AFAnalyst Field Notes
Analyst Career

How to Package a Data Analyst Portfolio Project

A portfolio project should feel like an analyst deliverable, not a homework notebook. Package the question, workflow, evidence, and recommendation clearly.

By Analyst Field Notes6 min read

Start with a business question

Good projects answer questions such as why conversion dropped, which segment retains better, or what drives repeat purchase. Avoid projects that only show charts without a decision.

A strong portfolio project sounds like work an analyst would actually do. "I built a dashboard" is weaker than "I analyzed which customer segment drove the decline in repeat purchase and recommended where to investigate next."

Explain the dataset

Describe the source, fields, time range, cleaning steps, and limitations. Interviewers need to know what the data can support.

Include a short data dictionary. List the key columns, data types, and definitions. If you removed rows or changed values, document the rule and why it was necessary.

Show your workflow

Include SQL, spreadsheet steps, Python notebooks, or dashboard logic. The reader should understand how raw data became insight.

Project workflow:
1. Define the business question.
2. Clean and validate the dataset.
3. Calculate core metrics.
4. Segment the result.
5. Visualize the finding.
6. Recommend the next action.

Use visuals sparingly

Charts should support the conclusion. A small number of well-labeled visuals is better than a dashboard full of disconnected widgets.

Use a line chart for trends, a bar chart for segment comparison, and a table when exact values matter. Do not use a chart just because it looks impressive.

End with a recommendation

Summarize the finding, explain confidence and caveats, then recommend the next action or experiment.

Portfolio project template

  • Title: one sentence describing the business problem.
  • Question: what decision the analysis supports.
  • Data: source, fields, date range, limitations.
  • Method: SQL, Excel, Python, or BI workflow.
  • Finding: the main pattern and supporting evidence.
  • Recommendation: what the business should do next.

Common mistakes

  • Using public datasets without a business question.
  • Showing every chart instead of the few that support the decision.
  • Hiding the SQL or analysis steps.
  • Ending with observations instead of recommendations.