REALLOCATION OF ELECTRONIC RESOURCES USING A PREDICTIVE MODEL OF ATTRIBUTION

Abstract

Systems and methods for predictive modeling of attribution are described. Systems and methods may include receiving one or more inputs; processing the one or more inputs using a general linear model; and providing predicted online and offline campaign impact.

Description

FIELD OF THE INVENTION

The present invention relates to systems and methods for marketing campaigns, and, more specifically, to systems and methods for improving speed of online and offline attribution.

BACKGROUND OF THE INVENTION

Targeted marketing is a commonly used tool for improving return on investment for advertising expenditures. In general, the more accurate the targeting is to consumers, the more benefit is received from the advertising campaign.

Measurement of the effectiveness of advertising campaigns provides feedback that can be used to determine whether the advertising campaign has been effective. The current industry technology uses stratified sample groups of campaign prospects separated into a treated and control group to measure effectiveness of a campaign incrementally. These determinations are made on a monthly basis. Existing technology does not optimize campaign return on investment because it does not utilize real time data to adjust for optimization. In addition, current industry technology targets based on cookies or sites and not based on email address.

Needs exist for improved systems and methods for improved systems and methods for marketing campaigns.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate preferred embodiments of the invention and together with the detailed description serve to explain the principles of the invention. In the drawings:

FIG. 1 shows an exemplary system for predictive modeling of attribution.

FIG. 2 shows an exemplary system for computational aspects of predictive modeling of attribution.

FIG. 3 shows an exemplary flow diagram for predictive modeling of attribution.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Systems and methods are described for using various tools and procedures for optimizing targeted advertising. In certain embodiments, the tools and procedures may be used in conjunction with improved attribution. The examples described herein relate to marketing campaigns, including email and Internet based advertising campaigns, for illustrative purposes only. The systems and methods described herein may be used for many different industries and purposes, including any type of marketing campaigns and/or other industries completely. In particular, the systems and methods may be used for any industry or purpose where customized customer identification is needed. For multi-step processes or methods, steps may be performed by one or more different parties, servers, processors, etc.

Certain embodiments may provide systems and methods for targeted advertising. A set of information may be accessed from one or more databases. The information may include various types of information, including, but not limited to, real time campaign information, audience profiles, and attribution data. A model may be accessed or created. The model may be a general linear model for determining factors for predicting results of an advertising campaign. The general linear model may be used to project online and offline impacts of marketing campaigns.

An email channel may be any communication sent electronically to an electronic address, i.e., sent via email. In certain embodiments, an email channel may refer to sending of third party advertisements through email.

In general, inventory may be a term for a unit of advertising space, such as a magazine page, television airtime, direct mail message, email messages, text messages, telephone calls, etc. Advertising inventory may be advertisements a publisher has available to sell to an advertiser. In certain embodiments, advertising inventory may refer to a number of email advertisements being bought and/or sold. The terms inventory and advertising inventory may be used interchangeably. For email marketing campaigns, advertising inventory is typically an email message.

A publisher may be an entity that sells advertising inventory, such as those produced by the systems and methods herein, to their email subscriber database. An advertiser may be a buyer of publisher email inventory. Examples of advertisers may include various retailers. A marketplace may allow advertisers and publishers to buy and sell advertising inventory. Marketplaces, also called exchanges or networks, may be used to sell display, video, and mobile inventory. In certain embodiments, a marketplace may be an email exchange/email marketplace. An email exchange may be a type of marketplace that facilitates buying and/or selling of inventory between advertisers and publishers. This inventory may be characterized based on customer attributes used in marketing campaigns. Therefore, an email exchange may have inventory that can be queried by each advertiser. This may increase efficiency of advertisers when purchasing inventory. A private network may be a marketplace that has more control and requirements for participation by both advertisers and publishers.

An individual record/prospect may be at least one identifier of a target. In certain embodiments, the individual record/prospect may be identified by a record identification mechanism, such as a specific email address (individual or household) that receives an email message.

An audience may be a group of records, which may be purchased as inventory. In certain embodiments, an audience may be a group of records selected from publisher databases of available records. The subset of selected records may adhere to a predetermined set of criteria, such as common age range, common shopping habits, and/or similar lifestyle situation (i.e., stay at home mother). Advertisers generally select the predetermined set of criteria when they are making an inventory purchase.

Although not required, the systems and methods are described in the general context of computer program instructions executed by one or more computing devices that can take the form of a traditional server/desktop/laptop; mobile device such as a smartphone or tablet; etc. Computing devices typically include one or more processors coupled to data storage for computer program modules and data. Key technologies include, but are not limited to, the multi-industry standards of Microsoft and Linux/Unix based Operating Systems; databases such as SQL Server, Oracle, NOSQL, and DB2; Business Analytic/Intelligence tools such as SPSS, Cognos, SAS, etc.; development tools such as Java,.NET Framework (VB.NET, ASP.NET, AJAX.NET, etc.); and other e-Commerce products, computer languages, and development tools. Such program modules generally include computer program instructions such as routines, programs, objects, components, etc., for execution by the one or more processors to perform particular tasks, utilize data, data structures, and/or implement particular abstract data types. While the systems, methods, and apparatus are described in the foregoing context, acts and operations described hereinafter may also be implemented in hardware.

FIG. 1 shows an exemplary system 100 for predictive modeling of online and offline attribution according to one embodiment. In this exemplary implementation, system 100 may include one or more servers/computing devices 102 (e.g., server 1, server 2, . . . , server n) operatively coupled over network 104 to one or more client computing devices 106-1 to 106-n, which may include one or more consumer computing devices, one or more provider computing devices, one or more remote access devices, etc. The one or more servers/computing devices 102 may also be operatively connected, such as over a network, to one or more third party servers/databases 114 (e.g., database 1, database 2, . . . , database n). The one or more servers/computing devices 102 may also be operatively connected, such as over a network, to one or more system databases 116 (e.g., database 1, database 2, . . . , database n). Various devices may be connected to the system, including, but not limited to, client computing devices, consumer computing devices, provider computing devices, remote access devices, etc. This system may receive inputs 118 and outputs 120 from the various computing devices, servers and databases.

Server/computing device 102 may represent, for example, any one or more of a server, a general-purpose computing device such as a server, a personal computer (PC), a laptop, a smart phone, a tablet, and/or so on. Networks 104 represent, for example, any combination of the Internet, local area network(s) such as an intranet, wide area network(s), cellular networks, WIFI networks, and/or so on. Such networking environments are commonplace in offices, enterprise-wide computer networks, etc. Client computing devices 106, which may include at least one processor, represent a set of arbitrary computing devices executing application(s) that respectively send data inputs to server/computing device 102 and/or receive data outputs from server/computing device 102. Such computing devices include, for example, one or more of desktop computers, laptops, mobile computing devices (e.g., tablets, smart phones, human wearable device), server computers, and/or so on. In this implementation, the input data comprises, for example, real time campaign data, audience profile, attribution data, and/or so on, for processing with server/computing device 102. In one implementation, the data outputs include, for example, emails, templates, forms, and/or so on. Embodiments of the present invention may also be used for collaborative projects with multiple users logging in and performing various operations on a data project from various locations. Embodiments of the present invention may be web-based, smart phone-based and/or tablet-based or human wearable device based.

In this exemplary implementation, server/computing device 102 includes at least one processor coupled to a system memory. System memory may include computer program. modules and program data.

In this exemplary implementation, server/computing device 102 includes at least one processor 202 coupled to a system memory 204, as shown in FIG. 2. System memory 204 may include computer program modules 206 and program data 208. In this implementation program modules 206 may include data module 210, model module 212, analysis module 214, and other program modules 216 such as an operating system, device drivers, etc. Each program module 210 through 216 may include a respective set of computer-program instructions executable by processor(s) 202. This is one example of a set of program modules and other numbers and arrangements of program modules are contemplated as a function of the particular arbitrary design and/or architecture of server/computing device 102 and/or system 100 (FIG. 1). Additionally, although shown on a single server/computing device 102, the operations associated with respective computer-program instructions in the program modules 206 could be distributed across multiple computing devices. Program data 208 may include campaign data 220, audience data 222, attribution data 224, and other program data 226 such as data input(s), third party data, and/or others.

As shown in FIG. 3, certain embodiments may take one or more types of information, passes them through one or more models, and projects online and offline campaign impacts.

A system 301 may include one or more input sources 303 that provide one or more items of data. Data may be accessed from and/or provided by one or more sources. In certain embodiments, input sources 303 may include, but are not limited to, real time campaign data 305, audience profiles 307, and/or attribution data 309. Items of data may be stored locally or remotely. Items of data may be stored in one or multiple databases.

Real time campaign data 305 may include one or more of the following:

• • opens (action of an email recipient opening an email, which may mean clicking “show images”);
  • clicks (action of an email recipient clicking on email content, which sends them to a landing page in a web browser);
  • landing page actions (action an email recipient complaining on an email, which may include indicating complain in a mail client program);
  • complaints (action of an email recipient unsubscribing on an email, which may include clicking unsubscribe to prevent further emails from the sender or advertiser);
  • unsubscribes (action of an email recipient unsubscribing on an email, which may include clicking unsubscribe to prevent further emails from the sender or advertiser);
  • metrics rates (calculated metrics that indicate (performance of a email campaign): Metrics may be calculations or computed values that are used to measure campaign performance. For example, open rate (the percentage of opens over possible opens) indicate the engagement levels of the email campaigns. Additional metrics may include, but are not limited to, click-through rate (the rate of clicks to possible clicks) s well as additional advertiser specific performance measurements. These metrics can also be considered over time, for example, if the open rate of a campaign starts at X and increases by a margin in the first 5 hours of the campaign, this increase can be used as an input as independent variable in predicting subsequent action;
  • rate of change of metric rates (additional calculated metrics): Velocity of a metric may be a calculation of the rate of change of a metric X. This calculation may yield a second metric, Y, which represents a new data point around which decisions can be made. If two metrics are comparable, the one that is moving in a directionally positive manner may be of greater use in computations; and
  • datetime (date, time, seasonality, as well as other time-based indications).

Audience profiles 307 may include individual and household level demographics from both self-reported sources and third party vendors, digital shopping behavior across other marketing campaigns, and offline shopping behavior sourced from catalogues, loyalty cards, retail stores, etc. Audience profiles 307 may include one or more of the following:

• • demographics explicit information on the email record individual such as, but not limited to, age, gender, income, marital status, etc.);
  • geographic (explicit information on the email record such as, but not limited to, postal address, zip code, state, etc.);
  • online sales (previous online behavior of an email recipient, such as, but not limited to, signing up for one or more services, purchasing one or more products, etc.);
  • offline sales (previous offline behavior of an email recipient, such as, but not limited to, signing up for one or more services, buying one or more products, etc. This may be based on offline SKU level data from retailers, catalogues, loyalty card activity, etc., and may be matched to email prospects based on various identifiers, such as name, postal address, etc.);
  • psychographic (description of personality, values, opinions, attitudes, interests, lifestyles, etc. that allow advertisers to customize content to improve response); and
  • purchase intent data. Purchase intent may be determined based on comparisons between the actions on a specific advertisement compared to a population average. For example, if females age 24-35 click on skin care advertisements at a rate of three times the national average, they may have a three times purchase intent multiplier.

Attribution data 309 may include measurements of the impact of an advertising campaign. Attribution may be a methodology behind measuring the impact of advertising campaigns. Attribution may be a process to identify a set of user actions (“events”) that contribute in some manner to a desired outcome, and then assigning a value to each of these events. In certain embodiments,attribution may determine a total impact of email campaigns not only based on activity online, hut also whether the advertisement contributes to offline activity, such as when the email recipient make a purchase in a brick and mortar store.

To measure campaign impact, an experiment may be performed in which the only difference between two groups of record sets is that one receives an advertisement (treatment group) and one does not (control group). These groups are created based on a stratified sampling process, which ensures that the attributes or characteristics of each group are proportional to each other. After a campaign is executed, the treatment group and the control group are compared to the new customer file provided by the advertiser. There may be specific criteria to determine a “match”. These criteria may include, but are not limited to, a time range (i.e., purchased within 30 days of receiving the advertisement) and a key d (i.e., email, or name and postal address).

With this match information, the new customer rate for both the treatment group and the control group are compared. The difference between these treatment group and the control group customer rates may be the incremental new customer rate of a campaign. The product of the treatment population and the incremental customer rate may be the incremental customers the campaign generated. Using this information, in addition to the cost of the advertising, may provide a true return on investment of the media spend.

In certain embodiments the above process may be executed in real time and/or close to real time.

Certain embodiments may allow for continuously matching the treatment and control files to an advertiser's customer file, and computing the incremental customer rate and the cost per new customer on a continuous and/or near continuous basis across campaigns. If multiple campaigns are launched simultaneous for a specific advertiser, certain embodiments may allow for measuring relative performance of the multiple campaigns and shifting media spend to a better performing campaign. Additionally, certain embodiments may use this modeling information to predict a final return on investment target for a particular campaign.

Attribution data 309 may be based on stratified micro-sampling. Micro-sampling may consider both control groups and treated groups. Control groups may be groups of email recipients that do not receive an advertisement. Treated groups may be groups of email recipients that do receive an advertisement. Attribution data 309 may allow measurement in real or near real time of an incremental lift of a campaign. Incremental lift may be a measured impact from campaigns by comparing response rates of treated and control groups. For example, a determination may be made as to whether a response to an advertisement by a treated group is greater than the response by a control group, which is not treated. A precise significance test may be performed in real time. Significance tests are well-known for determination of whether a value is considered “significant” (i.e., is not simply due to chance). The probability that a variable would assume a value greater than or equal to the observed value strictly by chance may also be determined by a significance test.

Attribution data 309 may include one or more of the following:

• • treatment group/treated prospects records: records that will receive an advertisement;
  • control group/control prospects records: records that will not receive an advertisement;
  • advertiser customer data/new customer file: sales information provided by an advertiser;
  • customer matches: matches between a treatment group or control group and the new customer file on specific criteria based on the advertiser;
  • treatment new customers: number of new customers that match the treatment group;
  • control new customers: number of new customers that match the control group;
  • treatment new customer rate: percent of new customers that were treated over the total treatment prospects;
  • control new customer rate: percent of new customers that were not treated over the total control treatment prospects;
  • incremental new customer rate/incremental customer rate: difference between the treatment new customer rate and the control new customer rate; and
  • incremental new customers: product of the treatment group population and the incremental new customer rate.

Note that customer rates can be measured for different windows of time. For example, in certain embodiments, customer rate may be measured over a set time, such as for five days. The customer rate over the set time may be used to predict a customer rate for a different time frame, such as a thirty date customer rate, for optimization purposes. All incremental customer rates cats be expressed as customer rates.

A general linear model 311 may determine differences in performance between a treated and control group in a marketing campaign based on the input variables. Certain embodiments may use real time campaign data, audience data and attribution data as independent variables in a general linear model. In certain embodiments, the model may use these variables and weight them against each other to determine their effect on a dependent variable (i.e., a projected cost per new customer rate for the entire campaign.) This output may be advertiser specific, but may be focused on return on investment for the marketing initiative in question. The outputs can be on a campaign or creative level, allowing optimization of advertising spend and business decisions.

The general linear model may allow for prediction of a 30-60 day attribution measurement in just days (compared to a traditional 30-60 day window) upon reaching a statistically relevant volume. A statistically relevant volume may depend on the advertising campaign in question, and may be based on a statistical significance test as described above. The input 303 may be provided to or accessed by the general linear model 311. The model 311 may determine one or more influential factors in predicting total sales generated by a campaign. The factors may be weighted based on their expected influence on a campaign.

The model may predict online and offline campaign impact 313. The results may allow for reallocation of advertising spending to top performing campaigns and audiences much faster than standard practices. Predictions 313 may project weekly cost per incremental customer across multiple campaigns. Time periods for various embodiments may vary, and may include real-time, near real-time, daily, weekly, monthly, quarterly, yearly, or other time periods. For example, the prediction 313 may project customer acquisition cost for a customer on a weekly basis giving the client the ability to shift advertising budget to the top performing campaigns. In direct mail, customer acquisition cost calculations take up to six weeks to actualize.

Although the foregoing description is directed to the preferred embodiments of the invention, it is noted that other variations and modifications will be apparent to those skilled in the art, and may be made without departing from the spirit or scope of the invention. Moreover, features described in connection with one embodiment of the invention may be used in conjunction with other embodiments, even if not explicitly stated above.

Claims

1. A computerized method of predictive modeling of attribution, the computerized method comprising the steps of: selecting a first set of control electronic mail recipients;selecting a second set of treatment electronic mail recipients;transmitting, by a computing server, a first electronic mail message to first remote computing devices associated with the first set of control electronic mail recipients using a data communication network;transmitting, by the computing server, a second electronic mail message containing an element of a messaging campaign to second remote computing devices associated with the second treatment set of electronic mail recipients;receiving, by the computing server in real time over the communication network, first data indicating interactions of the first set of control electronic mail recipients with the first electronic mail message and the second data indicating interactions of the second set of treatment electronic mail recipients with the second electronic mail message;receiving, by the computing server, customer information including match information;generating, by the computing server, campaign-related attribution data using the customer information and a difference between the first data and the second data in response to one or both of the first data and the second data satisfying the match information;based on the difference between the first data and the second data, generating an incremental new customer rate of a campaign to which the campaign-related attribution data relates;generating a product of the second set of treatment electronic mail recipients and the incremental new customer rate to identify a set of new customers generated by the campaign;generating an incremental lift by comparing response rates from the first set of control electronic mail recipients and the second set of treatment electronic mail recipients;acquiring, by the computing server, real time campaign data, audience data, and attribution data from one or more databases, and adding the campaign-related attribution data to the attribution data, the real time campaign data including rate of change of metric rates;generating, by the computing server, a general linear model having one or more inputs from each of the real time campaign data, the audience data, and the attribution data as independent input variables in the general linear model; andpredicting in real time, by the computing server, an online and offline campaign impact of the messaging campaign using the general linear model with the campaign-related attribution data as part of the independent input variables to determine differences in performance between the first set of electronic mail recipients and the second set of electronic mail recipients, the predicted campaign impact including data relating to at least one or more of the incremental new customer rate, the set of new customers, and the incremental lift;transmitting, by the computing server, a third electronic mail message based on the predicting.

2. The method of claim 1, wherein selecting the first set of control electronic mail recipients and the second set of treatment electronic mail recipients includes using a sampling process such that attributes of the first set and the second set are proportional to each other.

3. The method of claim 1, wherein the real time campaign data is further selected from the group consisting of: opens, clicks, landing page actions, complaints, unsubscribes, metrics rates, dates, times, and combinations thereof.

4. The method of claim 1, wherein the audience profiles are selected from the group consisting of: demographics, geographic, online sales, offline sales, psychographic, purchase intent data, and combinations thereof.

5. The method of claim 1, wherein the attribution data is selected from the group consisting of: advertiser customer data, treated prospects records, control prospects records, incremental customers, incremental customer rate, and combinations thereof.

6. The method of claim 1, wherein the match information includes a time range for completion of an event with respect to the transmission of the second electronic mail message containing the element of the messaging campaign.

7. The method of claim 1, wherein the method includes varying values of the real time campaign data, the audience data, and the attribution data that are input as independent input variables in the general linear model to determine weights indicating an importance of each of the real time campaign data, the audience data and the attribution data.

8. The method of claim 7, wherein the general linear model determines influential factors.

9. The method of claim 1, wherein the predicted online and offline campaign impact is determined on a weekly basis.

10. A system for predictive modeling of online and offline attribution, the system comprising: one or more databases;system memory comprising instructions; andone or more processors to execute the instructions to perform operations comprising:selecting a first set of control electronic mail recipients;selecting a second set of treatment electronic mail recipients;transmitting a first electronic mail message to first remote computing devices associated with the first set of control electronic mail recipients using a data communication network;transmitting a second electronic mail message containing an element of a messaging campaign to second remote computing devices associated with the second set of treatment electronic mail recipients;receiving, in real time over the communication network, first data indicating interactions of the first set of control electronic mail recipients with the first electronic mail message and the second data indicating interactions of the second set of treatment electronic mail recipients with the second electronic mail message;receiving customer information including match information;generating campaign-related attribution data using the customer information and a difference between the first data and the second data in response to one or both of the first data and the second data satisfying the match information;acquiring real time campaign data, audience data, and attribution data from the one or more databases, and adding the campaign-related attribution data to the attribution data, the real time campaign data including rate of change of metric rates;generating a general linear model having one or more inputs from each of the real time campaign data, the audience data, and the attribution data as independent input variables in the general linear model; andpredicting in real time an online and offline campaign impact of the messaging campaign using the general linear model with the campaign-related attribution data as part to the input independent input variables to determine differences in performance between the first set of electronic mail recipients and the second set of electronic mail recipients;transmitting, by the computing server, a third electronic mail message based on the predicting.

11. The system of claim 10, wherein selecting the first set of control electronic mail recipients and the second set of treatment electronic mail recipients includes using a sampling process such that attributes of the first set and the second set are proportional to each other.

12. The system of claim 10, wherein the real time campaign data is further selected from the group consisting of: opens, clicks, landing page actions, complaints, unsubscribes, metrics rates, dates, times, and combinations thereof.

13. The system of claim 10, wherein the audience profiles are selected from the group consisting of: demographics, geographic, online sales, offline sales, psychographic, purchase intent data, and combinations thereof.

14. The system of claim 10, wherein the attribution data is selected from the group consisting of: advertiser customer data, treated prospects records, control prospects records, incremental customers, incremental customer rate, and combinations thereof.

15. The system of claim 10, wherein the match information includes a time range for completion of an event with respect to the transmission of the second electronic mail message containing the element of the messaging campaign.

16. The system of claim 15, wherein the operations include varying values of the real time campaign data, the audience data, and the attribution data that are input as independent input variables in the general linear model to determine weights indicating an importance of each of the real time campaign data, the audience data and the attribution data.

17. The system of claim 16, wherein the general linear model determines influential factors.

18. The system of claim 10, wherein the predicted online and offline campaign impact is determined on a weekly basis.