Churn analysis and methods of intervention

1. A method for providing content recommendations to prevent user churn, the method comprising:

accessing content item consumption data for a plurality of users previously or currently subscribed to a media service, wherein the content item consumption data comprises information about time slots when content items are consumed by users;

determining, by querying a database, that a first subset of the plurality of users has unsubscribed from the media service during a predetermined time period and that a second subset of the plurality of users has not unsubscribed from the media service during the predetermined time period;

identifying a time slot typical for the first subset of users and atypical for the second subset of users based on content item consumption data of the first subset of users and content item consumption data of the second subset of users;

determining that a user is consuming a first content item at the identified time slot; and

in response to determining that the user is consuming the first content item at the identified time slot, generating for display a recommendation for a second content item, wherein the second content item is scheduled for a time slot that is different from the identified time slot.

2. The method of claim 1 , wherein the content item consumption data further comprises content item type information, the method further comprising:

identifying a content item type typical for the first subset of users and atypical for the second subset of users based on the content item consumption data of the first subset of users and the content item consumption data of the second subset of users; and

wherein the first content item is of a type that is the same as the identified content type, and the second content item is of a type that is the different from the identified content type.

3. The method of claim 1 , further comprising:

maintaining a machine learning churn prediction algorithm configured to identify a time slot predictive of churn based on the content item consumption data;

iteratively adjusting the machine learning churn prediction algorithm based on the identified time slot to accurately predict a user unsubscribing from the media service; and

wherein the identifying the time slot comprises identifying the time slot using the adjusted machine learning churn prediction algorithm.

4. The method of claim 1 , further comprising:

maintaining an intervention selection algorithm configured to select a content item based on data indicating that users are consuming content items at the identified time slot;

iteratively adjusting the intervention selection algorithm based on the user not unsubscribing from the media service after viewing a recommendation for the selected content item; and

wherein the generating for display the recommendation comprises generating for display a recommendation for the content item selected using the adjusted intervention content selection algorithm.

5. The method of claim 1 , wherein:

identifying the time slot that is typical for the first subset of users comprises:

calculating a first fraction of the first subset of users that have consumed content items at the time slot; and

determining that the first fraction exceeds a typicality threshold; and

identifying that the time slot is atypical for the second subset of users comprises:

calculating a second fraction of the second subset of users that have consumed content items at the time slot; and

determining that the second fraction does not exceed a typicality threshold.

6. The method of claim 1 , further comprising:

identifying a time slot typical for the second subset of users and atypical for the first subset of users based on content item consumption data of the first subset of users and content item consumption data of the second subset of users; and

wherein the generating for display the recommendation for the second content item is performed in response to determining that the user has not consumed content at the time slot typical for the second subset of users for a predetermined period of time.

7. The method of claim 6 , wherein the second content item is scheduled for the time slot typical for the second subset of users and atypical for the first subset of users.

8. The method of claim 1 , further comprising:

identifying a first set of time slots typical for the first subset of users and atypical for the second subset of users based on content item consumption data of the first subset of users and content item consumption data of the second subset of users;

identifying a second set of time slots typical for the second subset of users and atypical for the first subset of users based on content item consumption data of the first subset of users and content item consumption data of the second subset of users;

wherein generating for display the recommendation for the second content item is performed in response to determining that the user has consumed more than a first threshold amount of content items scheduled for time slots of the first set and has consumed less than a second threshold amount of content items scheduled for time slots of the second set over a predetermined time period.

9. The method of claim 1 , wherein the generating for display a recommendation for the second content item comprises generating for display an incentive to consume the second content item.

10. The method of claim 1 , wherein the identifying the time slot typical for the first subset of users and atypical for the second subset of users comprises using at least one of regression model and Shapley additive explanation model.

11. A system for providing content recommendations to prevent user churn, the system comprising:

control circuitry configured to:

access content item consumption data for a plurality of users previously or currently subscribed to a media service, wherein the content item consumption data comprises information about time slots when content items are consumed by users;

determine, by querying a database, that a first subset of the plurality of users has unsubscribed from the media service during a predetermined time period and that a second subset of the plurality of users has not unsubscribed from the media service during the predetermined time period;

identify a time slot typical for the first subset of users and atypical for the second subset of users based on content item consumption data of the first subset of users and content item consumption data of the second subset of users;

determine that a user is consuming a first content item at the identified time slot; and

in response to determining that the user is consuming the first content item at the identified time slot, generate for display, using display circuitry, a recommendation for a second content item, wherein the second content item is scheduled for a time slot that is different from the identified time slot.

12. The system of claim 11 , wherein the content item consumption data further comprises content item type information, and wherein the control circuitry is further configured to:

identifying a content item type typical for the first subset of users and atypical for the second subset of users based on the content item consumption data of the first subset of users and the content item consumption data of the second subset of users; and

wherein the first content item is of a type that is the same as the identified content type, and the second content item is of a type that is the different from the identified content type.

13. The system of claim 11 , wherein the control circuitry is further configured to:

maintain a machine learning churn prediction algorithm configured to identify a time slot predictive of churn based on the content item consumption data;

iteratively adjust the machine learning churn prediction algorithm based on the identified time slot to accurately predict a user unsubscribing from the media service; and

wherein the control circuitry is configured to identify the time slot by identifying the time slot using the adjusted machine learning churn prediction algorithm.

14. The system of claim 11 , wherein the control circuitry is further configured to:

maintain an intervention selection algorithm configured to select a content item based on data indicating that users are consuming content items at the identified time slot;

iteratively adjust the intervention selection algorithm based on the user not unsubscribing from the media service after viewing a recommendation for the selected content item; and

wherein the control circuitry is configured to generate for display the recommendation by generating for display a recommendation for the content item selected using the adjusted intervention content selection algorithm.

15. The system of claim 11 , wherein:

the control circuitry is configured to identify the time slot that is typical for the first subset of users by:

calculating a first fraction of the first subset of users that have consumed content items at the time slot; and

determining that the first fraction exceeds a typicality threshold; and

the control circuitry is configured to identify that the time slot is atypical for the second subset of users by:

calculating a second fraction of the second subset of users that have consumed content items at the time slot; and

determining that the second fraction does not exceed a typicality threshold.

16. The system of claim 11 , wherein the control circuitry is further configured to:

identifying a time slot typical for the second subset of users and atypical for the first subset of users based on content item consumption data of the first subset of users and content item consumption data of the second subset of users; and

wherein the control circuitry is configured to generate for display the recommendation for the second content item in response to determining that the user has not consumed content at the time slot typical for the second subset of users for a predetermined period of time, and

wherein the second content item is scheduled for the time slot typical for the second subset of users and atypical for the first subset of users.

17. The system of claim 11 , wherein the control circuitry is further configured to:

identifying a first set of time slots typical for the first subset of users and atypical for the second subset of users based on content item consumption data of the first subset of users and content item consumption data of the second subset of users;

identify a second set of time slots typical for the second subset of users and atypical for the first subset of users based on content item consumption data of the first subset of users and content item consumption data of the second subset of users;

wherein the control circuitry is configured to generate for display the recommendation for the second content item in response to determining that the user has consumed more than a first threshold amount of content items scheduled for time slots of the first set and has consumed less than a second threshold amount of content items scheduled for time slots of the second set over a predetermined time period.

18. The system of claim 11 , wherein the control circuitry is configured to generate for display a recommendation for the second content item by generating for display an incentive to consume the second content item.

19. The system of claim 11 , wherein the control circuitry is configured to identify the time slot typical for the first subset of users and atypical for the second subset of users by using at least one of partial least squares regression model and Shapley additive explanation model.

20. A system for providing content recommendations to prevent user churn, the system comprising:

control circuitry configured to:

access content item consumption data for a plurality of users previously or currently subscribed to a media service, wherein the content item consumption data comprises information about time slots when content items are consumed by users and information about content item types of the content items;

determine, by querying a database, that a first subset of the plurality of users has unsubscribed from the media service during a predetermined time period and that a second subset of the plurality of users has not unsubscribed from the media service during the predetermined time period, wherein the database is queried based on the plurality of users associated with the accessed content item consumption data;

identify a time slot typical for the first subset of users and atypical for the second subset of users based on content item consumption data of the first subset of users and content item consumption data of the second subset of users;

identify a content item type typical for the first subset of users and atypical for the second subset of users based on the content item consumption data of the first subset of users and the content item consumption data of the second subset of users;

monitor content item consumption of a user to determine whether the user is consuming a first content item at the identified time slot, wherein the first content item is of a type that is the same as the identified content type;

in response to determining that the user is consuming the first content item at the identified time slot, generate for display, using display circuitry, a recommendation for a second content item, wherein the second content item is scheduled for a time slot that is different from the identified time slot, and wherein the second content item is of a type that is the different from the identified content type; and

in response to determining that the user is not consuming the first content item at the identified time slot, continue to monitor content item consumption of the user.