and ANOVA, while the standard deviation (s = √s²) returns the measure to original units, enabling the empirical rule (≈ 68 % of data within ±1 SD) and the z-score transformation. Together, a measure of center and a measure of spread form the minimum descriptive profile that should accompany any business data summary — never report one without the other.