Build evidence, not just charts.
Three project briefs with synthetic datasets, realistic questions, and review criteria. Pick one, make the decisions yourself, and document what you learn.
All three datasets are original, synthetic practice data. They contain no real customers or employer information. Time estimates are planning guides.
Retail performance
A retail manager wants to know which products and regions deserve attention. Revenue alone does not explain performance. Use order-line data to separate sales volume, discounting, cost, and returns.
Questions to answer
- What changed in net revenue and gross margin by month?
- Which category and region combinations explain the change?
- How do discounts and returns affect the story?
Analysis rules
- Define net revenue as quantity × unit price × (1 − discount rate) for non-returned lines. Treat returned lines as zero revenue and zero cost in this simplified exercise.
- Aggregate line-level amounts before calculating margin rate; do not average row-level percentages.
- Build a monthly trend and a category/region drill-down. Explain partial periods and the dataset’s simplified return treatment.
What to deliver
- A documented SQL analysis or Power Query workflow.
- A one-page dashboard with defined KPIs and reconciliation checks.
- A 300-word decision memo with two recommendations and one limitation.
Add one question of your own. Explain why it matters, what the data supports, and what additional information you would need.
Review your work · 100 points
Use the data dictionary’s control totals to validate your import. This rubric supports self-review; it is not a professional credential.
Service operations
An operations lead sees inconsistent service times across teams. Use a fixed reporting snapshot to investigate ticket aging, resolution time, and SLA attainment without confusing open tickets with completed work.
Questions to answer
- Which queues have the oldest open work as of the snapshot?
- How does resolution time vary by priority and category?
- Where is the biggest opportunity to improve SLA attainment?
Analysis rules
- Use 2026-06-30 23:59 UTC as the fixed snapshot; do not use today’s date.
- Separate closed-ticket resolution time from open-ticket age. Both are elapsed hours, not business hours.
- Compare median and 90th-percentile resolution times. Calculate SLA attainment only for closed tickets, and separately flag open tickets already beyond their targets.
What to deliver
- A documented cleaning workflow, including missing resolution timestamps.
- A queue performance report with explicit denominators.
- A short recommendation that considers volume and case complexity.
Add one question of your own. Explain why it matters, what the data supports, and what additional information you would need.
Review your work · 100 points
Use the data dictionary’s control totals to validate your import. This rubric supports self-review; it is not a professional credential.
Subscription retention
A subscription business wants to understand retention across acquisition cohorts. Build customer-month logic from start and cancellation dates, and distinguish cohort retention from monthly churn.
Questions to answer
- How does month-end retention vary by acquisition cohort?
- What is monthly churn among customers active at month start?
- Do plan differences persist when cohort age is comparable?
Analysis rules
- Treat each subscription as one customer. There are no restarts, upgrades, or pauses in this simplified dataset.
- A cancellation date is the first inactive day. A customer is active on a date when start_date ≤ date and cancel_date is null or greater than date.
- For monthly churn, use cancellations in the month among customers active at month start, divided by that month-start population. Exclude customers who started later in the month.
- Only show complete calendar months through 2026-06. Keep unobserved cohort ages blank rather than treating them as zero retention.
What to deliver
- A SQL date spine and documented customer-month table.
- A retention matrix with cohort sizes and unobserved periods marked.
- A findings memo that distinguishes patterns from causal claims.
Add one question of your own. Explain why it matters, what the data supports, and what additional information you would need.
Review your work · 100 points
Use the data dictionary’s control totals to validate your import. This rubric supports self-review; it is not a professional credential.