Data Analytics Portfolio Projects for Freshers

A good analytics portfolio shows decisions and data quality, not just charts. Pick one of the three briefs below, use synthetic records or datasets whose reuse licence you have checked, and deliver a metric dictionary, an analysis and a short memo. These are practice briefs, not real client assignments.

How should you build a portfolio?

Each project needs a question, clean definitions, a calculation you can explain and checks you can repeat. Two or three finished projects beat ten half-finished ones. The skills behind them are in the data analytics track, and the query side is covered in SQL skills for a data analyst.

1. Library circulation report

Question it answers: which book categories have rising demand, and where are overdue loans concentrated?

Data to use: fictional loans with book category, checkout date, due date and return date.

Steps: 1. Clean the dates and separate open loans from returned ones. 2. Calculate circulation by period and category. 3. Calculate overdue volumes with a stated rule. 4. Handle renewed loans the same way every time.

Finished output: a workbook or SQL analysis, a dashboard and a metric dictionary. Check that open loans are not counted as returned on time. Raw loan counts may favour categories with more copies, so propose an availability measure before recommending purchases.

2. Subscription revenue reconciliation

Question it answers: does money billed match money collected, and which records do not match?

Data to use: synthetic monthly invoices, payments and credits.

Steps: 1. Define billed revenue separately from collected cash. 2. Join invoices to payments and credits with SQL or spreadsheet lookups. 3. List unmatched records. 4. Build an ageing view of unpaid amounts.

Finished output: a reconciliation table, an ageing report and an exception summary. Test a partial payment, a duplicate payment reference and a credit applied in a later month. Validate a small sample by hand. In your explanation, show why joining every invoice line to every payment can multiply values, and state your aggregation grain.

3. Student-club attendance report

Question it answers: which event formats bring people back, measured against sensible denominators?

Data to use: synthetic event registrations and check-ins with fictional identities only.

Steps: 1. Define registered people, attended people and repeat attendance. 2. Compare formats and periods using rates, not raw counts. 3. Test duplicate check-ins, cancelled events and missing registrations.

Finished output: a compact dashboard and a findings memo. Discuss selection bias, because people who choose one format may differ from people who choose another. Do not claim a format caused higher participation from an observational comparison.

What are the common mistakes?

• Using a dataset without checking its licence.
• Reporting counts where a rate is needed.
• Leaving definitions only in your head.
• Treating a screenshot as the whole project.
• Ending without any recommendation or stated limit.

How do you check a project is ready to share?

Ask someone to reproduce one headline number from your files alone. If they can, your definitions and steps are clear. Also check that every synthetic result is labelled as synthetic and that no real personal data is present.

How do you package it?

Provide the question, source conditions, cleaning steps, calculations and quality checks. Include a screenshot only as a preview, and make the underlying work understandable too.

Finish with one recommendation and the evidence still needed before acting. A report that admits a limitation can be more useful than one that offers an unjustified answer.

Worked example: reconciliation edge cases

For project 2, build a tiny set of three invoices and four payments. Make one invoice partly paid, one paid twice with the same reference, and one paid by a credit next month. Write the correct reconciliation by hand first. Then run your SQL or spreadsheet logic and compare. If your totals differ, the difference tells you which rule is missing. Keep this small test inside the project as evidence of your checking.

What should you build next?

Turn your best project into a dashboard using how to build a business dashboard. If you want feedback on your portfolio plan, see GullyMentor pricing for how mentorship works.

Sources & further reading

• PostgreSQL SQL tutorial

Examples and practice recommendations are educational illustrations. Tool documentation and programme inclusions should be checked for your intended use.