Mock interview. You're interviewing for a senior risk data scientist job on a payments team. The interviewer says: "Every payment gets a risk_score from 0 to 1 from a model built years ago. We're about to start blocking high-score payments. Tell me if the score actually works."
For this interview, a payment is bad if it was refunded or failed. A good risk score should give bad payments higher scores.
"Split all payments into 10 equal bands by risk_score (1 = lowest scores). Show each band's payments, its score range and its bad rate."
Expected output
band, payments, min_score, max_score, bad_rate (a percentage rounded to 1 decimal); sorted by bandExample of the shape (made-up values):
| band | payments | min_score | max_score | bad_rate |
|---|---|---|---|---|
| 1 | 250 | 0.000 | 0.010 | 4.0 |
Hints, the later steps and the answer come with Pro.
Step 1 of 4
Write your query here. Ctrl + Enter submits.