Closed 12 Oct
Subscription churn risk
Predict which subscribers cancel in the next 30 days.
- Scored by
- ROC AUC
- Higher is better
- Score to beat
- 0.763
- Logistic regression baseline
- Entrants
- 0
- 5 uploads a day
- Closed
- 12 Oct
- 00:00 UTC
Overview
A subscription business wants to know who is about to leave, early enough to do something about it.
You get 12,000 labelled subscribers and 4,000 unlabelled ones. Predict, for each subscriber in the test set, the probability that they churn within 30 days.
Submit a probability between 0 and 1, not a hard 0/1 label, the metric rewards getting the ordering right, and a hard label throws away all the confidence information you have.
Marking rubric
Exactly how your points are worked out. Mark your own work against it before you file.
All 100 points come from your score on the hidden test rows. There is no report in this competition.
The metric
100 points
Scored with ROC AUC (higher is better) on the private 70% of the test rows, which nobody sees until results. Your rank there becomes your points: first place earns the full 100, last place earns 0, and everyone between is spaced evenly by rank.
A score that does not beat the logistic regression baseline earns 0, whatever its rank. Ranks, not raw scores, so one runaway entry cannot squash the rest of the field.
points = 100 × (1 − (rank − 1) ÷ (entrants − 1))For example, 3rd of 21: 100 × (1 − 2 ÷ 20) = 90 points.
Medals and prizes
- With five or more entrants: gold for the top 10%, silver for the next 20%, bronze for the next 30%.
- Every medal needs a score that beats the baseline.
The full rules and terms apply to every entry.
Submission format
One thing to hand in: a predictions file, scored automatically.
Your predictions, as a CSV
- Header
customer_id,churned. Column order does not matter; extra columns are ignored.- Rows
- 4,000 rows, one per id in test.csv. A missing id rejects the file and tells you which.
- churned
- A probability between 0 and 1 that the answer is the positive class.
- File
- UTF-8, comma separated, up to 10 MB. sample_submission.csv is a valid starting point.
- Limits
- 5 scored uploads a day (resets 00:00 UTC). A file that fails these checks is rejected with the reason and does not use one. Pick up to 2 for final scoring; pick none and your best public score is used.
Leaderboard
0 entrants, best public score each. The baseline and our reference model are pinned where their scores fall.
| Rank | Entrant | Uploads | Best score |
|---|---|---|---|
| Logistic regression baselinebaseline to beat | 0.763 | ||
| No scored uploads yet. The first scored upload takes rank 1. | |||
Scored on the public 30% of the test rows while the competition runs. Final standings use the other 70%, so a public lead is not a final one. The two reference rows are scored on those final rows.
Data
| Column | What it holds |
|---|---|
customer_id | Unique subscriber id. Present in both train and test. |
tenure_months | Months since the subscription started. |
monthly_charges | Current monthly bill, in USD. |
support_tickets | Support tickets opened in the last 90 days. |
contract_type | One of month_to_month, one_year, two_year. |
auto_pay | 1 if automatic payment is enabled, else 0. |
late_payments | Late payments in the last 12 months. |
avg_session_minutes | Mean minutes per session over the last 30 days. |
region | One of north, south, east, west. |
churned predict this | TARGET. 1 if the subscriber cancelled within 30 days, else 0. Train only. |
Sign in to download train.csv, test.csv and sample_submission.csv.
A playground contest: the data card, sources and audit are published for Season 1 problems.
Timeline
OpensMon, 14 Sept 2026(done)
Data released; uploads accepted.
ClosesMon, 12 Oct 2026(done)
Selections are frozen. The last 48 hours are a quiet period.
Rules
- Submissions per day: 5, resetting at 00:00 UTC.
- Final scoring: pick up to 2 submissions to count. If you pick none, your best public submission is used automatically.
- Public vs private: the live leaderboard is scored on ~30% of the test rows. Final standings use the other ~70%, which nobody sees until the competition closes.
- Teams: individual entry only for now.
- External data: allowed, provided it is public and you say so in your writeup.
- One account per person. Multiple accounts, or sharing submissions between accounts, means disqualification.
Submit a CSV with exactly two columns: customer_id and churned, one row for each of the 4,000 test rows.
The full competition terms apply to every entry.
Read the brief, then enter.
Free with a student account. The brief and the public leaderboard need no account.