9 Basic Data Analysis Workflows

Frame the question and scope

A data analysis workflow turns a question into a supported, understandable answer. It is usually iterative: inspecting data may reveal a problem, and early results may suggest that the original question needs refining. A sound analysis defines the decision or question, checks whether the data are fit for that purpose, and communicates both results and limitations.

Inspect and prepare the data

Write down what you need to find out and what kind of answer would be useful. Make the question specific enough to guide the analysis. For example, replace “How are sales doing?” with “Which product category had the highest total sales revenue this quarter?” This identifies the time period, outcome, and groups to compare.

Also establish what the data represent: all sales or only a sample, and which records are included. These choices define the scope of the answer.

Inspect and prepare the data

Before calculating, examine the data’s structure and meaning. A useful spreadsheet format is a rectangular table with one header row, one record per row, and one variable per column. Use clear, unique column names; avoid merged cells and extra header rows. A sales table might include Date, Category, Region, Units, and Unit Price.

Check the data for several kinds of problems:

• Completeness: Are important values blank or marked “N/A”?

• Consistency: Are dates, category names, and units recorded in consistent formats?

• Validity: Are values plausible, such as nonnegative units and dates within the intended quarter?

• Duplicates: Do any records appear more than once?

Correct errors only when the correction can be justified. Decide how to handle missing or unusual values and record those decisions. Keep an unchanged copy of the original data so changes can be checked. Data quality is about whether the data are fit for a particular purpose, not simply whether they have been cleaned.

Calculate and explore

A Formula is an expression that calculates a result; a Function is a built-in operation used in a formula. If Units is in column D and Unit Price is in column E, a row’s revenue can be calculated with:

Copy the formula down the column. Summary functions can then use SUM to total revenue, AVERAGE to find a mean, or COUNTIF to count records matching a category. Check that formulas refer to the intended rows and columns.

A Filter temporarily displays only records matching a condition, such as sales from one region or quarter. Sorting changes the order of records, for example arranging revenue from highest to lowest. Filtering helps examine a subset, but does not by itself change the underlying question or prove that the subset is representative. Keep track of active filters so subset results are not mistaken for totals over the full dataset.

Summarize the data

Descriptive statistics summarize observed data. Count tells how many observations are included. Mean is the sum divided by the count, or $\frac{\text{sum}}{\text{count}}$. Median is the middle value after the data are sorted. Minimum and maximum show the extremes, and their difference is the Range.

Standard deviation describes spread around the mean, while Interquartile range describes the spread of the middle half of the data. Choose summaries that match the question and inspect the distribution rather than relying on one number. For example, mean revenue per transaction answers a different question from total revenue. A few unusually large transactions can pull the mean upward; comparing the mean with the median can help reveal that pattern.

Compare groups and visualize patterns

A Pivot table groups records by one or more fields and calculates summaries, such as total revenue by category or average revenue by region. To answer which product category had the highest total revenue, place Category in rows and Revenue in values, summarized by sum. Check that the value field is being summed rather than counted, and refresh the pivot table if its source data have changed. Pivot tables also make it useful to explore the same data using different groupings.

A chart can make a pattern easier to see. Use a bar or column chart to compare categories, a line chart to show change over time, and a histogram to examine the distribution of numerical values. Label axes, include units, and avoid visual effects that obscure comparisons. A chart should make the relevant result easier to interpret, not replace checking the underlying numbers.

Check and interpret results

Before drawing a conclusion, verify key totals against the source records. Check whether filters or missing values affected the results, and investigate surprising values. Ask whether the analysis actually answers the stated question. Describe what the data show, but do not claim that one factor caused another unless the analysis design supports a causal conclusion.

Communicate findings and limitations

Present the answer first, then provide the evidence that supports it. A concise report should state:

1. Question and scope: What was examined, and for which period or population?

2. Finding: What is the main result, with an appropriate number or chart?

3. Method: How were records prepared and summarized?

4. Limitations: Were there missing data, unusual values, assumptions, or other reasons for caution?

5. Implication: What may the result mean for the decision at hand?

For example: “Category A had the highest recorded revenue this quarter. Revenue was calculated as units multiplied by unit price and summed by category. The result covers recorded transactions only; missing or duplicated records could affect the comparison.” Clear communication includes relevant uncertainty and data limitations.

Connect the steps

The workflow is to clarify the question, inspect and prepare the data, summarize and explore it, check the results, and communicate the finding with its limitations. Tables, formulas, functions, filters, descriptive statistics, pivot tables, and charts are tools within that workflow. The question determines which tools and summaries are appropriate.