Automated targeting of content components

• US 8,799,814 B1
• Filed: 02/22/2008
• Issued: 08/05/2014
• Est. Priority Date: 02/22/2008
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Lloyd, Daniel B., Mason, Llewllyn J., Minter, Samuel A., Bryc, Michal M.
• Primary Examiner(s): TITCOMB, WILLIAM D
• Application Number:
  US12/035,618
• Time in Patent Office: 2,356 Days
• Field of Search: 715/764, 715/779, 715/736, 715/730, 715/732, 715/733, 715/738, 715/744, 715/745, 715/746, 715/788, 715/789, 703/3, 705/1, 705/25
• US Class Current: 715/811
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data
• CPC Class Codes: G06Q 30/0242 Determining effectiveness of advertisements

Abstract

Techniques for automated targeting of content components to users are described. Content components are selected and exposed through renderable pages for viewing by a population of users. User activity following exposure is tracked in an effort to identify which types of users (as characterized by certain attributes) are likely to act on the content components. The users are segmented into groups according to the attributes and the segments are fed back to aid in selection of content components to be exposed to the users. This enables more granular targeting of the content components to those users who exhibit the attributes that define the specific groups.

Claims

1. One or more non-transitory computer-readable media comprising computer-executable instructions that, when executed on one or more processors, perform a method for automated targeting of content components to a population of users, the method comprising:

selecting a content component to expose for viewing by the population of users, the population of users being characterized by a set of items purchased by the population of users;

placing the content component in a particular location on a page to be rendered electronically;

monitoring activity of the population of users in response to exposure of the content component to the population of users;

differentiating among the population of users based on one or more items in the set of items with respect to the content component, wherein the differentiating comprises determining which items in the one or more items are predictive of increased user activity with respect to the content component, the determining comprising;

finding, among the set of items, an item correlated to a largest difference between;

values of a ratio of counts of a number of exposures and counts of a number of click actions of users among the population of users who have previously purchased the one or more items in the set of items; and

values of a ratio of counts of a number of exposures and counts of a number of click actions of users among the population of users who have not previously purchased the one or more items in the set of items; and

segmenting the population of users into multiple groups of users based on the item found to be correlated to the largest difference.

2. The one or more non-transitory computer-readable media as recited in claim 1, wherein the content component is selected from a group of content components comprising advertisements, news, search results, weather, articles, entertainment, financial data, traffic information, recommendations, images, photographs, video, and e-commerce items.

3. The one or more non-transitory computer-readable media as recited in claim 1, wherein the monitoring of activity of the population of users comprises tracking exposures of the content component and user interaction with the content component.

4. The one or more non-transitory computer-readable media as recited in claim 1, wherein the monitoring of activity of the population of users comprises compiling counts of a number of exposures of the content component to the population of users and a number of affirmative actions taken by the users on the content component.

5. The one or more non-transitory computer-readable media as recited in claim 1, wherein the differentiating comprises determining which item purchased by the users in the set of items purchased by the users is correlated to a highest number of affirmative actions taken by the users on the content component.

6. The one or more non-transitory computer-readable media as recited in claim 1, wherein the segmenting comprises building a tree structure for individual placements of the content component, the tree structure having a root node representing a universe of users being observed and one or more subordinate nodes representing the items in the one or more items along which groups of users are segmented.

7. The one or more non-transitory computer-readable media as recited in claim 1, wherein the segmenting comprises:

segmenting the population of users into a first group of users who have purchased the item found to be correlated to the largest difference and a second group of users who have not purchased the item found to be correlated to the largest difference;

differentiating, a second time, among the first group of users based on the one or more items in the set of items purchased by the users, with the exception of the item found to be correlated to the largest difference along which the population was segmented; and

segmenting, a second time, the first group of users who have purchased the item found to be correlated to the largest difference.

8. One or more non-transitory computer-readable media comprising computer-executable instructions that, when executed on one or more processors, perform a method comprising:

monitoring, within a population of users, user activity when content components are exposed for viewing;

identifying, based on the user activity, one or more key items from a set of items that characterize the population of users, wherein the identifying of the key items comprises;

compiling counts, for each item in the set of items, of exposures and click actions taken by individual users on the content components;

associating the counts with one or more items in the set of items based at least in part on whether the one or more items in the set of items were previously purchased by one or more users of the population of users or whether the one or more items in the set of items were not previously purchased by one or more users of the population users; and

finding the one or more key items by finding a largest difference between;

values of a ratio of the counts associated with the one or more items in the set of items previously purchased by the one or more users of the population of users; and

values of a ratio of the counts associated with the one or more items in the set of items not previously purchased by the one or more users of the population of users;

segmenting the population of users into multiple groups based on the one or more key items, wherein the segmenting comprises building binary trees for associated content components, the binary trees having;

(i) a root node that defines a universe of users and that is associated with a feature count table, the feature count table containing;

the values of the ratio of the counts associated with the one or more items in the set of items previously purchased by the one or more users of the population of users; and

the values of the ratio of the counts associated with the one or more items in the set of items not previously purchased by the one or more users of the population of users, and

(ii) children nodes that differentiate segments of the population of users comprising users who have purchased the key items and users who have not purchased the key items; and

targeting different content components to the groups of users.

9. The one or more non-transitory computer-readable media as recited in claim 8, wherein the content components comprise advertising campaigns exposed on renderable pages.

10. The one or more non-transitory computer-readable media as recited in claim 8, wherein the content components comprise merchandising campaigns exposed on renderable pages.

11. The one or more non-transitory computer-readable media as recited in claim 8, wherein the monitoring comprises tracking click actions taken by the users after being exposed to the content components.

12. The one or more non-transitory computer-readable media as recited in claim 8, wherein the targeting comprises selecting different content components to expose to the groups of users.

13. The one or more non-transitory computer-readable media as recited in claim 8, wherein the targeting comprises arranging the content components in different locations for exposure to the groups of users.

14. One or more non-transitory computer-readable media comprising computer-executable instructions that, when executed on one or more processors, perform a method comprising:

tracking user activity with respect to a content component comprising an actionable component that a user can click on when the content component is exposed for viewing by a population of users;

assessing the user activity across a set of items that are descriptive of the population of users to identify one or more key items that are predictive of whether the users will act on the content component, the assessing to identify the one or more key items comprising;

compiling counts, for each item in the set of items, of exposures of the content component and counts of user clicks on the content component;

associating the counts of exposures and the counts of user clicks with one or more items in the set of items based at least in part on whether the one or more items in the set of items were previously purchased by one or more users of the population of users or whether the one or more items in the set of items were not previously purchased by one or more users of the population users; and

finding the one or more key items by finding a largest difference between;

values of a ratio of the counts of exposures and the counts of user clicks associated with the one or more items in the set of items previously purchased by the one or more users of the population of users; and

values of a ratio of the counts of exposures and the counts of user clicks associated with the one or more items in the set of items not previously purchased by the one or more users of the population of users; and

building a learning data structure that represents a universe of users to which the content component is exposed and differentiates groups of users by the one or more key items, wherein the building comprises;

constructing a decision tree structure that is associated with the content component, the decision tree structure having;

(i) a root node that represents the universe of users to which the content component is exposed and that is associated with a feature count table, the feature count table containing;

the values of the ratio of the counts associated with the one or more items in the set of items previously purchased by the one or more users of the population of users; and

the values of the ratio of the counts associated with the one or more items in the set of items not previously purchased by the one or more users of the population of users to measure the user activity, and

(ii) subordinate nodes that represent segments of the universe of users comprising users who have purchased the one or more key items and users who have not purchased the one or more key items.

15. A targeting manager for automated targeting of content components to a population of users, the targeting manager being stored on one or more non-transitory computer-readable media and comprising computer-executable instructions that, when executed on one or more processors, perform acts comprising:

monitoring user activity when a content component is placed in a particular location of a page and exposed for viewing;

identifying one or more items, from a set of predefined items used to characterize the users, that are predictive of user activity with the content component, wherein the identifying of the one or more items comprises;

compiling counts, for each item of the set of items, of (1) a number of exposures of the content component and a number of click actions on the content component made by users who have purchased the item and (2) a number of exposures of the content component and a number of click actions on the content component made by users who have not purchased the item;

finding, for each item of the set of items, a difference between;

a value of a ratio of the counts of the number of exposures of the content component and the number of click actions on the content component made by the users who have purchased an item of the set of items; and

a value of a ratio of the counts of the number of exposures of the content component and the number of click actions on the content component made by the users who have not purchased the item of the set of items;

finding the one or more items associated with a largest difference among the differences associated with the set of items;

using the one or more items to differentiate the population of users into groups; and

selecting and placing different content components to different groups of users.

16. The targeting manager as recited in claim 15, wherein the monitoring comprises at least one of counting clicks on the content components, measuring click-through frequency, measuring click-through trends, tracking page views, measuring duration of page views, or measuring session duration.

17. The targeting manager as recited in claim 15, further comprising computer-executable instructions that, when executed on one or more processors, perform acts comprising constructing a tree structure that is associated with the placement of the content component, the tree structure having a root node that defines a universe of users and one or more subordinate nodes that segment the users into multiple groups according to the one or more items.

18. A server system comprising:

one or more processors;

a non-transitory memory, accessible by the one or more processors;

a targeting manager stored in the memory and executable on the one or more processors to select and place content components on pages to be electronically rendered to a population of users; and

wherein the targeting manager monitors user activity when the content components are exposed, learns which items purchased by the population of users are predictive of user interest in the content components, and adjusts selection and placement of the content components to target different groups of users segmented by the items purchased by the population of users, the targeting manager comprising;

a segment tree constructor to build tree structures for associated placements of content components, the tree structures having a root node that represents a universe of users to which the content components are exposed and subordinate nodes that represent different groups of users who are differentiated by the items purchased by the population of users, wherein the differentiating comprises;

finding, among the items purchased by the population of users, items correlated to a largest difference between;

values of a ratio of counts of a number of exposures and counts of a number of click actions of users who have purchased items among the items purchased by the population of users; and

values of a ratio of counts of a number of exposures and counts of a number of click actions of users who have not purchased the items among the items purchased by the population of users.

19. The server system as recited in claim 18, wherein the content components are selected from a group of content components comprising advertisements, news, search results, weather, articles, entertainment, financial data, traffic information, recommendations, images, photographs, video, or e-commerce items.

20. The server system as recited in claim 18, further comprising a store to hold the content components.