EP2430605A2

EP2430605A2 - Targeting advertisements to videos predicted to develop a large audience

Info

Publication number: EP2430605A2
Application number: EP10770356A
Authority: EP
Inventors: Nandy Palash; Yu-To Chen; Nadine Harik; Shivakumar Rajaraman; Pius Fischer
Original assignee: Google LLC
Current assignee: Google LLC
Priority date: 2009-04-29
Filing date: 2010-04-29
Publication date: 2012-03-21
Also published as: WO2010127150A2; JP5544010B2; WO2010127150A3; EP2430605A4; JP2012525654A

Abstract

Advertisers target their advertisements to videos that, although not yet popular, are predicted to become popular in the near future. A video's popularity is measured by a number of users that have added the video to their list of favorite videos. A statistical model is formed to predict a video's favoriting velocity by analyzing a set of videos and related inputs using techniques such as classification and regression trees and logistic regression. The model is then applied to newer videos to determine a likelihood that the video will go viral. An advertising engine associates videos having a high likelihood of going viral with advertising creatives provided by advertisers for that purpose. A premium advertising fee is charged to advertisers in some embodiments in order to be associated with viral videos.

Description

TARGETING ADVERTISEMENTS TO VIDEOS PREDICTED TO DEVELOP A LARGE

AUDIENCE

Inventors:

Palash Nandy

Yu-To Chen

Nadine Harik

Shivakumar Rajaraman

Pius Fischer

BACKGROUND Field

[0001] The present invention relates generally to providing video on the Internet. In particular, embodiments of the present invention are directed toward predicting when a video will develop a large audience, and targeting advertisements to that video.

Description of Background Art

[0002] Various web sites exist that provide hosted video content to viewers. One such site is the YouTube site, provided by Google Inc. of Mountain View, California. Typically, videos are supplied to the video hosting web site by content providers, and are then made available for viewing by the public at large.

[0003] Some video hosting web sites allow advertisers to place advertisements on the web pages (also called watch pages) on which the video content is displayed. The advertiser pays the video hosting web site owner according to one or more revenue models, including, for example, a cost-per-impression (CPM) or cost-per-click (CPC) model.

[0004] Advertisers understandably would like their advertisements to be seen by a wide audience of viewers potentially interested in their products and services. Typically, advertisers can narrow the range of videos on which their advertising will appear by specifying keywords or other video metadata that they consider to be relevant. For example, a car company may specify that its advertisements should be shown alongside a video that has as its keywords or in its metadata the terms "car", "Ford", "Toyota", or

"truck". This helps advertisers to spend their money more efficiently by targeting their advertisements.

[0005] In addition to targeting advertisements to viewers of topical videos, advertisers would benefit from other ways to identify videos that maximize their return on advertising dollars. SUMMARY

[0006] Embodiments of the present invention enable advertisers to target their advertisements to videos that, although not yet popular, are predicted to become popular in the near future. According to various embodiments, a video's popularity is measured by a number of users that have added the video to their list of favorite videos, often referred to as "favoriting" a video. A statistical model is formed to predict a video's favoriting velocity by analyzing a set of videos and related inputs using techniques such as classification and regression trees and logistic regression. The model is then applied to newer videos to determine a likelihood that the video will go viral. In some embodiments, a different model is used for different geographic regions, such as continents, countries, states, and cities. An advertising engine associates videos having a high likelihood of going viral with advertising creatives provided by advertisers for that purpose. A premium advertising fee is charged to advertisers in some embodiments in order to be associated with viral videos.

BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Fig. 1 is a block diagram of the overall architecture of an embodiment of the present invention.

[0008] Fig. 2 illustrates a graph of the popularity of a viral video in accordance with an embodiment of the present invention.

[0009] Fig. 3 illustrates a method for building a statistical model to predict video popularity in accordance with an embodiment of the present invention.

[0010] Fig. 4 illustrates a method for providing advertisements in association with viral videos in accordance with an embodiment of the present invention.

[0011] The figures depict preferred embodiments of the present invention for purposes of illustration only. One skilled in the art will readily recognize from the following discussion that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles of the invention described herein.

DETAILED DESCRIPTION

[0012] One way to increase advertising effectiveness on a video hosting web site is to know in advance which of the millions of available videos is likely to become very popular with viewers in the near term, referred to as "going viral". These "viral videos" tend to exhibit a geometric increase in popularity, frequently over a very short period. Some of the videos will remain popular for an extended amount of time; others lose viewership rapidly, like a passing fad. By identifying videos that are likely to become viral in the near future, a video hosting web site can offer advertisers an opportunity to place their advertisements on those videos at the beginning of the viral process, thus enabling the advertiser to capture the attention of viewers during the video's increase in popularity. Advertisers may also obtain important but less tangible benefit from targeting viral videos, for example by developing their brand reputation for being associated with the latest trends in popular culture. [0013] Architecture

[0014] Fig. 1 illustrates a video hosting web site 100 adapted to identify videos predicted to develop a large audience and to target advertisements to those videos in accordance with an embodiment of the present invention.

[0015] The video hosting website 100 illustrated in Fig. 1 includes a front end interface 102, a video serving module 104, a video search module 106, an upload server 108, popularity engine 110, advertising engine 112, user database 114, video database 116, and advertisement database 118. Fig. 1 also illustrates a client 120, content provider 122, advertiser 124 and network 126. Other conventional features, such as firewalls, load balancers, authentication servers, application servers, failover servers, site management tools, and so forth are not shown so as to more clearly illustrate the features of the video hosting website 100. An example of a suitable website 100 is the YOUTUBE™ website, found at www.youtube.com. Other video hosting sites are known as well, and can be adapted to operate according to the teachings disclosed herein. The illustrated components of the video hosting website 100 can be implemented as single pieces of software or hardware or as multiple pieces of software or hardware. In general, functions described in one embodiment as being performed by one component, can also be performed by other components in other embodiments, or by a combination of components. Furthermore, functions described in one embodiment as being performed by components of the video hosting website 100 can also be performed by one or more clients 120 in other embodiments if appropriate.

[0016] The described servers can be implemented as server programs executing on server-class computers comprising a CPU, memory, network interface, peripheral interfaces, and other well known components. The functionality implemented by any of the elements can be provided from computer program products that are stored in computer-readable storage mediums (e.g., RAM, hard disk, or optical/magnetic media). [0017] A client 120 executes a browser or other application suitable for viewing video and connects to the video hosting site 100 through front end interface 102. Front end interface 102 may be, for example, a web server providing a user interface to client 120. Network 126, through which client 120 and video hosting site 100 communicate, is typically the Internet, but can also be any network, including but not limited to any combination of a LAN, a MAN, a WAN, a mobile, wired or wireless network, a private network, or a virtual private network. While only one client 120 is shown, a very large numbers (e.g., millions) of clients, or as many as can be supported by the hardware and software implementation, can be in communication with the video hosting website 100 at any time. The client 120 may include a variety of different computing devices that support playback of video. Examples of current client devices 120 are personal computers, digital assistants, personal digital assistants, cellular phones, smart phones, mobile phones, and laptop computers. [0018] Client 120 views videos from video hosting site 100 using software suited for that purpose. Typically that software is a browser such as Microsoft Internet Explorer, Netscape Navigator, Mozilla Firefox, Apple Safari, etc. Alternatively, client-specific or site-specific software may be used. The browser can also include a video player (e.g., Flash™ from Adobe Systems, Inc.), or any other player adapted for the video file formats used in the video hosting website 100. A user can access a set of videos from the video hosting website 100 by browsing a catalog of videos, conducting searches on keywords, reviewing play lists from other users or the system administrator (e.g., collections of videos forming channels), viewing videos associated with particular user group (e.g., communities), or by directly accessing the video resource, e.g., via a URL.

[0019] Content provider 122 provides videos to the video hosting website 100 via physical media or via the network 126. While only one content provider 122 is shown, any number of content providers are supported and can be in communication with the video hosting website 100 at any time. Content providers 122 and clients 120 (i.e., consumers of content) are illustrated as separate entities in Fig. 1. In practice, the same person or other entity can be both a client and a content provider. [0020] Video serving module 104 retrieves videos from video database 116 and makes them available to client 120, for example via front end interface 102. Video search module 106 enables a client 120 to search for videos meeting one or more specified criteria, or which are related to other videos the client has viewed, promoted by the video hosting site, and the like.

[0021] Upload server 108 receives videos from content providers 122 and stores them in video database 116. Upload server 108 typically provides additional functionality such as transcoding the uploaded video from one file type to another, obtaining metadata about the video from the content provider 122, creating thumbnails of the video, performing scene detection, and various other functions required as part of services provided by video hosting site 100.

[0022] Video database 116 is used to store uploaded videos. Database 116 stores video content and associated metadata provided by the content provider 122, or by the operator of video hosting website 100, or by third parties. The videos have metadata associated with each file such as a video ID, an ID of the user who provided the video, artist, video title, label, genre, time length, and optionally geo-restrictions that can be used for data collection or content blocking on a geographic basis. Additionally, video database 116 maintains a record of ratings received from users for each video, and a list of users who have added each video to their lists of favorite videos.

[0023] User database 114 maintains a record of all registered users of video hosting website 100, including login credentials, geographic location, demographic information, videos uploaded, videos watched, videos rated, videos listed as favorites, videos part of a playlist, user subscriptions, and other users who subscribe to the user. Additional information can also be stored as appropriate for a particular implementation. [0024] Advertising engine 112 performs functions related to providing advertising in conjunction with videos on video hosting site 100. Advertising engine 112 determines which videos should be monetized— that is, have advertisements displayed in association with the video — and which advertisements should be displayed along with particular videos, including videos predicted to go viral. Advertising creatives are stored in advertisement database 118. Advertisement database 118 also stores information related to each advertiser, such as account information and credentials, and campaign preferences for any advertising campaigns in which the advertiser 124 is involved.

[0025] Popularity engine 110 analyzes data associated with videos to determine their current and predicted popularity. Using a multitude of metrics such as view counts, ratings, age of the video, number of users who have made the video a favorite, number of different users who have rated the video and others, as well as metadata associated with the video, popularity engine 110 determines a likelihood that a video is about to enjoy a rapid increase in popularity.

[0026] Determining Popularity

[0027] Fig. 2 depicts a graph 200 that shows, for a particular video x, the number of users who have added x to their favorites list on each of a plurality of days. Graph 200 is merely illustrative— the particular graph for any individual video will obviously differ according to the relevant data. Graph 200 illustrates a logistic curve, typical for a viral video. The popularity of the video starts out small, then begins to increase. As the video goes viral, the slope becomes very large as more and more people add the video to their favorites each day. Eventually, the video reaches saturation, and the slope levels off. [0028] Accordingly, the number of users that have added the video to their favorites (which we refer to as "favoriting" the video) is one measure of the popularity of the video. By predicting a rapid increase in the velocity of favoriting of the video, popularity engine 110 thereby can determine that a video is about to go viral. Advertising engine 112 then uses the predicted increase in velocity to associate ads with the videos. [0029] Fig. 3 illustrates a method for building a statistical model that predicts the popularity of a video in accordance with an embodiment of the present invention. [0030] To begin, popularity engine 110 selects 302 an initial set of videos randomly from video database 116 and uses classification and regression trees (CART) to hierarchically partition 304 the space to find a highly-concentrated region of videos being favorited with high velocity. Logistic regression 306 is then used to build a model that can be applied 308 to all videos. A large number of video-related signals can be used as inputs to the CART analysis, including view counts and decay rates of view counts, ratings, age of the video, number of users who have already made the video a favorite, number of different users who have rated the video, number of comments received for the video, search queries resulting in video playback, identity of the content provider, number and identity of sites linking to the video, number of users sharing the video, and number of users posting the video to other sites such as social networking sites or bookmarking sites. Velocity and acceleration of each of these signals also act as inputs in various embodiments.

[0031] In one embodiment, separate models are built for different geographic regions.

For example, different models may be built for videos uploaded from North America, South

America, Europe, Asia, Oceana and Africa. This allows for varying geographical tastes to result in different decay rates in each model. The geographic source of a video is identified in one embodiment using information associated with the account of the user who uploaded the video to hosting site 100.

[0032] In one embodiment, videos in video database 116 are associated with a flag indicating controversial or racy content. Because such videos may have different viewing characteristics, in one embodiment they are excluded from the set of videos used to form the model.

[0033] Once the model has been constructed as described above, it is applied to each active video to form a prediction. Where multiple models are used, each model can be separately applied to videos to obtain relevant scores for that model. Popularity engine 110 stores 310 the prediction for each video analyzed in video database 116. In one embodiment, a prediction is a numeric value within a range such as 0 to 1, and represents a likelihood that the video will go viral.

[0034] In one embodiment, videos likely to go viral are identified periodically, for example each hour or each day. In one embodiment, the top n videos in terms of a score found by popularity engine 110 are selected as videos most likely to go viral. Alternatively, any video with a likelihood score of more than some threshold amount can be said to be likely to go viral. The value of n and the value of the threshold score can be chosen by the implementer.

[0035] Advertising

[0036] As described above, video hosting web site 100 provides advertisers 124 with the opportunity to place their advertisements on web pages associated with particular videos or types of videos. An advertiser typically runs one or more advertising campaigns, and each campaign may have one or more creatives associated with it. Advertisers specify criteria that are then used by advertising engine 112 to pair advertisements with videos. For example, a car manufacturer may specify that it wants its car-related advertisements to appear next to videos that are related to cars. Advertising engine 112 determines which videos meet the criteria specified by the advertiser 124 by examining information associated with the video, such as metadata including its title, description, and keywords; and additionally factors such as comments that have been left about the video. [0037] Advertising engine 112 offers advertisers the opportunity to additionally target their ads based on the velocity of a video's popularity. That is, an advertiser can specify that it would like its creatives to be associated with videos predicted to have an imminent increase in viewership.

[0038] Fig. 4 illustrates a method by which advertising engine 112 enables advertisers 124 to have their advertisements displayed on videos likely to go viral. Advertising engine 112 first identifies 402 an advertiser's campaign that the advertiser has indicated an interest in associating with a viral video. Next, advertising engine 112 locates 404 videos in video database 116 having at least a threshold probability of going viral as determined by popularity engine 110. In one embodiment, the threshold probability is predetermined by video hosting site 100; alternatively, it is specified by advertiser 124 as part of the campaign. The list of likely viral videos is then filtered 408 to take into account the advertiser's preferences— for example, keywords, demographic targets, etc. One or more of the remaining videos are then associated 410 with the advertiser's campaign. [0039] In other embodiments, the association of advertisements with likely viral videos is performed in alternative orders— for example, a set of likely viral videos may be determined first, and then matched with interested advertisers and their campaigns. The order of the matching and associating steps is chosen according to implementation details without departing from the concept described here.

[0040] In various embodiments, popularity engine 110 evaluates new and existing videos with varying frequency —for example, every hour, every two hours, twice a day, daily, weekly, etc., according to the preference of video hosting site 100. The association of advertisements with pre-viral and viral videos is also reevaluated periodically and advertising engine 112 reallocates advertisements to videos responsive to changes in the existing or predicted velocity of videos on the site. [0041] In addition to giving advertisers the opportunity to ride a wave of popularity by having their advertisements seen with the latest popular videos, such a method provides an additional revenue source for video hosting website 100, as advertising engine 112 in one embodiment imposes a premium on the advertising charge to advertiser 124 in exchange for having its ads associated with the soon-to-be-popular videos.

[0042] The present invention has been described in particular detail with respect to a limited number of embodiments. Those of skill in the art will appreciate that the invention may additionally be practiced in other embodiments. First, the particular naming of the components, capitalization of terms, the attributes, data structures, or any other programming or structural aspect is not mandatory or significant, and the mechanisms that implement the invention or its features may have different names, formats, or protocols. Further, the system may be implemented via a combination of hardware and software, as described, or entirely in hardware elements. Also, the particular division of functionality between the various system components described herein is merely exemplary, and not mandatory; functions performed by a single system component may instead be performed by multiple components, and functions performed by multiple components may instead performed by a single component. For example, the particular functions of the popularity engine 110 may be provided in many or one module.

[0043] Some portions of the above description present the feature of the present invention in terms of algorithms and symbolic representations of operations on information. These algorithmic descriptions and representations are the means used by those skilled in the art to most effectively convey the substance of their work to others skilled in the art. These operations, while described functionally or logically, are understood to be implemented by computer programs. Furthermore, it has also proven convenient at times, to refer to these arrangements of operations as modules or code devices, without loss of generality.

[0044] It should be borne in mind, however, that all of these and similar terms are to be associated with the appropriate physical quantities and are merely convenient labels applied to these quantities. Unless specifically stated otherwise as apparent from the present discussion, it is appreciated that throughout the description, discussions utilizing terms such as "processing" or "computing" or "calculating" or "determining" or "displaying" or the like, refer to the action and processes of a computer system, or similar electronic computing device, that manipulates and transforms data represented as physical (electronic) quantities within the computer system memories or registers or other such information storage, transmission or display devices.

[0045] Certain aspects of the present invention include process steps and instructions described herein in the form of an algorithm. It should be noted that the process steps and instructions of the present invention could be embodied in software, firmware or hardware, and when embodied in software, could be downloaded to reside on and be operated from different platforms used by real time network operating systems.

[0046] The present invention also relates to an apparatus for performing the operations herein. This apparatus may be specially constructed for the required purposes, or it may comprise a general-purpose computer selectively activated or reconfigured by a computer program stored in the computer. Such a computer program may be stored in a computer readable storage medium, such as, but is not limited to, any type of disk including floppy disks, optical disks, CD-ROMs, magnetic-optical disks, read-only memories (ROMs), random access memories (RAMs), EPROMs, EEPROMs, magnetic or optical cards, application specific integrated circuits (ASICs), or any type of media suitable for storing electronic instructions, and each coupled to a computer system bus. Furthermore, the computers referred to in the specification may include a single processor or may be architectures employing multiple processor designs for increased computing capability. [0047] The algorithms and displays presented herein are not inherently related to any particular computer or other apparatus. Various general-purpose systems may also be used with programs in accordance with the teachings herein, or it may prove convenient to construct more specialized apparatus to perform the required method steps. The required structure for a variety of these systems will appear from the description above. In addition, the present invention is not described with reference to any particular programming language. It is appreciated that a variety of programming languages may be used to implement the teachings of the present invention as described herein, and any references to specific languages are provided for disclosure of enablement and best mode of the present invention. [0048] Finally, it should be noted that the language used in the specification has been principally selected for readability and instructional purposes, and may not have been selected to delineate or circumscribe the inventive subject matter. Accordingly, the disclosure of the present invention is intended to be illustrative, but not limiting, of the scope of the invention. [0049] We claim:

Claims

1. A method for providing advertisements on a video hosting web site, the method comprising: receiving a video at the video hosting site; receiving requests for the video over a network at a first rate; providing the video in response to the requests; determining a likelihood of receiving requests for the video at a second rate; and responsive to the likelihood exceeding a threshold likelihood, associating an advertisement with the video, and providing the video and associated advertisement in responsive to additional requests for the video.

2. The method of claim 1, wherein the first rate is an average rate at which requests are received for the video during a first measurement period.

3. The method of claim 2, wherein the second rate is an average rate at which requests are received for the video during a second measurement period.

4. The method of claim 1 wherein determining the likelihood of receiving requests for the video at the second rate further comprises: applying a predictive model using input signals associated with the video; and determining the likelihood based on the output of the predictive model.

5. The method of claim 4 wherein the predictive model uses a logistic regression.

6. The method of claim 4 wherein the predictive model uses a classification and regression tree.

7. The method of claim 4 wherein the input signals include a number of users of the video hosting site that have added the video to a favorites list.

8. The method of claim 1 wherein the input signals include a number of times the video has been viewed by users of the video hosting site.

9. The method of claim 1 wherein the input signals include a number of times the video has been rated by users of the video hosting site.

10. The method of claim 1 wherein the input signals include a number of comments related to the video posted by users of the video hosting site.

11. The method of claim 1 wherein the input signals include a number of search queries resulting in playback of the video.

12. The method of claim 1 wherein the first rate and the second rate form points on a logistic curve.

13. The method of claim 1 wherein the associated advertisement is received from an advertiser and is associated with advertiser preference data.

14. The method of claim 13 wherein the advertiser preference data specifies the threshold likelihood.

15. The method of claim 13 wherein the advertiser preference data specifies an amount the advertiser agrees to pay to the video hosting site in exchange for providing the associated advertisement with the video.

16. A method of associating an advertisement with a video on a video hosting website, the method comprising; receiving an advertisement; receiving a video; receiving requests for the video over a network; providing the video over the network in response to each request; determining metrics associated with the video; determining based on the metrics a likelihood that the video will increase in popularity relative to other videos on the video hosting website in a predetermined time frame; responsive to the likelihood exceeding a threshold, associating the advertisement with the video; and displaying the advertisement in conjunction with the video.