Data Analyst Program

Guided Project · Week 1 of 4 · 5 hours

Dataset Selection, Problem Definition and Data Collection

Pick a dataset, define the problem you will solve, and gather the data.

Learning objectives

By the end of this week you will be able to:

  • Evaluate a dataset for size, quality, documentation and license.
  • Write a problem statement and three to five SMART questions.
  • Identify who would use the answer (the stakeholder).
  • Collect and store the raw data in an organized project folder.

Lessons

Dataset Selection

A good project dataset is interesting to you, big enough to analyze (aim for at least 1,000 rows and several columns), has a mix of numbers, categories and dates, and is free to use.

Where to look:

  • Kaggle Datasets: thousands of free datasets; check the "Usability" score and read the discussion tab.
  • Government open data: data.gov (US), data.gov.uk, city open data portals.
  • Google Dataset Search: a search engine for datasets.
  • Organizations: World Bank Open Data, Our World in Data, the UCI Machine Learning Repository.

Checklist before you commit: Is there a data dictionary explaining each column? When was it last updated? What license does it have (most open licenses allow portfolio use with credit)? Does it contain personal information that should not be shared? Avoid very famous beginner datasets (Titanic, Iris) for your main project; employers have seen them hundreds of times.

Problem Definition

Analysis without a question is just browsing. Start by writing a problem statement: who has the problem, what it is and why it matters.

Example: "A bike-share company's casual riders spend less per year than members. The marketing team wants to know how casual riders use bikes differently so it can design a campaign to convert them."

Then write SMART questions, questions that are Specific, Measurable, Action-oriented, Relevant and Time-bound:

  • Weak: "How do people use bikes?"
  • SMART: "In 2025, how did average ride length and peak riding days differ between casual riders and members?"

Also name the stakeholder (who will read your results) and the decision your analysis supports. If you cannot name a decision, sharpen the question.

Data Collection

Collecting data is more than clicking Download. Do it in an organized way from the start:

  • Create a project folder with subfolders: data/raw, data/clean, sql, notebooks, dashboard, docs.
  • Save the original files in data/raw and never edit them.
  • Write down the source URL, download date and license in a sources.md file.
  • Look at the data dictionary and note what each column means and its unit.
  • Load the data into your SQL database and into a Python notebook, and confirm the row counts match the source.

If your data comes in several files (for example one per month), combine them now with Power Query, a SQL UNION ALL or pandas pd.concat.

Watch

Practice

Shortlist three datasets and score each on interest, size, quality, documentation and license. Choose one. Write your problem statement, stakeholder, decision and three to five SMART questions. Set up the project folder and load the raw data.

Check your understanding

Try each question first, then open it to see the answer.

Why should you never edit the raw data files?

So you can always return to the original and prove how the clean data was produced.

Rewrite "What about sales?" as a SMART question.

For example: "Which three product categories grew fastest in revenue between Q1 and Q4 2025?"

Name two things to check before choosing a dataset.

Any two of: size, documentation, update date, license, personal information, data quality.

Assignment

Checkpoint 1: submit a one-page project proposal (problem statement, stakeholder, decision, SMART questions, dataset source and license) and a screenshot of your project folder.