Granted Patent B2

US 10,699,298 · App. 15/482,770 · Granted Jun 30, 2020

Inventors: John Haws (Durham, NC); Yohan Lejosne (Boston, MA); William Lefew (Cary, NC)

Assignee: Digital Turbine, Inc.

G06Q30/0254G06F7/16G06Q30/0242G06Q30/0244G06Q30/0246G06Q30/0255G06Q30/0269

Abstract

Claims (82)

1. A computer-implemented method for generating digital content, comprising:

receiving a plurality of samples, each comprises a request having a plurality of values of a plurality of attributes describing structural or functional properties of a device, said sample associated with a success value for a Bernoulli distributed user action event having a campaign having one or more digital content, and a bid rate (BR) of said campaign;

clustering said plurality of samples in a plurality of homogenous sample groups according to respective said plurality of values;

identifying a group campaign with a highest valuation for each one of said plurality of sample groups using triangular approximation of the Bernoulli distribution of user action events in said sample group;

receiving a query from a device including a plurality of other values of said plurality of attributes describing structural or functional properties of said device, wherein said plurality of other values includes at least one certain value for one certain attribute describing a property of said device;

selecting one of said plurality of sample groups, wherein said selected sample group has a sample having said certain value of said certain attribute describing said property of another device;

selecting a digital content of said group campaign with highest valuation identified for said selected sample group; and

generating, utilizing at least one hardware processor, advertising digital content in response to said query, including said selected digital content based on the highest valuation calculation.

2. The method of claim 1 , wherein said selected digital content is an advertisement for the purpose of displaying or playing said advertisement on said device;

wherein said plurality of attributes are members of a group including: an identifier identifying a software application or web site running on said device, a country where said device is located, an identifier of an operating system running on said device, an operating system version identifier, a size of advertisement requested, a make of said device, a model of said device, a network connection type, a gender of a user, and a city where said device is located; and

wherein said user action event is a member of a group including: a user clicking on a certain location in said software application or web site and said user installing another software.

3. The method of claim 1 , wherein said identifying a group campaign with a highest valuation comprises:

generating a list of samples comprising all samples clustered in said sample group;

generating a list of events comprising all Bernoulli distributed user action events associated with said list of samples;

assuming for each one event in said list of events a conversion rate of said event in said list of samples (CVR);

producing a plurality of valuations by calculating for each one event in said list of events a valuation in said list of samples (EV), comprising:

using a confidence level (τ);

applying numerical methods to find a possible conversion rate (θ) such that the probability that said CVR is greater than or equal to said θ is equal to said τ; and

multiplying said θ by said campaign's BR to compute said EV;

identifying a highest valuation of said plurality of valuations in said list of samples (EV best );

producing a plurality of probabilistic scores by calculating a probabilistic score (P i ) for each one event in said list of events, comprising:

calculating a probability that said EV is greater than said EV best by using triangular approximation of the Bernoulli distribution of said one event in said list of events; and

associating said P i with said one event;

identifying a highest probabilistic score of said plurality of probabilistic scores;

choosing said campaign of said one event associated with said highest probabilistic score; and

associating said chosen campaign with said sample group.

4. The method of claim 3 , wherein using triangular approximation of the Bernoulli distribution of said one event in said list of events comprises using an isosceles triangle centered around said EV best , and having a base equaling four times the standard deviation of the distribution of said event and a height equaling the inverse of two times the standard deviation of the distribution of said event.

5. The method of claim 4 , wherein said numerical methods include using a numerical solver.

6. The method of claim 1 , wherein said clustering said plurality of samples in a plurality of homogenous sample groups according to respective said plurality of values comprises:

for each one of said plurality of samples:

choosing another at least one certain value of another at least one certain attribute describing a property of a device of said plurality of values;

selecting one of said plurality of sample groups, having another of said plurality of samples having said other at least one certain value of said other at least one certain attribute describing said property of another device; and

adding said sample to said selected sample group.

7. The method of claim 1 , further comprising storing said plurality of sample groups in a storage.

8. The method of claim 7 , wherein said storage is at least one database, said at least one database is a member of a group including: a local database and a remote database.

9. The method of claim 1 , further comprising:

receiving another sample comprising a plurality of third values of said plurality of attributes describing said properties of a third device and associated with another success value for another Bernoulli distributed user action event, having a second campaign having one or more other digital content, and a second bid rate (BR i ) of said second campaign;

selecting another of said plurality of sample groups;

adding said other sample to said other sample group; and

identifying a new group campaign with a highest valuation using triangular approximation of the Bernoulli distribution of user action events in said other sample group.

10. The method of claim 9 , wherein said identifying a new campaign with a highest valuation using triangular approximation of the Bernoulli distribution of user action events in said other sample group comprises:

generating a new list of samples comprising all samples clustered in said other sample group;

generating a new list of events comprising all Bernoulli distributed user action events associated with said new list of samples;

assuming for each one event in said new list of events a conversion rate of said one event in said new list of samples (CVR new );

producing a plurality of new valuations by calculating for each one event in said new list of events a new valuation in said new list of samples (EV new ), comprising:

using another confidence level (τ new );

applying numerical methods to find a new possible conversion rate (θ new ) such that the probability that said CVR new is greater than or equal to said θ new is equal to said τ new ; and

multiplying said θ new by said second campaign's second bid rate BR i ;

identifying a new highest valuation of said plurality of new valuations in said new list of samples (EV newbest );

producing a plurality of new probabilistic scores by calculating a new probabilistic score (P inew ) for each one other event in said new list of events, comprising:

calculating a new probability that said EV new is greater than said EV newbest by using triangular approximation of the Bernoulli distribution of said one other event in said new list of events; and

associating said Pi new with said one other event;

identifying a new highest probabilistic score of said new plurality of probabilistic scores;

choosing said second campaign of said one other event associated with said new highest probabilistic score; and

associating said chosen second campaign with said other sample group.

11. The method of claim 10 , wherein said numerical methods include using a numerical solver.

12. A software program product for generating digital content, comprising:

a non-transitory computer readable storage medium;

first program instructions for receiving a plurality of samples, each comprises a request having a plurality of values of a plurality of attributes describing structural or functional properties of a device, said sample associated with a success value for a Bernoulli distributed user action event having a campaign having one or more digital content, and a bid rate (BR) of said campaign;

second program instructions for clustering said plurality of samples in a plurality of homogenous sample groups according to respective said plurality of values;

third program instructions for identifying a group campaign with a highest valuation for each one of said plurality of sample groups using triangular approximation of the Bernoulli distribution of user action events in said sample group;

fourth program instructions for receiving a query from a device including a plurality of other values of said plurality of attributes describing structural or functional properties of said device, wherein said plurality of other values includes at least one certain value for one certain attribute describing a property of said device;

fifth program instructions for selecting one of said plurality of sample groups wherein said selected sample group has a sample having said certain value of said certain attribute describing said property of another device;

sixth program instructions for selecting a digital content of said group campaign with highest valuation identified for said selected sample group; and

seventh program instructions for generating advertising digital content in response to said query, including said selected digital content based on the highest valuation calculation;

wherein said first, second, third, fourth, fifth, sixth and seventh program instructions are executed by at least one computerized processor from said non-transitory computer readable storage medium.

13. A system for generating digital content, comprising:

at least one code storage storing a code;

at least one hardware processor coupled to said at least one code storage for executing said code for:

receiving a plurality of samples, each comprises a request having a plurality of values of a plurality of attributes describing structural or functional properties of a device, said sample associated with a success value for a Bernoulli distributed user action event having a campaign having one or more digital contents, and a bid rate (BR) of said campaign;

clustering said plurality of samples in a plurality of homogenous sample groups according to respective said plurality of values;

identifying a group campaign with a highest valuation for each one of said plurality of sample groups using triangular approximation of the Bernoulli distribution of user action events in said sample group;

receiving a query from a device including a plurality of other values of said plurality of attributes describing structural or functional properties of said device, wherein said plurality of other values includes at least one certain value for one certain attribute describing a property of said device;

selecting one of said plurality of sample groups, wherein said selected sample group has a sample having said certain value of said certain attribute describing said property of another device;

selecting a digital content of said group campaign with highest valuation identified for said selected sample group; and

generating advertising digital content in response to said query, including said selected digital content based on the highest valuation calculation.

14. The system of claim 13 , wherein said receiving a plurality of samples is by reading said plurality of samples from said code storage.

15. The system of claim 13 , wherein said receiving a plurality of samples is via a network.

16. The system of claim 13 , further comprising a local digital memory hardware electrically coupled to said at least one hardware processor.

17. The system of claim 16 , wherein said receiving a plurality of samples is by reading said plurality of samples from said local digital memory hardware.

18. The system of claim 13 , further comprising a remote storage, connected to said at least one hardware processor via a network.

19. The system of claim 18 , wherein said receiving a plurality of samples is by reading said plurality of samples from said remote storage.

Assignments (9)

Continuity (1)