The provided data asset is relational and consists of four distinct data files.

1. address.csv: contains address information

2. customer.csv: contains customer information.

3. demographic.csv: contains demographic data

4. termination.csv: includes customer termination information.

5. autoinsurance_churn.csv: includes merged customer churn data generated from this notebook.

All data sets are linked using either ADDRESS_ID or INDIVIDUAL_ID. The ADDRESS_ID pertains to a specific postal service address, while the INDIVIDUAL_ID is unique to each individual. It is important to note that multiple customers may be assigned to the same address, and not all customers have demographic information available.

Size of the data set

The data set includes 1,536,673 unique addresses and 2,280,321 unique customers, of which 2,112,579 have demographic information. Additionally, 269,259 customers cancelled their policies within the previous year.

Note

Please note that the data is synthetic, and all customer information provided is fictitious. While the latitude-longitude information can be mapped at a high level and generally refers to the Dallas-Fort Worth Metroplex in North Texas, it is important to note that drilling down too far may result in some data points that are located in the middle of Jerry World, DFW Airport, or Lake Grapevine. The physical addresses provided are fake and are unrelated to the corresponding lat/long.

The termination table includes the ACCT_SUSPD_DATE field, which can be used to derive a binary churn/did not churn variable. The data set is modelable, meaning that the other data available can be used to predict which customers churned and which did not. The underlying logic used to make these predictions should align with predicting auto insurance churn in the real world.