Data analysis is the process of collecting, organising, examining and interpreting data to answer questions and support decisions. It is at work when a shop decides how much of each product to buy, when a hospital plans its shifts or when an app chooses what to show on its home screen.
The four types of data analysis
Descriptive: answers what happened. Example: how much we sold each month last year.
Diagnostic: answers why it happened. Example: sales fell in March because a product ran out of stock.
Predictive: estimates what is likely to happen, using statistics and machine learning. Example: how much we should sell next quarter.
Prescriptive: suggests what to do. Example: how much of each item to buy to avoid both shortages and excess stock.
The steps of a data analysis
1. Define the question. Good analysis starts with a clear question tied to a decision, such as which customers are most likely to cancel.
2. Collect the data. It can come from management systems, spreadsheets, websites, apps, surveys or public sources.
3. Clean and prepare. Fix errors, standardise formats, handle missing values and remove duplicates. This step often takes most of the time.
4. Explore and analyse. Calculate indicators, compare groups, look for trends and test hypotheses.
5. Visualise and communicate. Simple charts and dashboards help show the result to whoever has to decide. The conclusion should fit in a few sentences.
6. Decide and follow up. Analysis only creates value when it leads to action, and the result of that action should be measured afterwards.
The most widely used tools
Spreadsheets such as Excel and Google Sheets remain the entry point and solve many problems. SQL is the language for querying databases and one of the most requested skills in the field. Python and R are used for statistical analysis, automation and machine learning. Visualisation tools such as Power BI, Tableau and Looker Studio turn data into interactive dashboards.
Metrics and careful conclusions
Correlation is not causation: two things that happen together do not mean one causes the other. Averages can hide important differences between groups. Small samples produce fragile conclusions. And a chart can mislead depending on the scale chosen. Good data analysis is as much about rigour as about honesty regarding its limits.
Personal data and privacy
Analysing personal data must follow data protection laws such as the GDPR in Europe and the LGPD in Brazil: have a legal basis for the use, use only what is necessary, protect the information and respect people's rights. Whenever possible, work with aggregated or anonymised data.
How to get started
Begin with spreadsheets and basic statistics, then learn SQL and a visualisation tool. Practise with public datasets, such as those from statistics offices and open data portals, and build a portfolio of questions, analyses and clearly explained conclusions. With that foundation, Python and machine learning are the next step; see our guide to machine learning explained.