Analyze a CSV export without touching a spreadsheet

A defensible answer from a messy export in about ten minutes — data-quality flags, the cut you asked for, a chart, and the code that produced it.

Tuned for Julius AI

First

Upload your CSV/Excel file to Julius; send the first block, eyeball the data-quality report, then send the analysis block (edit it to your real question).

The prompt

Here is a data export. Before analyzing:
1. List the columns, row count, and data types.
2. Flag data quality problems: nulls, duplicate rows, obvious outliers,
   and any column that looks mislabeled.
Do NOT analyze yet — just report what you found and wait for me.

---

Now the analysis:
- Group users into weekly cohorts by their signup date.
- For each cohort, compute week-4 retention.
- Define "retained in week 4" explicitly and tell me the definition you used.
- Show the chart AND the Python code you ran.
Flag every assumption you made in bold.

Built-in guardrail: Gates analysis behind a data-quality pass (nulls, dupes, outliers) and forces the model to state its "retained" definition and show the Python code, so the number is checkable rather than trusted blind.

Then

Read the code block, not just the chart — the usual failure is a wrong "retained" definition or a silent duplicate inflating a cohort; correct the definition, re-run, and paste the chart with a one-line note on how retention was defined.

Field-tested Jul 21, 2026. If it stops working, tell us.