US20060230053A1 - Consumer profiling and advertisement selection system - Google Patents

Abstract

A consumer profiling and advertisement selection system is presented in which consumers or subscribers can be characterized based on their purchase or viewing habits. The result of this process is a consumer characterization vector describing the probabilistic demographics and product preferences of the subscriber or viewer. Advertisement characterization vectors describing an actual or hypothetical market for a product or desired viewing audience can be determined. The ad characteristics including an ad demographic vector, an ad product category and an ad product preference vector is transmitted along with a consumer ID. The consumer ID is used to retrieve a consumer characterization vector which is correlated with the ad characterization vector to determine the suitability of the advertisement to the consumer. A price for displaying the advertisement can be determined based on the results of the correlation of the ad characteristics with the consumer characterization vector.

Description

CROSS REFERENCE TO RELATED APPLICATION
• This application is a continuation of U.S. patent application Ser. No. 09/807,887, filed Apr. 19, 2001, and entitled Consumer Profiling and Advertisement Selection System, the entire disclosure of which is incorporated herein by reference.
• This application claims the benefit of International Application No. PCT/US99/28628, filed Dec. 2, 1999, entitled Consumer Profiling and Advertisement Selection System, which claims the benefit of co-pending U.S. patent application Ser. No. 09/204,888, filed Dec. 3, 1998, entitled Subscriber Characterization System; U.S. patent application Ser. No. 09/268,526, filed Mar. 12, 1999, entitled Advertisement Selection System Supporting Discretionary Target Market Characteristics, now U.S. Pat. No. 6,216,129; and U.S. patent application Ser. No. 09/268,519, filed Mar. 12, 1999, entitled Consumer Profiling System, now U.S. Pat. No. 6,298,348.
BACKGROUND OF THE INVENTION
• The advent of the Internet has resulted in the ability to communicate data across the globe instantaneously, and will allow for numerous new applications which enhance consumer's lives. One of the enhancements which can occur is the ability for the consumer to receive advertising which is relevant to their lifestyle, rather than a stream of ads determined by the program they are watching. Such “targeted ads” can potentially reduce the amount of unwanted information which consumers receive in the mail, during television programs, and when using the Internet. Examples of editorial targeting can be found on the World Wide Web, where banners are delivered based on the page content. The product literature from DoubleClick, “Dynamic Advertising Reporting and Targeting (DART),” printed from the World Wide Web site http://www.doubleclick.net/dart on Jun. 19, 1998 discloses DoubleClick's advertising solution for matching advertiser's selected targeted profiles with individual user profiles and deliver an appropriate banner. The user and advertisements are matched based on geographic location or keywords on the page content. The product literature from Imgis, “Ad Force,” printed from the World Wide Web site http://www.starpt.com/core on Jun. 30, 1998 discloses an ad management system for targeting users and delivering advertisements to them. Users are targeted based on the type of content they are viewing or by keywords.
• From an advertiser's perspective the ability to target ads can be beneficial since they have some confidence that their ad will at least be determined relevant by the consumer, and therefore will not be found annoying because it is not applicable to their lifestyle. Different systems for matching a consumer profile to an advertisement have been proposed such as the U.S. Pat. No. 5,774,170, which discloses a system for delivering targeted advertisement to consumers. In this system, a set of advertisements is tagged with commercial identifier (CID) and, from the existing marketing database, a list of prospective viewers is also identified with CID. The commercials are displayed to the consumers when the CIDs match.
• Other systems propose methods for delivering programming tailored to subscribers' profile. U.S. Pat. No. 5,446,919 discloses a communication system capable of targeting a demographically or psychographically defined audience. Demographic and psychographic information about audience member are downloaded and stored in the audience member receiver. Media messages are transmitted to audience member along with a selection profile command, which details the demographic/psychographic profile of audience members that are to receive each media message. Audience members which fall within a group identified by the selection profile command are presented with the media message.
• U.S. Pat. No. 5,223,924 discloses a system and method for automatically correlating user preferences with a TV program information database. The system includes a processor that performs “free text” search techniques to correlate the downloaded TV program information with the viewer's preferences. U.S. Pat. No. 5,410,344 discloses a method for selecting audiovideo programs based on viewers' preferences, wherein each of the audiovideo programs has a plurality of programs attributes and a corresponding content code representing the program attributes. The method comprises the steps of storing a viewer preference file, which includes attributes ratings, which represents the degree of impact of the programs attributes on the viewer and, in response to the comparison of viewer preference file with the program content codes, a program is selected for presentation to the viewer.
• In order to determine the applicability of an advertisement to a consumer, it is necessary to know something about their lifestyle, and in particular to understand their demographics (age, household size and income). In some instances, it is useful to know their particular purchasing habits. Purchasing habits are being used by E-commerce to profile their visitors. As an example, the product literature from Aptex software Inc., “SelectCast for Commerce Servers,” printed from the World Wide Web site http://www.aptex.com/products-selectcast-commerce.htm on Jun. 30, 1998 discloses the product SelectCast for Commerce Servers. The product personalizes online shopping based on observed user behavior. User interests are learned based on the content they browse, the promotions they click and the products they purchase.
• Knowledge of the purchasing habits of a consumer can be beneficial to a product vendor in the sense that a vendor of soups would like to know which consumers are buying their competitor's soup, so that they can target ads at those consumers in an effort to convince them to switch brands. That vendor will probably not want to target loyal customers, although for a new product introduction the strategy may be to convince loyal customers to try the new product. In both cases it is extremely useful for the vendor to be able to determine what brand of product the consumer presently purchases.
• There are several difficulties associated with the collection, processing, and storage of consumer data. First, collecting consumer data and determining the demographic parameters of the consumer can be difficult. Surveys can be performed, and in some instances the consumer will willingly give access to normally private data including family size, age of family members, and household income. In such circumstances there generally needs to be an agreement with the consumer regarding how the data will be used. If the consumer does not provide this data directly, the information must be “mined” from various pieces of information which are gathered about the consumer, typically from specific purchases.
• A relatively intrusive method for collecting consumer information is described in U.S. Pat. No. 4,546,382, which discloses a television and market research data collection system and method. A data collection unit containing a memory, stores data as to which of the plurality of TV modes are in use, which TV channel is being viewed as well as input from a suitable optical scanning device for collecting consumer product purchases.
• Once data is collected, usually from one source, some type of processing can be performed to determine a particular aspect of the consumer's life. As an example, processing can be performed on credit data to determine which consumers are a good credit risk and have recently applied for credit. The resulting list of consumers can be solicited, typically by direct mail. Although information such as credit history is stored on multiple databases, storage of other information such as the specifics of grocery purchases is not typically performed. Even if each individual's detailed list of grocery purchases was recorded, the information would be of little use since it would amount to nothing more than unprocessed shopping lists.
• Privacy concerns are also an important factor in using consumer purchase information. Consumers will generally find it desirable that advertisements and other information is matched with their interests, but will not allow indiscriminate access to their demographic profile and purchase records.
• The Internet has spawned the concept of “negatively priced information” in which consumers can be paid to receive advertising. Paying consumers to watch advertisements can be accomplished interactively over the Internet, with the consumer acknowledging that they will watch an advertisement for a particular price. Previously proposed schemes such as that described in U.S. Pat. No. 5,794,210, entitled “Attention Brokerage,” of which A. Nathaniel Goldhaber and Gary Fitts are the inventors, describe such a system, in which the consumer is presented with a list of advertisements and their corresponding payments. The consumer chooses from the list and is compensated for viewing the advertisement. The system uses also software agents representing consumers to match the consumer interest profiles with advertisements. The matching is done using “relevance indexing” which is based on hierarchical tree structures. The system requires real-time interactivity in that the viewer must select the advertisement from the list of choices presented.
• The ability to place ads to consumers and compensate them for viewing the advertisements opens many possibilities for new models of advertising. However, it is important to understand the demographics and product preferences of the consumer in order to be able to determine if an advertisement is appropriate.
• Although it is possible to collect statistical information regarding consumers of particular products and compare those profiles against individual demographic data points of consumers, such a methodology only allows for selection of potential consumers based on the demographics of existing customers of the same or similar products.
• U.S. Pat. No. 5,515,098, entitled “System and method for selectively distributing commercial messages over a communications network,” of which John B. Carles is the inventor, describes a method in which target household data of actual customers of a product are compared against subscriber household data to determine the applicability of a commercial to a household. Target households for a product or service are characterized by comparing or correlating the profile of the customer household to the profile of all households. A rating is established for each household for each category of goods/services. The households within a predefined percentile of subscribers, as defined by the rating, are targeted by the advertiser of the product or service.
• It will also frequently be desirable to target an advertisement to a market having discretionary characteristics and to obtain a measure of the correlation of these discretionary features with probabilistic or deterministic data of the consumer/subscriber, rather than being forced to rely on the characteristics of existing consumers of a product. Such correlation should be possible based both on demographic characteristics and product preferences.
• Another previously proposed system, described in U.S. Pat. No. 5,724,521, entitled “Method and apparatus for providing electronic advertisements to end users in a consumer best-fit pricing manner,” of which R. Dedrick is the inventor, utilizes a consumer scale as the mechanism to determine to which group an advertisement is intended. A consumer scale matching process compares the set of characteristics stored in a user profile database to a consumer scale associated with the electronic advertisement. The fee charged to the advertiser is determined by where the set of characteristics fall on the consumer scale. Such a system requires specification of numerous parameters and weighting factors, and requires access to specific and non-statistical personal profile information.
• For the foregoing reasons, there is a need for a consumer profiling system which can profile the consumer, provide access to the consumer profile in a secure manner, and return a measurement of the potential applicability of an advertisement. There is also a need for an advertisement selection system which can match an advertisement with discretionary target market characteristics, and which can do so in a manner which protects the privacy of the consumer data and characterizations.
SUMMARY OF THE INVENTION
• The present invention supports the receipt of consumer purchase information with which consumer characterization vectors are updated based on product characterization information. The consumer characterization vectors include a consumer demographic vector which provides a probabilistic measure of the demographics of the consumer, and a product preference vector which describes which products the consumer has typically purchased in the past, and therefore is likely to purchase in the future. The product characterization information includes vector information which represents probabilistic determinations of the demographics of purchasers of an item, heuristic rules which can be applied to probabilistically describe the demographics of the consumer based on that purchase, and a vector representation of the purchase itself.
• In a preferred embodiment a computer-readable detailed purchase record is received, along with a unique consumer identifier. A demographic characterization vector corresponding to the consumer can be retrieved. In the event that there is no existing demographic characterization vector for that consumer, a new demographic characterization vector can be created. In a preferred embodiment the new demographic characterization vector contains no information. A set of heuristic rules is retrieved and contains a probabilistic measure of the demographic characteristics of a typical purchaser of an item. A new demographic characterization vector is calculated based on the purchase, the existing demographic characterization vector, and the heuristic rules.
• In a preferred embodiment the calculation of the demographic characterization vector is performed by calculating a weighted average of a product demographics vector and the existing demographic characterization vector. A weighting factor is used in which the weighting factor is determined based on the ratio of the current product purchase amount to a cumulative product purchase amount. The cumulative product purchase amount can be measured as the amount spent on a particular category of items (e.g. groceries, clothes, accessories) over a given period of time such as one month or one year.
• In a preferred embodiment the heuristic rules are in the form of a product demographics vector which states the demographics of known purchasers of an item. Each product can have an associated product demographics vector.
• The present invention can be used to develop product preference descriptions of consumers which describe the brand and size product that they purchase, and which provide a probabilistic interpretation of the products they are likely to buy in the future. The product preference description can be generated by creating a weighted average of an existing product preference vector describing the consumer's historical product preferences (type of product, brand, and size) and the characteristics of recent purchases.
• The present invention can be realized as a data processing system or computer program which processes consumer purchase records and updates their demographic and product preference profiles based on the use of product characterization information. The data processing system can also be used to receive information regarding an advertisement and to perform a correlation between the advertisement and the consumer's demographic and product preferences.
• The present invention can be realized as software resident on one or more computers. The system can be realized on an individual computer which receives information regarding consumer purchases, or can be realized on a network of computers in which portions of the system are resident on different computers.
• One advantage of the present invention is that it allows consumer profiles to be updated automatically based on their purchases, and forms a description of the consumer including demographic characteristics and product preferences. This description can be used by advertisers to determine the suitability of advertisements to the consumer. Consumers benefit from the system since they will receive advertisements which are more likely to be applicable to them.
• The present invention can be used to profile consumers to support the correlation of an advertisement characterization vector associated with an advertisement with the consumer characterization vector to determine the applicability of the advertisement to the consumer.
• Another feature of the present invention is the ability to price access to the consumer based on the degree of correlation of an advertisement with their profile. If an advertisement is found to be very highly correlated with a consumer's demographics and product preferences, a relatively high price can be charged for transmitting the advertisement to the consumer. From the consumer's perspective, if the correlation between the advertisement and the consumer's demographics or product preferences is high the consumer can charge less to view the ad, since it is likely that is will be of interest.
• The present invention also describes a system for determining the applicability of an advertisement to a consumer, based on the reception of an ad characterization vector and use of a unique consumer ID. The consumer ID is used to retrieve a consumer characterization vector, and the correlation between the consumer characterization vector and the ad characterization vector is used to determine the applicability of the advertisement to the consumer. The price to be paid for presentation of the advertisement can be determined based on the degree of correlation.
• The price to present an advertisement can increase with correlation, as may be typical when the content/opportunity provider is also the profiling entity. The price can decrease with correlation when the consumer is the profiler, and is interested in, and willing to charge less for seeing advertisements which are highly correlated with their demographics, lifestyle, and product preferences.
• The present invention can be used to specify purchasers of a specific product. In a preferred embodiment the advertisement characterization vector contains a description of a target market including an indicator of a target product, i.e., purchasers of a particular product type, brand, or product size. The advertisement characterization vector is correlated with a consumer characterization vector which is retrieved based on a unique consumer ID. The correlation factor is determined and indicates if the consumer is a purchaser of the product the advertisement is intended for. This feature can be used to identify purchasers of a particular brand and can be used to target ads at those consumers to lure them away from their present product provider. Similarly, this feature can be used to target ads to loyal consumers to introduce them to a new product in a product family, or different size of product.
• One advantage of the present invention is that discretionary target market parameters can be specified and do not necessarily need to correspond to an existing market, but can reflect the various market segments for which the advertisement is targeted. The market segments can be designated by demographic characteristics or by product preferences.
• Another advantage of the present invention is that demographic samples of present purchasers of a product are not required to define the target market.
• The present invention can be used to determine the applicability of an advertisement to a consumer based on demographics, product preferences, or a combination of both.
• In a preferred embodiment of the present invention the correlation is calculated as the scalar product of the ad characterization vector and the consumer characterization vector. The ad characterization vector and consumer characterization vector can be composed of demographic characteristics, product purchase characteristics, or a combination of both.
• In a preferred embodiment pricing for the displaying of said advertisement is developed based on the result of the correlation between the ad characterization vector and the consumer characterization vector. In a first embodiment the pricing increases as a function of the correlation. This embodiment can represent the situation in which the party which determines the correlation also controls the ability to display the advertisement.
• In an alternate embodiment the price for displaying the advertisement decreases as a function of the degree of correlation. This embodiment can represent the situation in which the consumer controls access to the consumer characterization vector, and charges less to view advertisements which are highly correlated with their interests and demographics. A feature of this embodiment is the ability of the consumer to decrease the number of unwanted advertisements by charging a higher price to view advertisements which are likely to be of less interest.
• One advantage of the present invention is that it allows advertisements to be directed to new markets by setting specific parameters in the ad characterization vector, and does not require specific statistical knowledge regarding existing customers of similar products.
• Another advantage is that the system allows ads to be directed at consumers of a competing brand, or specific targeting at loyal customers. This feature can be useful for the introduction of a new product to an existing customer base.
• Another advantage of the present invention is that the correlation can be performed by calculating a simple scalar (dot) product of the ad characterization and consumer characterization vectors. A weighted sum or other statistical analysis is not required to determine the applicability of the advertisement.
• The present invention can be realized as a data processing system and as a computer program. The invention can be realized on an individual computer or can be realized using distributed computers with portions of the system operating on various computers.
• An advantage of the present invention is the ability to direct advertisements to consumers which will find the advertisements of interest. This eliminates unwanted advertisements. Another advantage is the ability of advertisers to target specific groups of potential customers.
• These and other features and objects of the invention will be more fully understood from the following detailed description of the preferred embodiments which should be read in light of the accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
• The accompanying drawings, which are incorporated in and form a part of the specification, illustrate the embodiments of the present invention and, together with the description serve to explain the principles of the invention.
• In the drawings:
• FIGS. 1A and 1B show user relationship diagrams for the present invention;
• FIGS. 2A, 2B,2C and2D illustrate a probabilistic consumer demographic characterization vector, a deterministic consumer demographic characterization vector, a consumer product preference characterization vector, and a storage structure for consumer characterization vectors respectively;
• FIGS. 3A and 3B illustrate an advertisement demographic characterization vector and an advertisement product preference characterization vector respectively;
• FIG. 4 illustrates a computer system on which the present invention can be realized;
• FIG. 5 illustrates a context diagram for the present invention;
• FIGS. 6A and 6B illustrate pseudocode updating the characteristics vectors and for a correlation operation respectively;
• FIG. 7 illustrates heuristic rules;
• FIGS. 8A and 8B illustrate flowcharts for updating consumer characterization vectors and a correlation operation respectively; and
• FIG. 9 represents pricing as a function of correlation.
• FIG. 10 illustrates a representation of a consumer characterization as a set of basis vectors and an ad characterization vector.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
• In describing a preferred embodiment of the invention illustrated in the drawings, specific terminology will be used for the sake of clarity. However, the invention is not intended to be limited to the specific terms so selected, and it is to be understood that each specific term includes all technical equivalents which operate in a similar manner to accomplish a similar purpose.
• With reference to the drawings, in general, andFIGS. 1 through 10 in particular, the method and apparatus of the present invention is disclosed.
• FIG. 1A shows a user relationship diagram which illustrates the relationships between a consumer profiling system and various entities. As can be seen inFIG. 1, aconsumer100 can receive information and advertisements from a consumer personal computer (PC)104, displayed on atelevision108 which is connected to asettop106, or can receive a mailedad182.
• Advertisements and information displayed onconsumer PC104 ortelevision108 can be received over anInternet150, or can be received over the combination of theInternet150 with another telecommunications access system. The telecommunications access system can include but is not limited to cable TV delivery systems, switched digital video access systems operating over telephone wires, microwave telecommunications systems, or any other medium which provides connectivity between theconsumer100 and acontent server162 andad server146.
• A content/opportunity provider160 maintains thecontent server162 which can transmit content including broadcast programming across a network such as theInternet150. Other methods of data transport can be used including private data networks and can connect the content sever160 through an access system to a device owned byconsumer100.
• Content/opportunity provider160 is termed such since ifconsumer100 is receiving a transmission fromcontent server162, the content/opportunity provider can insert an advertisement. For video programming, content/opportunity provider is typically the cable network operator or the source of entertainment material, and the opportunity is the ability to transmit an advertisement during a commercial break.
• The majority of content that is being transmitted today is done so in broadcast form, such as broadcast television programming (broadcast over the air and via cable TV networks), broadcast radio, and newspapers. Although the interconnectivity provided by the Internet will allow consumer specific programming to be transmitted, there will still be a large amount of broadcast material which can be sponsored in part by advertising. The ability to insert an advertisement in a broadcast stream (video, audio, or mailed) is an opportunity foradvertiser144. Content can also be broadcast over the Internet and combined with existing video services, in which case opportunities for the insertion of advertisements will be present.
• AlthoughFIG. 1A represents content/opportunity provider160 andcontent server162 as being independently connected toInternet150, with the consumer's devices being also being directly connected to theInternet150, the content/opportunity provider160 can also control access to the subscriber. This can occur when the content/opportunity provider is also the cable operator or telephone company. In such instances, the cable operator or telephone company can be providing content toconsumer100 over the cable operator/telephone company access network. As an example, if the cable operator has control over the content being transmitted to theconsumer100, and has programmed times for the insertion of advertisements, the cable operator is considered to be a content/opportunity provider160 since the cable operator can provide advertisers the opportunity to accessconsumer100 by inserting an advertisement at the commercial break.
• In a preferred embodiment of the present invention, a pricing policy can be defined. The content/opportunity provider160 can chargeadvertiser144 for access toconsumer100 during an opportunity. In a preferred embodiment the price charged for access toconsumer100 by content/opportunity provider varies as a function of the applicability of the advertisement toconsumer100. In analternate embodiment consumer100 retains control of access to the profile and charges for viewing an advertisement.
• The content provider can also be a mailing company or printer which is preparing printed information forconsumer100. As an example,content server162 can be connected to aprinter164 which creates a mailedad182 forconsumer100. Alternatively,printer164 can produce advertisements for insertion into newspapers which are delivered toconsumer100. Other printed material can be generated byprinter162 and delivered toconsumer100 in a variety of ways.
• Advertiser144 maintains anad server146 which contains a variety of advertisements in the form of still video which can be printed, video advertisements, audio advertisements, or combinations thereof.
• Profiler140 maintains aconsumer profile server130 which contains the characterization ofconsumer100. The consumer profiling system is operated byprofiler140, who can useconsumer profile server130 or another computing device connected toconsumer profile server130 to profileconsumer100.
• Data to perform the consumer profiling is received from a point ofpurchase110. Point ofpurchase110 can be a grocery store, department store, other retail outlet, or can be a web site or other location where a purchase request is received and processed. In a preferred embodiment, data from the point of purchase is transferred over a public orprivate network120, such as a local area network within a store or a wide area network which connects a number of department or grocery stores. In an alternate embodiment the data from point ofpurchase110 is transmitted over theInternet150 toprofiler140.
• Profiler140 may be a retailer who collects data from its stores, but can also be a third party who contracts withconsumer100 and the retailer to receive point of purchase data andprofile consumer100.Consumer100 may agree to such an arrangement based on the increased convenience offered by targeted ads, or through a compensation arrangement in which they are paid on a periodic basis for revealing their specific purchase records.
• Consumer profile server130 can contain a consumer profile which is determined from observation of the consumer's viewing habits ontelevision108 orconsumer PC104. Determination of demographic and product preference information based on the consumer's use of services such as cable television and Internet access can be performed by monitoring the channel selections that a subscriber makes, and determining household demographics based on the subscriber selections and information associated with the programming being watched.
• In one embodiment the channel selections are recorded, and based on the time of day during which the programming is watched and duration of viewing, heuristic rules are applied to make probabilistic determinations regarding the household demographics including age, gender, household size and income, as illustrated inFIG. 2A. This can be accomplished by applying heuristic rules which associate the programming with known and assumed characteristics for viewers of the programming. As an example, it is known that the probability that the viewer of a cartoon in the morning is in the 3-8 year old age group is high, thus if the household viewing habits consistently record viewing of cartoons the probability that the household will contain one or more viewers in the 3-8 year old age group is high.
• In one embodiment information regarding the program is extracted from the Electronic Program Guide (EPG) which contains information regarding the scheduled programming. In another embodiment information regarding the programming is retrieved from the closed caption channel transmitted in the broadcast signal.
• The volume at which the program is watched can also be stored and forms an additional basis for subscriber characterization, wherein the muting of a channel indicates limited interest in a particular program or advertisement. In the case of an advertisement, muting of the advertisement can be used as a measure of the effectiveness (or ineffectiveness) of the advertisement and can serve as part of the basis for the subscriber characterization. The muting of a program, as well as the duration for which the program is watched, can also be used in the determination of the subscriber characterization vector.
• By processing the recorded viewing habits in conjunction with programming related information and heuristic rules similar to those illustrated inFIG. 7 but related to programming rather than purchases, it is possible to construct a subscriber characterization vector which contains a probabilistic demographic profile of the household.
• When used herein, the term consumer characterization vector also represents the subscriber characterization vector previously described. Both the consumer characterization vector and the subscriber characterization vector contain demographic and product preference information which is related toconsumer100.
• FIG. 1B illustrates an alternate embodiment of the present invention in which theconsumer100 is alsoprofiler140.Consumer100 maintainsconsumer profile server130 which is connected to a network, either directly or throughconsumer PC104 orsettop106.Consumer profile server130 can contain the consumer profiling system, or the profiling can be performed in conjunction withconsumer PC104 orsettop106. A subscriber characterization system which monitors the viewing habits ofconsumer100 can be used in conjunction with the consumer profiling system to create a more accurate consumer profile.
• When theconsumer100 is also theprofiler140, as shown inFIG. 1B, access to the consumer demographic and product preference characterization is controlled exclusively byconsumer100, who will grant access to the profile in return for receiving an increased accuracy of ads, for cash compensation, or in return for discounts or coupons on goods and services.
• FIG. 2A illustrates an example of a probabilistic demographic characterization vector. The demographic characterization vector is a representation of the probability that a consumer falls within a certain demographic category such as an age group, gender, household size, or income range.
• In a preferred embodiment the demographic characterization vector includes interest categories. The interest categories may be organized according to broad areas such as music, travel, and restaurants. Examples of music interest categories include country music, rock, classical, and folk. Examples of travel categories include “travels to another state more than twice a year,” and travels by plane more than twice a year.”
• FIG. 2B illustrates a deterministic demographic characterization vector. The deterministic demographic characterization vector is a representation of the consumer profile as determined from deterministic rather than probabilistic data. As an example, ifconsumer100 agrees to answer specific questions regarding age, gender, household size, income, and interests the data contained in the consumer characterization vector will be deterministic.
• As with probabilistic demographic characterization vectors, the deterministic demographic characterization vector can include interest categories. In a preferred embodiment,consumer100 answers specific questions in a survey generated byprofiler140 and administered over the phone, in written form, or via theInternet150 andconsumer PC104. The survey questions correspond either directly to the elements in the probabilistic demographic characterization vector, or can be processed to obtain the deterministic results for storage in the demographic characterization vector.
• FIG. 2C illustrates a product preference vector. The product preference represents the average of the consumer preferences over past purchases. As an example, a consumer who buys the breakfast cereal manufactured by Post under the trademark ALPHABITS about twice as often as purchasing the breakfast cereal manufactured by Kellogg under the trademark CORN FLAKES, but who never purchases breakfast cereal manufactured by General Mills under the trademark WHEATIES, would have a product preference characterization such as that illustrated inFIG. 2C. As shown inFIG. 2C, the preferred size of the consumer purchase of a particular product type can also be represented in the product preference vector.
• FIG. 2D represents a data structure for storing the consumer profile, which can be comprised of aconsumer ID field237, a deterministicdemographic data field239, a probabilisticdemographic data field241, and one or more product preference data fields243. As shown inFIG. 2D, the productpreference data field243 can be comprised of multiple fields arranged byproduct categories253.
• Depending on the data structure used to store the information contained in the vector, any of the previously mentioned vectors may be in the form of a table, record, linked tables in a relational database, series of records, or a software object.
• Theconsumer ID512 can be any identification value uniquely associated withconsumer100. In a preferredembodiment consumer ID512 is a telephone number, while in an alternateembodiment consumer ID512 is a credit card number. Other unique identifiers include consumer name with middle initial or a unique alphanumeric sequence, the consumer address, social security number.
• The vectors described and represented in FIGS.2A-C form consumer characterization vectors that can be of varying length and dimension, and portions of the characterization vector can be used individually. Vectors can also be concatenated or summed to produce longer vectors which provide a more detailed profile ofconsumer100. A matrix representation of the vectors can be used, in which specific elements, such aproduct categories253, are indexed. Hierarchical structures can be employed to organize the vectors and to allow hierarchical search algorithms to be used to locate specific portions of vectors.
• FIGS. 3A and 3B represent an ad demographics vector and an ad product preference vector respectively. The ad demographics vector, similar in structure to the demographic characterization vector, is used to target the ad by setting the demographic parameters in the ad demographics vector to correspond to the targeted demographic group. As an example, if an advertisement is developed for a market which is the 18-24 and 24-32 age brackets, no gender bias, with a typical household size of 2-5, and income typically in the range of $20,000-$50,000, the ad demographics vector would resemble the one shown inFIG. 3A. The ad demographics vector represents a statistical estimate of who the ad is intended for, based on the advertisers belief that the ad will be beneficial to the manufacturer when viewed by individuals in those groups. The benefit will typically be in the form of increased sales of a product or increased brand recognition. As an example, an “image ad” which simply shows an artistic composition but which does not directly sell a product may be very effective for young people, but may be annoying to older individuals. The ad demographics vector can be used to establish the criteria which will direct the ad to the demographic group of 18-24 year olds.
• FIG. 3B illustrates an ad product preference vector. The ad product preference vector is used to select consumers which have a particular product preference. In the example illustrated inFIG. 3B, the ad product preference vector is set so that the ad can be directed a purchasers of ALPHABITS and WHEATIES, but not at purchasers of CORN FLAKES. This particular setting would be useful when the advertiser represents Kellogg and is charged with increasing sales of CORN FLAKES. By targeting present purchasers of ALPHABITS and WHEATIES, the advertiser can attempt to sway those purchasers over to the Kellogg brand and in particular convince them to purchase CORN FLAKES. Given that there will be a payment required to present the advertisement, in the form of a payment to the content/opportunity provider160 or to theconsumer100, theadvertiser144 desires to target the ad and thereby increase its cost effectiveness.
• In the event thatadvertiser144 wants to reach only the purchasers of Kellogg's CORN FLAKES, that category would be set at a high value, and in the example shown would be set to 1. As shown inFIG. 3B, product size can also be specified. If there is no preference to size category the values can all be set to be equal. In a preferred embodiment the values of each characteristic including brand and size are individually normalized.
• Because advertisements can be targeted based on a set of demographic and product preference considerations which may not be representative of any particular group of present consumers of the product, the ad characterization vector can be set to identify a number of demographic groups which would normally be considered to be uncorrelated. Because the ad characterization vector can have target profiles which are not representative of actual consumers of the product, the ad characterization vector can be considered to have discretionary elements. When used herein the term discretionary refers to a selection of target market characteristics which need not be representative of an actual existing market or single purchasing segment.
• In a preferred embodiment the consumer characterization vectors shown in FIGS.2A-C and the ad characterization vectors represented inFIGS. 3A and 3B have a standardized format, in which each demographic characteristic and product preference is identified by an indexed position. In a preferred embodiment the vectors are singly indexed and thus represent coordinates in n-dimensional space, with each dimension representing a demographic or product preference characteristic. In this embodiment a single value represents one probabilistic or deterministic value (e.g. the probability that the consumer is in the 18-24 year old age group, or the weighting of an advertisement to the age group).
• In an alternate embodiment a group of demographic or product characteristics forms an individual vector. As an example, age categories can be considered a vector, with each component of the vector representing the probability that the consumer is in that age group. In this embodiment each vector can be considered to be a basis vector for the description of the consumer or the target ad. The consumer or ad characterization is comprised of a finite set of vectors in a the vector space that describes the consumer or advertisement.
• FIG. 4 shows the block diagram of a computer system for a realization of the consumer profiling system. Asystem bus422 transports data amongst theCPU203, theRAM204, Read Only Memory—Basic Input Output System (ROM-BIOS)406 and other components. TheCPU203 accesses ahard drive400 through adisk controller402. The standard input/output devices are connected to thesystem bus422 through the I/O controller201. A keyboard is attached to the I/O controller201 through akeyboard port416 and the monitor is connected through amonitor port418. The serial port device uses aserial port420 to communicate with the I/O controller201. Industry Standard Architecture (ISA)expansion slots408 and Peripheral Component Interconnect (PCI)expansion slots410 allow additional cards to be placed into the computer. In a preferred embodiment, a network card is available to interface a local area, wide area, or other network. The computer system shown inFIG. 4 can be part ofconsumer profile server130, or can be a processor present in another element of the network.
• FIG. 5 shows a context diagram for the present invention. Context diagrams are useful in illustrating the relationship between a system and external entities. Context diagrams can be especially useful in developing object oriented implementations of a system, although use of a context diagram does not limit implementation of the present invention to any particular programming language. The present invention can be realized in a variety of programming languages including but not limited to C, C++, Smalltalk, Java, Perl, and can be developed as part of a relational database. Other languages and data structures can be utilized to realize the present invention and are known to those skilled in the art.
• Referring toFIG. 5, in a preferred embodimentconsumer profiling system500 is resident onconsumer profile server130. Point ofpurchase records510 are transmitted from point ofpurchase110 and stored onconsumer profile server130.Heuristic rules records530,pricing policy570, andconsumer profile560 are similarly stored onconsumer profile server130. In a preferred embodiment advertisement records540 are stored onad server146 and connectivity betweenadvertisement records540 andconsumer profiling system500 is via the Internet or other network.
• In an alternate embodiment the entities represented inFIG. 5 are located on servers which are interconnect via the Internet or other network.
• Consumer profiling system500 receives purchase information from a point of purchase, as represented by point of purchase records510. The information contained within the point ofpurchase records510 includes aconsumer ID512, aproduct ID514 of the purchased product, thequantity516 purchased and theprice518 of the product. In a preferred embodiment, the date and time ofpurchase520 are transmitted by point ofpurchase records510 toconsumer profiling system500.
• Theconsumer profiling system500 can access theconsumer profile560 to update the profiles contained in it.Consumer profiling system500 retrieves aconsumer characterization vector562 and aproduct preference vector564. Subsequent to retrieval one or more data processing algorithms are applied to update the vectors. An algorithm for updating is illustrated in the flowchart inFIG. 8A. The updated vectors termed herein as newdemographic characterization vector566 andnew product preference568 are returned toconsumer profile560 for storage.
• Consumer profiling system500 can determine probabilistic consumer demographic characteristics based on product purchases by applyingheuristic rules519.Consumer profiling system500 provides aproduct ID514 toheuristic rules records530 and receives heuristic rules associated with that product. Examples of heuristic rules are illustrated inFIG. 7.
• In a preferred embodiment of the present invention,consumer profiling system500 can determine the applicability of an advertisement to theconsumer100. For determination of the applicability of an advertisement, acorrelation request546 is received byconsumer profiling system500 fromadvertisements records540, along withconsumer ID512.Advertisements records540 also provides advertisement characteristics including anad demographic vector548, an ad product category552 and an adproduct preference vector554.
• Application of a correlation process, as will be described in accordance withFIG. 8B, results in ademographic correlation556 and aproduct correlation558 which can be returned to advertisement records540. In a preferred embodiment,advertiser144 usesproduct correlation558 anddemographic correlation556 to determine the applicability of the advertisement and to determine if it is worth purchasing the opportunity. In a preferred embodiment,pricing policy570 is utilized to determine anad price572 which can be transmitted fromconsumer profiling system500 toadvertisement records540 for use byadvertiser144.
• Pricing policy570 is accessed byconsumer profiling system500 to obtainad price572.Pricing policy570 takes into consideration results of the correlation provided by theconsumer profiling system500. An example of pricing schemes are illustrated inFIG. 9
• FIGS. 6A and 6B illustrate pseudocode for the updating process and for a correlation operation respectively. The updating process involves utilizing purchase information in conjunction with heuristic rules to obtain a more accurate representation ofconsumer100, stored in the form of a newdemographic characterization vector562 and a newproduct preference vector568.
• As illustrated in the pseudocode inFIG. 6A the point of purchase data is read and the products purchased are integrated into the updating process.Consumer profiling system500 retrieves a product demographics vector obtained from the set ofheuristic rules519 and applies the product demographics vector to thedemographics characterization vector562 and theproduct preference vector564 from theconsumer profile560.
• The updating process as illustrated by the pseudocode inFIG. 6A utilizes a weighting factor which determines the importance of that product purchase with respect to all of the products purchased in a particular product category. In a preferred embodiment the weight is computed as the ratio of the total of products with aparticular product ID514 purchased at that time, to the product total purchase, which is the total quantity of the product identified by itsproduct ID514 purchased byconsumer100 identified by itsconsumer ID512, purchased over an extended period of time. In a preferred embodiment the extended period of time is one year.
• In the preferred embodiment the product category total purchase is determined from a record containing the number of times thatconsumer100 has purchased a product identified by a particular product ID.
• In an alternate embodiment other types of weighting factors, running averages and statistical filtering techniques can be used to use the purchase data to update the demographic characterization vector. The system can also be reset to clear previous demographic characterization vectors and product preference vectors.
• The newdemographic characterization vector566 is obtained as the weighted sum of the product demographics vector thedemographic characterization vector562. The same procedure is performed to obtain the newproduct preference vector568. Before storing those new vectors, a normalization is performed on the said new vectors. When used herein the term product characterization information refers product demographics vectors, product purchase vectors or heuristic rules, all of which can be used in the updating process. The product purchase vector refers to the vector which represents the purchase of a item represented by a product ID. As an example, a product purchase vector for the purchase of Kellogg's CORN FLAKES in a 32 oz. size has a product purchase vector with a unity value for Kellogg's CORN FLAKES and in the 32 oz. size. In the updating process the weighted sum of the purchase as represented by the product purchase vector is added to the product preference vector to update the product preference vector, increasing the estimated probability that the consumer will purchase Kellogg's CORN FLAKES in the 32 oz. size.
• InFIG. 6B the pseudocode for a correlation process is illustrated.Consumer profiling system500, after receiving the product characteristics and theconsumer ID512 from the advertisement records retrieves the consumerdemographic characterization vector562 and itsproduct preference vector564. The demographic correlation is the correlation between thedemographic characterization vector562 and the ad demographics vector. The product correlation is the correlation between the adproduct preference vector554 and theproduct preference vector564.
• In a preferred embodiment the correlation process involves computing the dot product between vectors. The resulting scalar is the correlation between the two vectors.
• In an alternate embodiment, as illustrated inFIG. 10, the basis vectors which describe aspects of the consumer can be used to calculate the projections of the ad vector on those basis vectors. In this embodiment, the result of the ad correlation can itself be in vector form whose components represent the degree of correlation of the advertisement with each consumer demographic or product preference feature. As shown inFIG. 10 the basis vectors are the age of theconsumer1021, the income of theconsumer1001, and the family size of theconsumer1031. Thead characterization vector1500 represents the desired characteristics of the target audience, and can include product preference as well as demographic characteristics.
• In this embodiment the degree of orthogonality of the basis vectors will determine the uniqueness of the answer. The projections on the basis vectors form a set of data which represent the corresponding values for the parameter measured in the basis vector. As an example, if household income is one basis vector, the projection of the ad characterization vector on the household income basis vector will return a result indicative of the target household income for that advertisement.
• Because basis vectors cannot be readily created from some product preference categories (e.g. cereal preferences) an alternate representation to that illustrated inFIG. 2C can be utilized in which the product preference vector represents the statistical average of purchases of cereal in increasing size containers. This vector can be interpreted as an average measure of the cereal purchased by the consumer in a given time period.
• The individual measurements of correlation as represented by the correlation vector can be utilized in determining the applicability of the advertisement to the subscriber, or a sum of correlations can be generated to represent the overall applicability of the advertisement.
• In a preferred embodiment individual measurements of the correlations, or projections of the ad characteristics vector on the consumer basis vectors, are not made available to protect consumer privacy, and only the absolute sum is reported. In geometric terms this can be interpreted as disclosure of the sum of the lengths of the projections rather than the actual projections themselves.
• In an alternate embodiment the demographic and product preference parameters are grouped to form sets of paired scores in which elements in the consumer characterization vector are paired with corresponding elements of the ad characteristics vector. A correlation coefficient such as the Pearson product-moment correlation can be calculated. Other methods for correlation can be employed and are well known to those skilled in the art.
• When the consumer characterization vector and the ad characterization vector are not in a standardized format, a transformation can be performed to standardize the order of the demographic and product preferences, or the data can be decomposed into sets of basis vectors which indicate particular attributes such as age, income or family size.
• FIG. 7 illustrates an example of heuristic rules including rules for defining a product demographics vector. From the product characteristics, a probabilistic determination of household demographics can be generated. Similarly, the monthly quantity purchased can be used to estimate household size. The heuristic rules illustrated inFIG. 7 serve as an example of the types of heuristic rules which can be employed to better characterizeconsumer100 as a result of their purchases. The heuristic rules can include any set of logic tests, statistical estimates, or market studies which provide the basis for better estimating the demographics ofconsumer100 based on their purchases.
• InFIG. 8A the flowchart for updating the consumer characterization vectors is depicted. The system receives data from the point of purchase at receive point ofpurchase information step800. The system performs a test to determine if a deterministic demographic characterization vector is available at deterministic demographic informationavailable step810 and, if not, proceeds to update the demographic characteristics.
• Referring toFIG. 8A, at read purchaseID info step820, theproduct ID514 is read, and at update consumer demographiccharacterization vector step830, an algorithm such as that represented inFIG. 6A is applied to obtain a newdemographic characterization vector566, which is stored in theconsumer profile560 at store updated demographiccharacterization vector step840.
• Theend test step850 can loop back to the readpurchase ID info820 if all the purchased products are not yet processed for updating, or continue to the branch for updating theproduct preference vector564. In this branch, the purchased product is identified at read purchaseID info step820. An algorithm, such as that illustrated inFIG. 6A for updating theproduct preference vector564, is applied in update productpreference vector step870. The updated vector is stored inconsumer profile560 at store productpreference vector step880. This process is carried out until all the purchased items are integrated in the updating process.
• FIG. 8B shows a flowchart for the correlation process. Atstep900 the advertisement characteristics described earlier in accordance withFIG. 5 along with the consumer ID are received byconsumer profiling system500. Atstep910 thedemographic correlation556 is computed and atstep920 theproduct preference correlation558 is computed. An illustrative example of an algorithm for correlation is presented inFIG. 6b. The system returnsdemographic correlation556 andproduct preference correlation558 to the advertisement records540 before exiting the procedure atend step950.
• FIG. 9 illustrates two pricing schemes, one for content/opportunity provider160 basedpricing970, which shows increasing cost as a function of correlation. In this pricing scheme, the higher the correlation, the more the content/opportunity provider160 charges to air the advertisement.
• FIG. 9 also illustrates consumer basedpricing960, which allows a consumer to charge less to receive advertisements which are more highly correlated with their demographics and interests.
• As an example of the industrial applicability of the invention, aconsumer100 can purchase items in a grocery store which also acts as aprofiler140 using aconsumer profiling system500. The purchase record is used by the profiler to update the probabilistic representation ofcustomer100, both in terms of their demographics as well as their product preferences. For each item purchased byconsumer100, product characterization information in the form of a product demographics vector and a product purchase vector is used to update the demographic characterization vector and the product preference vector forconsumer100.
• A content/opportunity provider160 may subsequently determine that there is an opportunity to present an advertisement toconsumer100. Content/opportunity provider160 can announce this opportunity toadvertiser144 by transmitting the details regarding the opportunity and theconsumer ID512.Advertiser144 can then queryprofiler140 by transmittingconsumer ID512 along with advertisement specific information including thecorrelation request546 andad demographics vector548. Theconsumer profiling system500 performs a correlation and determines the extent to which the ad target market is correlated with the estimated demographics and product preferences ofconsumer100. Based on thisdetermination advertiser144 can decide whether to purchase the opportunity or not.
• Although this invention has been illustrated by reference to specific embodiments, it will be apparent to those skilled in the art that various changes and modifications may be made which clearly fall within the scope of the invention. The invention is intended to be protected broadly within the spirit and scope of the appended claims.

Claims (22)

1. A method of profiling users in an computer environment, said method comprising:

(a) receiving a purchase history corresponding to a unique user and including information related to the purchase of at least one item by said unique user;

(b) receiving demographic information corresponding to said unique user;

(c) receiving product characterization information describing a statistical relationship between a particular product and demographic characteristics of purchasers of the product;

(d) extrapolating additional demographic information from said purchase history and said product characterization information; and

(e) updating said demographic information based on said additional demographic information.

2. The method ofclaim 1, wherein said product characterization information is developed based on a market study of a user population that visits a particular website.

3. The method ofclaim 1, wherein said demographic information is received from a user demographic profile corresponding to said unique user.

4. The method ofclaim 3, further comprising:

(f) identifying missing demographic information in said user demographic profile.

5. The method ofclaim 5, wherein at least a portion of said missing demographic information in said user demographic profile is obtained through said additional demographic information.

6. The method ofclaim 5, further comprising:

(g) returning said updated demographic information of step (e) to said user demographic profile.

7. The method ofclaim 1, wherein said product characterization information does not include information about specific users.

8. The method ofclaim 1, further comprising:

(f) targeting advertisements based on said updated demographic information.

9. The method ofclaim 1, further comprising:

(f) comparing said updated demographic information to a target expression;

(g) generating a score based on said comparing; and

(h) delivering an advertisement based on said score.

10. A method of targeting ads in an computer environment based on user segmentation, said method comprising:

(a) receiving user profile information corresponding to a unique user;

(b) assigning, based on said user profile information, said unique user to a population segment; and

(c) comparing said population segment of said unique user to an ad segment characterization corresponding to at least one advertisement.

11. The method ofclaim 10, further comprising:

(d) calculating a correlation factor between said ad segment characterization and said population segment of said unique user; and

(e) targeting an ad based on said correlation factor.

12. The method ofclaim 10, wherein said assigning is realized by creating a population segment vector that describes the population segment of said unique user.

13. The method ofclaim 12, wherein said ad segment characterization is represented by an ad segment vector.

14. The method ofclaim 13, wherein said comparing is realized by comparing said population segment vector to said ad segment vector.

15. The method ofclaim 14, further comprising:

(d) calculating a correlation factor between said ad segment vector and said population segment vector; and

(e) targeting an ad based on said correlation factor.

16. The method ofclaim 10, wherein said user profile information includes purchase history information.

17. The method ofclaim 16, wherein said purchase history information includes a record of at least one purchase.

18. The method ofclaim 10, wherein said user profile information includes demographic information.

19. A method of presenting cross-sell products in a computer environment, said method comprising:

(a) receiving a purchase history corresponding to a unique user and including information related to the purchase of at least one product;

(b) receiving product characterization information for said at least one product wherein said product characterization describes a relationship between said at least one product and demographic characteristics of purchasers of said at least one product;

(c) calculating a consumer characterization vector based on said purchase history and said product characterization information; and

(d) suggesting items based on said consumer characterization vector.

20. The method ofclaim 19, wherein said consumer characterization vector is based on a combination of a demographic characterization vector and a product preference vector.

21. The method ofclaim 20, wherein the suggestion made in step (d) is based on the product preference vector.

22. A method of profiling users in a computer environment, said method comprising:

(a) receiving a purchase history wherein said purchase history corresponding to a unique user and said purchase history including information related to the purchase of at least one item by said unique user;

(b) receiving demographic information corresponding to said unique user;

(c) receiving product characterization information describing a statistical relationship between a particular product and demographic characteristics of purchasers of the product;

(d) calculating additional demographic information from said purchase history and said product characterization information; and

(e) updating said demographic information based on said additional demographic information.

Images