vaibhavkhurana/subscriber-retention-intelligence

Subscriber Retention Intelligence

Complete transformed analytical corpus behind the Subscriber Retention Intelligence product. Release public-m12 contains 442,211,685 accepted rows in 32 Zstandard-compressed Parquet files.

Contents

Configuration	Grain	Rows
member	One row per subscriber	6,769,473
subscription_transaction	One row per accepted transaction	22,975,416
churn_label	One subscriber per observed label window	1,963,891
listening_day	One subscriber activity day	410,502,905

Total compressed size is 7,522,254,856 bytes.

Identity and privacy

Subscriber and transaction keys are release-specific, collision-checked 64-bit one-way pseudonyms. Source identifiers and the private publication salt are excluded. Exact reported age is generalized to a 10-year band. Exact registration date is reduced to registration month. Implausible or missing age values remain null.

Load

from datasets import load_dataset

members = load_dataset(
    "vaibhavkhurana/subscriber-retention-intelligence",
    "member",
    split="full",
)

Choose subscription_transaction, churn_label, or listening_day for the other table configurations. Parquet files also support direct reads from DuckDB, Polars, Spark, and pandas.

Interpretation

This is a historical consumer-subscription dataset. Payment amounts are gross receipts in source currency, not recognized revenue. Churn labels are observed challenge outcomes, not treatment effects. The release does not contain enterprise accounts, contracts, seats, outreach records, or causal intervention results.

Source and terms

Source: KKBox, WSDM Churn Prediction Challenge, distributed through Kaggle. Public analytical use and this transformed detailed release were approved by the dataset owner for this project. Source-provider terms continue to apply. This dataset card does not relicense provider-owned data.

Product and methodology: https://github.com/vaibhavkhuranaaa/subscriber-retention-intelligence

Live analytical product: https://subscriber-retention-intelligence.pages.dev/