Data Analyst Portfolio: A Reproducible Project Example
Build a data analyst portfolio around a clear question, reproducible analysis, and a useful finding. Explore Edgar Owen's page and download the worked example.

Javier Martínez Samblas
Founder of Self · AI-powered resume-to-website platform

A data analyst portfolio should make your reasoning easy to follow: the question, data checks, method, finding, and next action. When professional datasets are confidential, a focused public or synthetic project gives employers concrete work to inspect.
Self's portfolio website builder turns your CV into a professional page where each analysis has a clear project description and links to its technical evidence. Edgar Owen's portfolio combines SQL and Python work with operational reporting experience.
A finished data analyst portfolio: Edgar Owen

Explore Edgar Owen's website · Download Edgar's CV text (English)
Choose a question you can answer publicly
Edgar's worked project asks: Which support queue should a lead investigate first? It uses purpose-built synthetic tickets to demonstrate a familiar operational decision. The project includes the data, definitions, queries, notebook, and an output a reader can reproduce.
Choose a similarly narrow question around backlog, stock availability, recurring errors, or service demand. A manageable dataset and one useful decision give the case a clear shape.
Show the evidence behind the analysis
Question
What Edgar shows
Data
What Edgar shows
Quality checks
What Edgar shows
Method
What Edgar shows
Finding
What Edgar shows
Next action
What Edgar shows
| Layer | What Edgar shows |
|---|---|
| Question | Which queue needs the first backlog investigation? |
| Data | 624 raw rows representing 600 unique tickets across eight weeks and three queues. |
| Quality checks | Exact duplicates, timestamp order, queue labels, and status/resolution agreement. |
| Method | Deduplicate, fix the reporting snapshot, count open tickets, and group their elapsed age. |
| Finding | 96 tickets remain open; 58 are in onboarding, including 32 aged at least 15 days. |
| Next action | Inspect older onboarding tickets and compare incoming with resolved volume. |
Count distinct ticket IDs before comparing queues. The 624 source rows include 24 duplicate copies; counting raw rows would count some tickets more than once.
Download and reproduce the worked project
Download the complete analysis kit (ZIP, English) · Open the CSV dataset · Open the SQLite query · Open the Python notebook
- Extract the ZIP. It includes the CSV, analysis script, SQL query, notebook, metric definitions, expected results, and chart.
- Run python analyze.py with Python 3.10 or later, or run the notebook cells. The analysis uses the standard library.
- Compare your output with results.json. Both SQL and Python produce the same queue and age-band counts.
Use a fixed snapshot: 29 June 2026 at 00:00 UTC. Count open tickets separately from resolved-ticket durations. The example measures elapsed hours rather than business hours, and documents the denominator for each duration summary.
Onboarding
Open tickets
Aged 15+ days
Billing
Open tickets
Aged 15+ days
Technical support
Open tickets
Aged 15+ days
| Queue | Open tickets | Aged 15+ days |
|---|---|---|
| Onboarding | 58 | 32 |
| Billing | 20 | 6 |
| Technical support | 18 | 4 |
Explain the finding and the next action
A concise case conclusion could say: After removing 24 duplicate rows, I analysed 600 tickets at a fixed snapshot. Onboarding held 58 of 96 open tickets and 32 of the 42 tickets aged at least 15 days. I recommended reviewing that queue's oldest work and comparing arrival with resolution volume before deciding on capacity changes.
These are findings from the synthetic example. Keep the next investigation distinct from an implemented business outcome. For a professional project, describe the decision you supported and your own contribution at the level you can publish.
Keep one publication checklist: use data you own or can publish, label public or synthetic sources, remove internal identifiers and credentials, and obtain permission for employer material. A rebuilt case can demonstrate the method with its own records and outputs.
Give each technical link a clear job
Use the website for the question, method, selected visual, finding, and next action. Link the notebook or repository beside the case so someone can inspect calculations without having to discover the story inside a code folder.
Edgar's second project adds a different capability: aligning equipment status definitions across six depots and replacing three reporting spreadsheets with one weekly inventory view. Its recreated schema and example records explain the reporting design.
Question
Prompt
Data and checks
Prompt
Method
Prompt
Output
Prompt
Action
Prompt
Technical evidence
Prompt
| Part | Prompt |
|---|---|
| Question | Who needs to decide what? |
| Data and checks | What is one row, and what did you validate? |
| Method | Which calculation answers the question, and why? |
| Output | Which chart or table makes the finding clear? |
| Action | What should happen next, and what remains to investigate? |
| Technical evidence | Where can the reader reproduce the work? |
Create your data analyst portfolio with Self
Upload your CV, choose Cybercore, and customize your introduction and project cases in Studio. Add a chart as project media and link the notebook or repository from the project. Your experience and skills give the analysis a professional frame.
Explore Edgar Owen's website · Create my data analyst portfolio with Cybercore
Cybercore publishing is included with Pro and Team. A Free page can use a Free template such as Minimalistic and is accessible by direct link without search indexing; Pro and Team add custom domains and search visibility controls.
Data analyst portfolio questions
Two or three well-explained cases are a useful start. Give each a distinct purpose, such as SQL analysis, reporting design, or experimentation, with enough detail to discuss the decisions.
Use one when interaction helps answer the question. A clear chart, reproducible query, and concise explanation can already show analytical skill. Choose the format around the work you want to demonstrate.
Show what you completed: a checked dataset, a reproducible calculation, a documented finding, or a recommendation. Those outputs let the reader assess the work before any operational change is tested.