|
Video | Introduction | The Five Steps | 1. Define | 2. Collect | 3. Clean | 4. Analyze | 5. Share | Tools | Conclusion | Sources |
In this video, a data analytics instructor walks through the five steps of the data analysis process and the tools used at each step.
Introduction
Data analysis is how you find the story in the data.
A chart communicates one finding to one audience. Before a chart can be built, someone has to work out what the data shows. The data analysis process is the sequence of steps that gets from a question to that finding. Organizations use it to answer business questions, evaluate programs, and support decisions.
Last week covered Data Storytelling: choosing a chart that makes the right comparison easy, emphasizing the takeaway, calling out the key number, and presenting data ethically. Those skills assume a finding already exists. This week covers where the finding comes from. The Data Analysis Process allows us to determine a finding, which we use in turn to tell our story.
The Five Steps at a Glance
Each step produces something the next step uses.
| Step | Action | Deliverable |
| 1. Define the Question | State the problem and what you want to find out | A problem statement |
| 2. Collect the Data | Gather the data that can answer the question | A dataset |
| 3. Clean the Data | Fix errors and prepare the data for use | A usable dataset |
| 4. Analyze the Data | Examine the data to find the story | A finding |
| 5. Share the Results | Present the finding with context | A chart with context |
Step 1: Define the Question
Every analysis starts with a question.
The first step is to define the objective. In data analytics, this is the problem statement. Defining the objective means forming a hypothesis and deciding how to test it. A useful starting question is: What problem am I trying to solve?
The question a manager asks may not describe the underlying problem. An analyst needs to understand the organization and its goals well enough to frame the problem. Consider a company that builds custom training software. It wins many new clients, but few of them return. Management asks, “Why are we losing customers?” An analyst might reframe the question as “Which factors are negatively impacting the customer experience?” or “How can we boost customer retention while minimizing costs?” The reframed questions point to data that can answer them.
Business metrics and key performance indicators (KPIs) help define the question. KPIs are measures an organization tracks to see how it is performing. Regular reports and dashboards show where problems appear.
Connection to storytelling. Once we begin to figure out the answer or answers to our problem statement, we can start to see the story we’re going to tell.
Step 2: Collect the Data
The question determines the data.
Once the question is defined, the next step is a plan for collecting the data that can answer it. A finding can only be as accurate as the data behind it. Data that is incomplete, out of date, or measured differently from what the question asks can lead to a finding that looks solid but is wrong, and no amount of cleaning or analysis fixes data that does not fit the question.
Things to check when collecting data.
- Fit. Does the data measure what the question asks about? Sales figures show what customers bought, not why they stopped buying.
- Source. Who collected the data, how, and for what purpose?
- Date. When was the data collected, and does that time period match the question?
- Definitions and units. What does each column measure, and in what units? Two sources can use the same label for different things.
- Coverage. Does the data include the people, places, or time periods the question is about, or only some of them?
- Permission and privacy. Is the data available to use, and does it contain personal information that needs protecting?
Types of data.
- Quantitative data is numeric: sales figures, test scores, counts.
- Qualitative data is descriptive: customer reviews, interview responses, open-ended survey answers.
Sources of data.
| Source | Definition | Examples |
| First-party data | Data an organization collects directly from its own customers or activity | Surveys, interviews, focus groups, observation, transaction records, CRM data |
| Second-party data | Another organization’s first-party data, obtained directly from that organization or through a marketplace | Website or app activity, purchase history, shipping data |
| Third-party data | Data collected and aggregated from many sources by an outside party | Industry reports, market research, open data repositories, government portals |
First-party data is collected for the question at hand and tends to be structured. Second-party data tends to be structured and reliable. Third-party data often includes unstructured or large-volume data.
Students and small teams typically work with first-party data they gather themselves (a survey, for example) or with third-party data from public sources such as government data portals.
Step 3: Clean the Data
Data should be reviewed and prepared before it is used as a part of the analysis process.
Collected data needs to be checked and corrected before analysis. Cleaning (also called scrubbing) matters because errors in the data carry through to the results. Analyzing flawed data can produce a wrong finding and send the work back to the start.
Common cleaning tasks.
- Remove errors, duplicates, and outliers. These show up when data comes from several sources.
- Remove irrelevant observations. Rows that have no bearing on the question are taken out.
- Fix structure and layout. Typos, inconsistent labels, and formatting problems are corrected so the data is easier to sort, group, and chart.
- Fill gaps. Cleaning often reveals missing values. Once identified, they can be filled, flagged, or removed, depending on the question.
Rules and step-by-step procedures for Google Sheets and Excel: Rules for Better Data Cleaning.
Step 4: Analyze the Data
Analysis is where the story is found.
Analysis means examining the cleaned data to answer the question defined in Step 1. The goal is a finding that is specific enough to state in a sentence: what changed, what differs, what stands out.
How analysis works. Analysis typically moves from a broad look at the data to a specific finding.
- Explore. Get a sense of the whole dataset. Sort columns, calculate totals and averages, find the highest and lowest values, and build quick, unpolished charts to see the shape of the data.
- Compare. Break the data into groups and time periods and compare them. Pivot tables summarize data by group. Patterns tend to appear here: one group is larger, a value is rising or falling, a share has grown, two measures move together, or one value stands apart from the rest.
- Test. Check the pattern against the question and hypothesis from Step 1. Does the pattern hold across the whole dataset, or only part of it? Is there another explanation? Two measures that move together do not show that one causes the other.
- State the finding. Write the finding as one sentence. That sentence becomes the takeaway title of the chart in Step 5.
Last week introduced the difference between exploratory analysis, where an analyst looks for findings, and explanatory communication, where a writer hands a finding to a reader. The charts built while exploring are for the analyst. The chart built in Step 5 is for the reader.
The kinds of patterns that turn up during analysis match the comparisons covered in Genre Focus: Choosing the Right Chart, so the finding also points to the chart type.
Four types of analysis.
| Type | Question It Answers | Example |
| Descriptive | What happened? | Monthly visits to the tutoring center over a semester |
| Diagnostic | Why did it happen? | Visits dropped in October. Did exam schedules, staffing, or hours change? |
| Predictive | What is likely to happen? | Using past semesters to estimate next semester’s visits |
| Prescriptive | What should be done? | Recommending staffing levels based on predicted visits |
Descriptive analysis is a common first step. Prescriptive analysis draws on the other three. Microsoft Learn lists a fifth type, cognitive analytics, in its overview of data analysis.
Pivot tables. A pivot table groups rows by one column and totals another, so comparisons that take many sorts appear in one table. Steps: Google Sheets (Step 5) and Excel (Step 5) in Rules for Better Data Cleaning.
Step 5: Share the Results
The finding reaches an audience.
The final step is sharing the insights with stakeholders. Sharing involves more than handing over raw results. It involves interpreting the outcomes and presenting them so the audience can follow them. Analysts typically use reports, dashboards, and interactive visualizations.
How results are interpreted and presented can influence decisions, including whether an organization restructures, launches a product, or closes a division. For that reason:
- Present the evidence gathered, including evidence that does not support the finding. Selecting only supportive data (cherry-picking) misrepresents the analysis.
- State gaps in the data and any insights open to more than one interpretation.
- Write clearly and plainly so the audience reads the result the same way.
Tools at a Glance
Analysts use a range of tools, from spreadsheets to programming languages. The specific tools change from year to year.
Tools fall into a few broad groups. Spreadsheets handle collection, cleaning, analysis, and charts for small and medium datasets. Dedicated software handles specific jobs, such as dashboards, data cleaning, or visualization, often with more capacity or polish than a spreadsheet. Programming languages such as Python, R, and SQL handle large datasets and repeatable work, and they require coding. AI assistants are now built into spreadsheet programs.
The tools below include the ones named in this week’s video, updated where products have been renamed or retired, plus newer tools. This list is a glimpse of what is available, not a requirement. In this course, Excel or Google Sheets covers each step.
Spreadsheets
- Microsoft Excel. Spreadsheet program with cleaning tools, formulas, pivot tables, and charts. Power Query, built into Excel for Windows, records cleaning steps so they can be repeated on new data.
- Google Sheets. Free, browser-based spreadsheet with cleaning tools, formulas, pivot tables, and charts. Files can be shared and edited by several people at once.
Collecting Data
- Google Forms and Microsoft Forms. Survey tools that send responses straight to a spreadsheet.
- data.gov. The U.S. government’s open data portal, with datasets from federal agencies.
- data.census.gov. U.S. Census Bureau data on population, housing, income, and business.
- Data management platforms (DMPs). Enterprise software that gathers and combines data from many sources. Several DMPs named in the video have changed: Salesforce’s DMP (Audience Studio) was retired in 2024 in favor of Salesforce Data Cloud, Xplenty now operates as Integrate.io, and the open-source D:SWARM project was archived in 2022. SAS remains a major enterprise analytics vendor.
Cleaning Data
- OpenRefine. Free, open-source tool that runs on a personal computer. It imports and exports spreadsheet and CSV files and handles larger or messier datasets than spreadsheet tools.
- Data Ladder. Enterprise software for matching and deduplicating records across datasets.
- Python (pandas) and R. Programming languages with libraries for cleaning, reshaping, and analyzing large datasets.
Dashboards and Reporting
- Data Studio. Google’s free tool for building reports and dashboards from Google Sheets and other sources. Google renamed it Looker Studio in 2022 and changed the name back to Data Studio in April 2026.
- Power BI Desktop. Microsoft’s business intelligence tool for building reports and dashboards. The desktop app is a free download for Windows.
- Grafana. Open-source dashboard software, used to monitor data that updates over time.
- Databox and Dasheroo. Commercial dashboard services that pull KPIs from business apps into one view.
Visualizing and Sharing
- Datawrapper. Web-based tool for charts, maps, and tables, used by many newsrooms. The free plan includes unlimited charts, maps, and tables.
- Tableau Public. Free version of Tableau for creating and sharing visualizations online. Visualizations published there are visible to anyone.
- Infogram. Web-based tool for infographics, charts, and maps built from templates. The free version is limited to 10 projects.
- Google Charts. Free charting library for websites. It requires some coding.
- Matplotlib, Seaborn, and Plotly. Python libraries for building charts in code.
AI Assistants
- Copilot in Excel. Creates formulas, charts, and pivot tables, summarizes data, and answers questions about a workbook. Access depends on the Microsoft 365 subscription.
- Gemini in Google Sheets. Creates tables, formulas, charts, and pivot tables and analyzes data. It requires an eligible Google Workspace or Google AI plan. Google notes that Gemini “may suggest inaccurate or inappropriate information.”
Output from an AI assistant is checked against the data before it is used.
Last updated: October 2026.
Conclusion
Analysis finds the story. The other four steps set it up and deliver it.
The data analysis process runs in five steps: define the question, collect the data, clean the data, analyze the data, and share the results. Each step produces something the next step uses: a problem statement, a dataset, a usable dataset, a finding, and a chart with context.
Defining the question gives the story a direction. Cleaning makes sure the story rests on reliable data (see Rules for Better Data Cleaning). Analysis finds the story, and sharing the results delivers it to an audience.
In the in-class activity, your group applies the process to a historical dataset and reports its findings. In the take-home assignment, you confirm the analysis and produce a chart with context.
Sources & Further Reading
Everything referenced above, if you want to go to the originals.
- “A Step-by-Step Guide to the Data Analysis Process” (video), CareerFoundry — embedded at the top of this page. Source of the five-step process and the training-software example.
- Ault, Liao, and Musolino, “1.1 What Is Data Science?” Principles of Data Science (OpenStax, 2025) — problem definition, data collection, data preparation, data analysis, data reporting.
- Wild and Pfannkuch, “Statistical Thinking in Empirical Enquiry,” International Statistical Review 67(3), 1999 — the PPDAC cycle.
- American Statistical Association, GAISE II (2020) — formulating statistical investigative questions.
- Wickham, Çetinkaya-Rundel, and Grolemund, R for Data Science, 2nd ed. — import, tidy, transform, visualize, model, communicate.
- Chapman et al., CRISP-DM 1.0: Step-by-Step Data Mining Guide (2000) — business understanding; time spent on data preparation.
- LiveRamp, “1st-Party Data, 2nd-Party Data, and 3rd-Party Data,” 2019, and Quantcast, “Data: First-Party, Second-Party, Third-Party” — data sources.
- Microsoft Learn, “Data Analysis” — descriptive, diagnostic, predictive, prescriptive, and cognitive analytics.
- Microsoft, “About Power Query in Excel” and “Import and Shape Data in Excel for Mac (Power Query)” (Excel Support).
- OpenRefine — free, open-source data cleaning tool.
- Microsoft, “Get Started with Copilot in Excel” and Google, “Collaborate with Gemini in Google Sheets” — AI assistants.
- Microsoft Learn, “Get Power BI Desktop” — free download, Windows requirements.
- “Data Studio Is Back: Google Kills Looker Studio Name for Good,” PPC Land, April 2026.
- Tableau, “Tableau Public”; Datawrapper, “Create Data Visualization for Free”; ECML, “Infogram” — free plans.
- Hightouch, “Salesforce Audience Studio”; Heroku Dev Center, “Integrate.io (Xplenty)”; D:SWARM documentation (archived 2022) — status of tools named in the video.