README.md
Sample learner — SQL & data analysis portfolio
Sample portfolio. Atterna generated this repository from a synthetic learner profile (“Sample learner”, not a real person) on Atterna’s fictional marketplace data. Your own export looks like this, built from your own queries and words.
Work samples from Atterna, a simulator where you learn data analysis by doing a junior analyst’s job. Each project below is work I actually did there: the business question, my SQL, the result and the call I made.
7 projects · 36 SQL queries
The data is fictional. Meridian is a synthetic online marketplace built for training: no real customers, sellers or company. The tasks and data are Atterna’s; the SQL and the decisions are mine.
Projects
| Project | Business question | What I did | Headline | My call |
|---|---|---|---|---|
| 1. | Take rate fell by more than a point in the promo week. | Case study · 3 SQL queries · decision memo from my choices | Take rate from the case data: 12.6% in W37, 11.5% in the promo week. | No reversal: the mix explains the drop |
| 2. | Vera wants to drop the most-refunded products from the December promo. | Case study · 4 SQL queries · decision memo from my choices | Refunds booked: €16,073 on 323 refunds. | No defensible cut yet: the ranking moves and item-level evidence is thin |
| 3. | The dashboard shows record GMV. What really happened to the money? | Meridian episode 1 · 6 SQL queries · memo in my own words | −€9,984.50 Platform contribution in the promo week | Stop the promo and run a test |
| 4. | Product's report says conversion went up. | Meridian episode 2 · 5 SQL queries · recommendation | On the fair part the badge probably raises conversion: +1.5 pp, 95% interval +0.2 pp to +2.9 pp. | Turn the badge on now for web and app 5.1 and 5.3 |
| 5. | The retention dashboard shows a collapse. | Meridian episode 3 · 5 SQL queries · recommendation | 7.20% Month-1 retention of promotion customers | A discount only on a new customer's first order of €40 or more |
| 6. | In January the platform launches the Trusted Shop programme. | Meridian episode 4 · 5 SQL queries · recommendation | The problem sits with a few shops: 17 shops break the rules, 9 also beyond chance; together they carry 11% of GMV and up to 40%… | Judge shops with 30+ orders on cancellations (≤5%) and late shipments (≤15%) |
| 7. | Friday. I need a final report for the board: for each category — orders, GMV, average order value and share of total GMV. | Course final project · 8 SQL queries | All 8 queries passed the check | SUM() OVER (), ROW_NUMBER and PARTITION BY, Running total |
About this work
- Every SQL answer passed Atterna’s automatic checker, which compares the result with a hidden reference on two versions of the data, so a hard-coded number fails. It is Atterna’s own check, not a third-party certification.
- These are my saved SQL submissions. Queries supplied by the task and recognised starter templates are labelled. The automatic checks assess the results, not independent authorship.
- Words are labelled. “My own words” is text I typed. “Assembled from my choices” is built from options I selected in the exercise, so it is not free text.
For hiring managers
- Per query: each project’s “How it was checked” lists attempts and hint use. Queries the task supplied or pre-filled, and answers identical to the finished query in the hints, are not counted as first tries. Hint use that was not recorded is reported as not recorded, not as zero.
- Not shown: who typed the work. Practice is not proctored, and every Atterna learner gets the same tasks on the same fictional data. Ask me to walk you through any query.
How to read this repository
- Each folder in
projects/is one project. Start with itsREADME.md: the question, my approach, the result and the decision. queries.sqlholds all the SQL of a project in one file.results/has the result of each query as CSV, plus a chart where the result is one label and one number.
Next step
Next: I repeat this method on a public dataset. Edit this line and link the repository when it is ready.
Sample generated by Atterna from synthetic progress on fictional Meridian data. Task and data: Atterna. SQL and decisions: the sample learner’s. 10 October 2026. Data version: 2026-09-26.en1.