US20190279230A1

US20190279230A1 - Online Content Delivery Based on Information from Social Networks

Info

Publication number: US20190279230A1
Application number: US16/426,490
Authority: US
Inventors: Jeffrey B. Winner; Nicholas Galbreath
Original assignee: Facebook Inc
Current assignee: Meta Platforms Inc
Priority date: 2004-06-14
Filing date: 2019-05-30
Publication date: 2019-09-12
Also published as: US20050278443A1; US10373173B2

Abstract

In one embodiment, a social-networking system identifies a first plurality of users of the online social network, wherein the first plurality of users each share one or more user attributes, accesses, from a tracking database, tracking information of online activities of the first plurality of users with respect to a plurality of content objects, each content object having an associated stored value, calculates, for each content object, a first probability of interaction with the content object by the first plurality of users based on the accessed tracking information, calculates, for each content object, an expected value based on the associated stored value and the first probability of interaction, and sends, to a client device of a first user of the first plurality of users, one or more of the content objects based on the calculated expected values.

Description

PRIORITY

This application is a continuation under 35 U.S.C. § 120 of U.S. patent application Ser. No. 10/867,844, filed 14 Jun. 2004, which is incorporated herein by reference.

TECHNICAL FIELD

The present invention generally relates to online content management, and more specifically, to a system and method for managing online content delivery based on information from a person's social network.

BACKGROUND

Various statistical models have been used to make predictions about the future behavior and interests of users in an online environment. Online booksellers have used collaborative filtering techniques to recommend books that may be of interest to their customers based on the purchasing behavior of their other customers with similar interests and demographics.
Statistical models, including those based on collaborative filtering techniques, are, however, imperfect, and more accurate predictors are desired on many fronts. Sellers want them so that they can recommend more relevant products to their customers. Advertisers want them so that they can present more relevant ads to their audience. Web site operators want them so that they can deliver more relevant content to their visitors.

SUMMARY OF PARTICULAR EMBODIMENTS

The present invention provides a method for preparing or selecting relevant content for delivery to a member of a network. The selection is based, in part, on prior online activities of the other members of the network, and the closeness of the member's relationship with the other members of the network.
The relevant content may be an online ad that is selected from a number of candidate online ads based on click-through rates of groups within the online social network that are predefined with respect to the member or with respect to certain attributes. A predefined group contains one or more members of the network and may be any of the following: a group containing all members who are friends of the member; a group containing all members who are friends of friends of the member; a group containing all members who are friends of friends of friends of the member; a group containing members who have expressed a particular interest (e.g., music, cooking, travel, etc.), and a group containing members who fit a particular demographic (e.g., gender, age group, income level, ethnicity, etc.).
An online ad's revenue-generating potential may be considered in the selection process. For example, an online ad that has a high per-click revenue associated therewith is to be preferred in the selection process over an online ad that has a lower per-click revenue associated therewith, assuming that the probability of the member clicking on either online ad is the same. Certain online ads may be displayed without considering the per-click revenue associated therewith. For example, an invitation to purchase a particular item, e.g., a particular book, may be delivered to a member if the probability of the member clicking on that invitation is greater than a set threshold.

BRIEF DESCRIPTION OF THE DRAWINGS

So that the manner in which the above recited features of the present invention can be understood in detail, a more particular description of the invention briefly summarized above may be had by reference to embodiments, some of which are illustrated in the appended drawings. It is to be noted, however, that the appended drawings illustrate only typical embodiments of this invention and are therefore not to be considered limiting of its scope, for the invention may admit to other equally effective embodiments.

FIG. 1 is a diagram that conceptually represents the relationships between members in a social network;

FIG. 2 is a block diagram illustrating the system for managing an online social network;

FIG. 3 is a sample log for tracking the display and click history of a particular ad;

FIG. 4 is a sample adjacency list that is maintained by the graph servers used in the system for managing an online social network;

FIG. 5 is a sample table of click-through rates that have been computed for certain attributes;

FIG. 6 illustrates in a tabular form the method by which click probabilities for several ads are calculated for a member;

FIG. 7 illustrates in a tabular form the method by which an ad is selected for delivery to a member; and

FIG. 8 is a flow diagram that illustrates the method by which an ad is selected for delivery to a member.

DESCRIPTION OF EXAMPLE EMBODIMENTS

A social network is generally defined by the relationships among groups of individuals, and may include relationships ranging from casual acquaintances to close familial bonds. A social network may be represented using a graph structure. Each node of the graph corresponds to a member of the social network. Edges connecting two nodes represent a relationship between two individuals. In addition, the degree of separation between any two nodes is defined as the minimum number of hops required to traverse the graph from one node to the other. A degree of separation between two members is a measure of relatedness between the two members.
FIG. 1 is a graph representation of a social network centered on a given individual (ME). Other members of this social network include A-U whose position, relative to ME's, is referred to by the degree of separation between ME and each other member. Friends of ME, which includes A, B, and C, are separated from ME by one degree of separation (1 d/s). A friend of a friend of ME is separated from ME by 2 d/s. As shown, D, E, F and G are each separated from ME by 2 d/s. A friend of a friend of a friend of ME is separated from ME by 3 d/s. FIG. 1 depicts all nodes separated from ME by more than 3 degrees of separation as belonging to the category ALL.
Degrees of separation in a social network are defined relative to an individual. For example, in ME's social network, H and ME are separated by 2 d/s, whereas in G's social network, H and G are separated by only 1 d/s. Accordingly, each individual will have their own set of first, second and third degree relationships.
As those skilled in the art understand, an individual's social network may be extended to include nodes to an Nth degree of separation. As the number of degrees increases beyond three, however, the number of nodes typically grows at an explosive rate and quickly begins to mirror the ALL set.
FIG. 2 is a block diagram illustrating a system for managing an online social network. As shown, FIG. 2 illustrates a computer system 100, including an application server 200 and distributed graph servers 300. The computer system 100 is connected to a network 400, e.g., the Internet, and accessible over the network by a plurality of computers, which are collectively designated as 500.
The application server 200 manages a member database 210, a relationship database 220, a search database 230, an ad database, and a CTR database 250.
The member database 210 contains profile information for each of the members in the online social network managed by the computer system 100. The profile information may include, among other things: a unique member identifier, name, age (e.g., <30 years old or 30 years old and older), gender (male or female), location, hometown, a pointer to an image file, listing of interests and other attributes (e.g., music, cooking, travel), etc. The profile information also includes VISIBILITY and CONTACTABILITY settings, the uses of which are described in U.S. Pat. No. 8,010,458, filed May 26, 2004, which is incorporated by reference. The relationship database 220 stores information relating to the first degree relationships between members. In addition, the contents of the member database 210 are indexed and optimized for search, and stored in the search database 230. The member database 210, the relationship database 220, and the search database 230 are updated to reflect inputs of new member information and edits of existing member information that are made through the computers 500.
The ad database 240 contains ad information about banner ads, paid links, and specific product ads that are served by the application server 200. The ad information includes for each ad: a unique ad identifier, advertiser identifier, URL of the advertiser, click-through revenue, hyperlink to an image or text that contains the ad content, ad type indicator (e.g., 1=banner ad, 2=paid link, 3=product ad), file address of the log that contains the display and click history of the ad, and other typical information require to display the ad and to track the traffic on the ad. FIG. 3 is a sample log for tracking the display and click history of an ad having the ad identifier LINK1. The log shows the member ID of the member to whom the ad was displayed, whether there was a click or no click, and, if there was a click, the time stamp of that click.
The member database 210, the relationship database 220, and the search database 230 are updated to reflect inputs of new member information and edits of existing member information that are made through the computers 500. The ad database 240 is updated to reflect inputs of new ad information and edits of existing ad information that are made by a third party or the operator of the online social network through the computers 500 or a dedicated computer (not shown) connected to the computer system 100.
The application server 200 also manages the information exchange requests that it receives from the remote computers 500. The graph servers 300 receive a query from the application server 200, process the query and return the query results to the application server 200. The graph servers 300 manage a representation of the social network for all the members in the member database. The graph servers 300 have a dedicated memory device 310, such as a random access memory (RAM), in which an adjacency list that indicates all first degree relationships in the social network is stored.
A sample adjacency list that reflects the social network map of FIG. 1 is shown in FIG. 4. A list item is generated for each member and contains a member identifier for that member and member identifier(s) corresponding to friend(s) of that member. As an alternative to the adjacency list, an adjacency matrix or any other graph data structure may be used. The graph servers 300 and related components are described in U.S. Pat. No. 8,572,221, filed May 26, 2004, which is incorporated by reference.
The CTR database 250 stores, for each ad in the ad database 240, a set of click-through rates (CTRs) relating to groups within the online social network that are defined with respect to certain attributes, and to groups within the online social network that are defined with respect to a particular member. FIG. 5 shows CTRs relating to groups within the online social network that are defined with respect to certain attributes. These groups include: gender=male, gender=female, age<30, age=30+, interest=music, interest=cooking, interest=travel. Additional groups may be defined, e.g., location=94043, marital status=single, annual income<$100,000, annual income=$100,000+. The age groups and income groups may be defined in different ways as well.
FIG. 6 additionally shows CTRs relating to groups within the social network that are defined with respect to the member ME. These groups include: d/s=1 (members who are friends of ME), d/s=2 (members who are friends of friends of ME), and d/s=3 (members who are friends of friends of friends of ME). Additional groups may be defined, e.g., d/s=4. Combination groups may be defined, e.g., d/s=1 & interest=music (members who are friends of ME and whose expressed interest includes music).
A particular member may also predefine custom groups, e.g., C & D (group consisting of just member C and member D). Alternatively, the system may track the behavior of the members with respect to other members and define the custom groups for the members. As one example, a custom group for ME is populated with those members whose online activities in the past have influenced (i.e., have predicted well) the online activities of member ME.
The CTRs for a particular ad is computed using the display and click history for that ad. In general, CTR=(number of clicks)/(number of times displayed). The CTR for a group defined with respect to a particular attribute=(number of clicks by members who possess that particular attribute)/(number of times displayed to members who possess that particular attribute). The CTR for a group defined with respect to a particular member=(number of clicks by members who belong in that group)/(number of times displayed to members who belong in that group). For example, the CTR for member ME's d/s=1 group=(number of clicks by members who are friends of ME)/(number of times displayed to members who are friends of ME).
The click probability for a member with respect to any particular ad is estimated as the maximum of the CTRs computed for that member with respect to the particular ad. FIG. 6 shows that the click probability for member ME with respect to ad LINK1 is 0.077, with respect to ad LINK is 0.010, with respect to ad LINK3 is 0.011, and with respect to ad LINK4 is 0.041. FIG. 6 also shows the CTRs for the groups defined with respect to attributes that member ME does not possess (e.g., gender=female, age<30, interest=cooking, and interest=travel) noted as N/A. The CTRs associated with these groups are not considered when estimating member ME's click probability with respect to each of the ads.
After the click probabilities are estimated in the manner described above, expected revenues are derived from the click probabilities. FIG. 7 is a sample table showing revenues expected to be earned from different ads (LINK1, LINK2, LINK3, LINK4) as a result of member ME's predicted online behavior. The expected revenue for each of the ads is derived by multiplying the click probability computed for that ad with the per click revenue figure stored in the ad database 240 for that ad.
FIG. 8 is a flow diagram that illustrates the method by which an ad is selected for delivery to the member ME. In Step 810, the CTRs for a first set of groups defined with respect to particular attributes and a second set of groups defined with respect to the member ME are computed for each of the online ads stored in the ad database 240. In Step 820, for each of the online ads, the maximum CTR is selected as the estimated probability that member ME will click on that online ad. In Step 830, the expected revenues from displaying the online ads to member ME are estimated by multiplying the estimated probability for each of the online ads with the per-click revenue associated with that ad. In Step 840, the online ad with the highest estimated expected revenue is selected for delivery to member ME.
Ads are selected by the computer system 100 and delivered to the computers 500 for display at the computers 500. The ads may be delivered, for example, in response to a member logging on and accessing his or her home page. In one embodiment of the invention, the selection of one or more ads to be delivered to the member is based on the expected revenue of all the ads. For example, the system may be designed to deliver only the ad with the highest expected revenue, or alternatively, ads that are in the top N in expected revenue (where N>1). In another embodiment of the invention, the selection of one or more ads to be delivered to the member is based on the click probability. If the click probability is above a certain threshold, e.g., 0.50, the ad is to be delivered without regard to what the expected revenue is.
The computations of the CTRs, click probabilities, and expected revenues may be performed in real-time or off-line as a batch process. However, the batch process is preferred so that ads can be served more quickly to the members.
While particular embodiments according to the invention have been illustrated and described above, it will be clear that the invention can take a variety of forms and embodiments within the scope of the appended claims.

Claims

What is claimed is:

1. A method comprising, by one or more computing systems of an online social network:

identifying a first plurality of users of the online social network, wherein the first plurality of users each share one or more user attributes;

accessing, from a tracking database, tracking information of online activities of the first plurality of users with respect to a plurality of content objects, each content object having an associated stored value;

calculating, for each content object, a first probability of interaction with the content object by the first plurality of users based on the accessed tracking information;

calculating, for each content object, an expected value based on the associated stored value and the first probability of interaction; and

sending, to a client device of a first user of the first plurality of users, one or more of the content objects based on the calculated expected values.

2. The method of claim 1, wherein the user attributes comprise one or more of age, gender, location, hometown, interests, relationship status, or income.

3. The method of claim 1, wherein the first plurality of users have previously influenced the online activities of the first user.

4. The method of claim 1, wherein the first plurality of users are members of a predefined custom group.

5. The method of claim 1, wherein the first plurality of users are friends of the first user on the online social network.

6. The method of claim 1, further comprising:

accessing a social graph comprising a plurality of nodes and a plurality of edges connecting the nodes, wherein each node corresponds to a user of the online social network, wherein each edge between two nodes represents a single degree of separation between the two nodes, and wherein a degree of separation between any two nodes is a minimum number of edges required to traverse the social graph data from one user node to the other.

7. The method of claim 6, wherein the node corresponding to the first user is within a threshold degree of separation of the nodes corresponding to each other user of the first plurality of users.

8. The method of claim 1, wherein the one or more content objects sent to the client device of the first user comprise content objects having expected values greater than a threshold expected value.

9. The method of claim 8, wherein the one or more content objects sent to the client device of the first user further comprise content objects having expected values below a threshold expected value and first probabilities of interaction greater than a threshold probability of interaction.

10. The method of claim 1, further comprising:

identifying a second plurality of users of the online social network, wherein the second plurality of users each share one or more user attributes;

accessing, from a tracking database, tracking information of online activities of the second plurality of users with respect to the plurality of content objects; and

calculating, for each content object, a second probability of interaction with the content object by the second plurality of users based on the accessed tracking information.

11. The method of claim 10, wherein calculating the expected value for each content object is based on the associated stored value and the greater of the first probability of interaction and the second probability of interaction.

12. The method of claim 11, wherein the one or more content objects sent to the client device of the first user comprise content objects having expected values greater than a threshold value.

13. The method of claim 12, wherein the one or more content objects sent to the client device of the first user further comprise content objects having expected values below a threshold value and a first probability of interaction or a second probability of interaction greater than a threshold probability of interaction.

14. The method of claim 1, wherein the one or more content objects are sent to the client device of the first user in response to a request to access a home page of the online social network associated with the first user.

15. The method of claim 1, wherein one or more of the content objects sent to the client device of the first user comprises a hyperlink associated with the respective content object.

16. The method of claim 15, wherein each hyperlink associated with the content object may be selected by the first user to access third-party content associated with the content object on a third-party website external to the online social network.

17. The method of claim 15, wherein the one or more content objects sent to the client device of the first user further comprises one or more of:

an image associated with the content object;

a text associated with the content object;

a content type indicator; or

a content object identifier.

18. The method of claim 1, further comprising:

receiving, from the client device of the first user, an indication of one or more interactions with one or more of the content objects sent to the client device of the first user; and

updating the tracking information of the first user from the tracking database based on the interactions.

19. A system comprising: one or more processors; and a memory coupled to the processors comprising instructions executable by the processors, the processors operable when executing the instructions to:

identify a first plurality of users of the online social network, wherein the first plurality of users each share one or more user attributes;

access, from a tracking database, tracking information of online activities of the first plurality of users with respect to a plurality of content objects, each content object having an associated stored value;

calculate, for each content object, a first probability of interaction with the content object by the first plurality of users based on the accessed tracking information;

calculate, for each content object, an expected value based on the associated stored value and the first probability of interaction; and

send, to a client device of a first user of the first plurality of users, one or more of the content objects based on the calculated expected values.

20. One or more computer-readable non-transitory storage media embodying software that is operable when executed to: