Method and system for identifying users across mobile and desktop devices

Patent metadata

Field	Value
Publication number	US-10423985-B1
Application number	US-201615019959-A
Country	US
Kind code	B1
Filing date	Feb 9, 2016
Priority date	Feb 9, 2015
Publication date	Sep 24, 2019
Grant date	Sep 24, 2019

Abstract

A method for conversion attribution. The method includes obtaining a first identifier associated with a first device, obtaining a second identifier associated with a second device, bridging the first identifier and the second identifier based on a determination, made by a probabilistic classifier, that the first identifier and the second identifier are associated with a common user, and attributing, using the bridge, a conversion on a website accessed using the second device.

First claim

What is claimed is:

1. A computer-implemented method, comprising:
obtaining a collection of mappings between pairs of identifiers using deterministic ID bridging, wherein the mapping between a pair of identifiers indicates that the identifiers are associated with a common user;
training a probabilistic classifier using the collection of mappings between pairs of identifiers, wherein the probabilistic classifier is trained to generate a probability that two input identifiers are associated with a particular common user;
obtaining a first identifier associated with a first device;
obtaining a second identifier associated with a second device;
using the probabilistic classifier to determine a probability that the first identifier and the second identifier are associated with a first user:
bridging the first identifier and the second identifier based on a determination, made by the probabilistic classifier, that the probability that first identifier and the second identifier are associated with the first user satisfies a specified level of confidence; and
attributing, using the bridge, a conversion on a website accessed using the second device following presentation of content on the first device.

2. The method of claim 1 , wherein training the probabilistic classifier comprises:
labeling a first subset of the collection of mappings between pairs of identifiers as true positive;
labeling a second subset of the collection of mappings between pairs of identifiers as false positive;
selecting features for each of the pairs in the first subset and in the second subset; and
training the probabilistic classifier using the features for the pairs in the first subset and in the second subset.

3. The method of claim 2 , wherein a feature of the features includes an IP address affiliated with a pairs of identifiers, a time window during which a pair was seen, and a weight.

4. The method of claim 3 , wherein the weight corresponds to a number of unique device identifiers seen by an ad exchange, and associated with the IP address, over all time windows.

5. The method of claim 2 , wherein a feature of the features includes an IP address affiliated with a pair of identifiers, a time window during which a pair was seen, and a probability.

6. The method of claim 5 , wherein the probability corresponds to ratio of a number of times a device ID was associated with the IP address and the number of times any device ID was associated with the IP address.

7. The method of claim 2 , wherein labeling the first subset of the collection of mappings between pairs of identifiers comprises using results of deterministic ID bridging.

8. The method of claim 2 , wherein the first feature set includes at least one selected from a group consisting of an IP address associated with each identifier of a pair of identifiers, and a probability associated with a candidate pair of identifiers.

9. The method of claim 1 , further comprising:
obtaining a third identifier from a third device; and
determining, using the probabilistic classifier, that the first identifier and the third identifier are associated with different users.

10. The method of claim 1 , wherein the content is an advertisement and wherein conversion is attributed when the user accesses the webpage within a predetermined period of time after viewing the advertisement on the first device.

11. The method of claim 1 , wherein an Internet Protocol (IP) address is associated with the first device and the IP address is associated with the second device.

12. The method of claim 1 , wherein the first identifier is a device ID.

13. The method of claim 12 , wherein the second identifier is a guest ID, assigned to the user of the second device when the user navigated to a second website, wherein the user is not logged onto a service associated with the website.

14. A non-transitory computer readable medium comprising instructions for performing a method, the method comprising:
obtaining a collection of mappings between pairs of identifiers using deterministic ID bridging, wherein the mapping between a pair of identifiers indicates that the identifiers are associated with a common user;
training a probabilistic classifier using the collection of mappings between pairs of identifiers, wherein the probabilistic classifier is trained to generate a probability that two input identifiers are associated with a particular common user;
obtaining a first identifier associated with a first device and a second identifier associated with a second device;
using the probabilistic classifier to determine a probability that the first identifier and the second identifier are associated with a first user:
bridging the first identifier and the second identifier based on a determination, made by the probabilistic classifier, that the probability that first identifier and the second identifier are associated with the first user satisfies a specified level of confidence; and
attributing, using the bridge, a conversion on a website accessed using the second device, wherein the conversion is attributed when the user accesses the webpage within a predetermined period of time after viewing an advertisement on the first device.

15. The non-transitory computer readable medium of claim 14 , wherein training the probabilistic classifier comprises:
labeling a first subset of the collection of mappings between pairs of identifiers as true positive;
labeling a second subset of the collection of mappings between pairs of identifiers as false positive;
selecting features for each of the pairs in the first subset and in the second subset; and
training the probabilistic classifier using the features for the pairs in the first subset and in the second subset.

16. The non-transitory computer readable medium of claim 15 , wherein a feature of the features includes an IP address affiliated with pair of identifiers, a time window during which a pair was seen, and a weight.

17. The non-transitory computer readable medium of claim 16 , wherein the weight corresponds to a number of unique device identifiers seen by an ad exchange, and associated with the IP address, over all time windows.

18. The non-transitory computer readable medium of claim 15 , wherein a feature of the features includes an IP address affiliated with pair of identifiers, a time window during which a pair was seen, and a probability, wherein the probability corresponds to ratio of a number of times a device ID was associated with the IP address and the number of times any device ID was associated with the IP address.

19. The method of claim 1 , wherein the probabilistic classifier is applied to features of the first identifier and the second identifier, and wherein the features include one or more IP addresses associated with each identifier over a specified time window.

20. A system comprising:
one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
obtaining a collection of mappings between pairs of identifiers using deterministic ID bridging, wherein the mapping between a pair of identifiers indicates that the identifiers are associated with a common user;
training a probabilistic classifier using the collection of mappings between pairs of identifiers, wherein the probabilistic classifier is trained to generate a probability that two input identifiers are associated with a particular common user;
obtain

Assignees

• Twitter Inc

Inventors

• Dutta Chinmoy
• Kancha Santosh
• Li Junjun
• Lu Wanchen
• Mahajan Milind
• Pandey Sandeep
• Qin Xiaochuan
• Ranadive Ameet
• Rastogi Vibhor
• Rizvi Shariq
• Shrivastava Abhishek
• Wu Yimin
• Zhang Lei
• Zhou Ke

Classifications

• G06Q10/40

  Business processes related to social networking or social networking services · CPC title

• H04L67/306

  User profiles · CPC title

• H04L67/02

  based on web technology, e.g. hypertext transfer protocol [HTTP] · CPC title

• G06Q30/0267

  Wireless devices · CPC title

• G06Q30/0201 Primary

  Market modelling; Market analysis; Collecting market data · CPC title