Timedora

Descriptive Analytics: Summarizing Observed Data

Learn what descriptive analytics can and cannot tell you, and how classifying variables guides the choice of summaries.

Start with the Data Type

Descriptive analytics: describe, don’t explain

Descriptive analytics organizes and summarizes observed data to help people understand what happened. Tables, statistics, and charts can reveal patterns in the data, but a description alone does not establish why something happened or predict what will happen next.

Before choosing a summary, identify the type of each variable. Categorical data sorts observations into groups, such as sales channel or customer status. Numerical data records quantities with meaningful magnitudes and intervals, such as delivery time or order value. This distinction matters: a method that makes sense for quantities may be meaningless for category labels.

Sales channels
Sales channelOrdersShare
Online4545%
Store3535%
Partner2020%
Total100100%
Sales-channel orders and each channel’s share of all 100 orders.

Match the summary to the variable

For categorical data, a is the count in a category, while a is that count divided by the overall total. In the table, online orders have a of 45 and a of 45/100=0.4545/100=0.45, or 45%. The denominator makes the percentage interpretable. A bar chart is usually useful because separated bars compare category counts or percentages. A pie chart can show parts of a whole, though comparing bar lengths is often easier. For comparisons across groups, use grouped or stacked bars and state what each bar represents.

The mode is the most frequent category. Means and standard deviations do not describe labels such as “online” and “store”—even if someone has assigned the labels numbers—because those numbers do not represent meaningful quantities.

For numerical data, consider both center and spread. For delivery times of 2, 3, 3, 4, and 8 hours, the mean is (2+3+3+4+8)/5=4(2+3+3+4+8)/5=4 hours. The sorted middle value is 3 hours, so the median is 3 hours; the range is 8−2=68-2=6 hours. The high value pulls the mean above the median, so reporting both helps describe the data. For roughly symmetric data without influential outliers, mean and standard deviation are often suitable. For skewed data or data with outliers, median and IQR are often more representative.

The same principle guides charts: histograms show numerical values grouped into adjacent intervals, while bar-chart bars represent distinct categories and are separated. Classifying variables first helps prevent mismatched summaries and keeps a descriptive report focused on what the data support.

  • : The spread of the middle half of the data.