Conversion timing prediction for networked advertising
A computer-implemented method for conversion timing inference. A conversion timing model is model is configured to predict a likelihood of conversion based on an entity's elapsed time since a qualified entry event. The conversion timing model is constructed based on a distribution of the conversion timespans of converters. Each conversion timespan describes a length of time between a qualified entry event and a conversion event for a converted entity. A notification of an opportunity to expose a candidate entity to networked content is received and the likelihood of conversion for the candidate entity is determined by: determining an elapsed time since a qualified entry event for the candidate entity and applying the conversion timing model to the elapsed time. A response to the notification based on the likelihood of conversion for the candidate entity is prepared. Timely responses may include the selection of customized content, customized advertising content or bid values.
1. A computer-implemented method comprising:
receiving, from a data repository, histories of a plurality of converted entities, each history comprising records of events associated with a respective converted entity;
calculating a respective conversion timespan for each converted entity, the respective conversion timespan comprising a time difference between an entry event of the respective converted entity and a conversion event of the respective converted entity;
fitting a conversion timing model to a distribution of the calculated conversion timespans, wherein the conversion timing model is configured to predict, from an elapsed time since an entry event of an entity, a likelihood of conversion for the entity;
receiving, from a real time bidding exchange, notification of an opportunity to bid on an advertising exposure associated with a candidate entity;
responsive to receiving notification of the opportunity to bid, predicting a likelihood of conversion for the candidate entity by applying the conversion timing model to an elapsed time since an entry event of the candidate entity;
and
preparing a response to the notification, the response comprising a bid price determined from the predicted likelihood of conversion for the candidate entity.
2. The method of claim 1 further comprising:
sending the prepared response to the exchange.
3. The method of claim 1 wherein:
predicting a likelihood of conversion comprises predicting a likelihood of conversion within a time window.
4. The method of claim 1 wherein:
fitting comprises fitting the distribution of calculated conversion timespans with a gamma mixture model configured with a number of conversion modes.
5. The method of claim 1 wherein:
preparing a response comprises selecting an advertising creative according to the predicted likelihood of conversion.
6. The method of claim 5 further comprising:
sending the selected advertising creative to the candidate entity.
7. The method of claim 1 wherein:
preparing comprises customizing content.
8. A system comprising:
a processor; and
a computer readable storage medium storing processor-executable computer program instructions, the instructions comprising instructions for:
receiving, from a data repository, histories of a plurality of converted entities, each history comprising records of events associated with a respective converted entity;
calculating a respective conversion timespan for each converted entity, the respective conversion timespan comprising a time difference between an entry event of the respective converted entity and a conversion event of the respective converted entity;
fitting a conversion timing model to a distribution of the calculated conversion timespans, wherein the conversion timing model is configured to predict, from an elapsed time since an entry event of an entity, a likelihood of conversion for the entity;
receiving, from a real time bidding exchange, notification of an opportunity to bid on an advertising exposure associated with a candidate entity;
responsive to receiving notification of the opportunity to bid, predicting a likelihood of conversion for the candidate entity by applying the conversion timing model to an elapsed time since an entry event of the candidate entity;
and
preparing a response to the notification, the response comprising a bid price determined from the predicted likelihood of conversion for the candidate entity.
9. The system of claim 8 wherein the instructions further comprise:
sending the prepared response to the exchange.
10. The system of claim 8 wherein:
predicting a likelihood of conversion comprises predicting a likelihood of conversion within a time window.
11. The system of claim 8 wherein:
fitting comprises fitting the distribution of calculated conversion timespans with a gamma mixture model configured with a number of conversion modes.
12. The system of claim 8 wherein:
preparing a response comprises selecting an advertising creative according to the predicted likelihood of conversion.
13. The system of claim 12 wherein the instructions further comprise:
sending the selected advertising creative to the candidate entity.
14. The system of claim 8 wherein:
preparing comprises customizing content.
15. A non-transitory computer readable storage medium executing computer program instructions, the computer program instructions comprising instructions for:
receiving, from a data repository, histories of a plurality of converted entities, each history comprising records of events associated with a respective converted entity;
calculating a respective conversion timespan for each converted entity, the respective conversion timespan comprising a time difference between an entry event of the respective converted entity and a conversion event of the respective converted entity;
fitting a conversion timing model to a distribution of the calculated conversion timespans, wherein the conversion timing model is configured to predict, from an elapsed time since an entry event of an entity, a likelihood of conversion for the entity;
receiving, from a real time bidding exchange, notification of an opportunity to bid on an advertising exposure associated with a candidate entity;
responsive to receiving notification of the opportunity to bid, predicting a likelihood of conversion for the candidate entity by applying the conversion timing model to an elapsed time since an entry event of the candidate entity;
and
preparing a response to the notification, the response comprising a bid price determined from the predicted likelihood of conversion for the candidate entity.
16. The medium of claim 15 , the instructions further comprising:
sending the prepared response to the exchange.
17. The medium of claim 15 wherein:
predicting a likelihood of conversion comprises predicting a likelihood of conversion within a time window.
18. The medium of claim 15 wherein:
fitting comprises fitting the distribution of calculated conversion timespans with a gamma mixture model configured with a number of conversion modes.
19. The medium of claim 15 wherein:
preparing a response comprises selecting an advertising creative according to the predicted likelihood of conversion.
20. The medium of claim 19 wherein the instructions further comprise:
sending the selected advertising creative to the candidate entity.