US20220207548A1 - Systems and methods for contract based offer generation - Google Patents

Abstract

Systems and methods for a contract-based offer generator is provided. A contract for a promotional offer on a product is received. Data is extracted from the contract. An offer band is accessed, and a plurality of test offers are selected from the offer bank by scoring each offer in the offer bank against the extracted data. The promotional offer and the selected plurality of test offers are deployed in a plurality of retail locations. This is done by maximizing orthogonality between the following variables: store sales, store out of stock rates, number of relevant SKUs carried in each store, temporal effects, discount depth, buy quantity and offer structure.

Description

CROSS REFERENCE TO RELATED APPLICATIONS
• This non-provisional application claims priority to U.S. Provisional Application No. 63/143,847, filed Jan. 30, 2021, of the same inventors and title, pending, which application is incorporated herein in its entirety by this reference.
• Additionally, this is a continuation-in-part application which claims the benefit of U.S. application entitled “Systems and Methods for Intelligent Promotion Design in Bric and Mortar Retailers with Promotion Scoring,” U.S. application Ser. No. 16/120,178, filed Aug. 31, 2018, by Rapperport et al., (Attorney Docket No. EVS-1801-C1), pending, which is a continuation application and claims the benefit of U.S. application Ser. No. 15/990,005, filed May 25, 2018, of the same title, (Attorney Docket No. EVS-1801), abandoned, which is a continuation-in-part application and claims the benefit of U.S. application Ser. No. 14/209,851, filed Mar. 13, 2014, entitled “Architecture and Methods for Promotion Optimization,” by Moran (Attorney Docket No. EVS-1401), now U.S. Pat. No. 9,984,387 issued May 29, 2018, which claims priority under 35 U.S.C. 119(e) to a commonly owned U.S. Provisional Application No. 61/780,630, filed Mar. 13, 2013, entitled “Architecture and Methods for Promotion Optimization,” by Moran (Attorney Docket No. PRCO-P001P1), expired. application Ser. No. 15/990,005 also claims the benefit of U.S. Provisional Application No. 62/576,742, filed Oct. 25, 2017, entitled “Architecture and Methods for Generating Intelligent Offers with Dynamic Base Prices” by Rapperport et al. (Attorney Docket No. EVS-1703-P), expired. Additionally, U.S. application Ser. No. 16/120,178 claims priority to U.S. Provisional Application No. 62/553,133, filed Sep. 1, 2017, entitled “Systems and Methods for Promotion Optimization” by Rapperport et al. (Attorney Docket No. EVS-170X-P), expired.
• This continuation-in-part application also claims the benefit of U.S. application entitled “Systems and Methods for Price Testing and Optimization in Brick and Mortar Retailers,” U.S. application Ser. No. 16/157,018, filed Oct. 10, 2018, by Montero et al., (Attorney Docket No. EVS-1802), now U.S. Pat. No. 10,915,912 issued Feb. 9, 2021, and claims the benefit of U.S. application entitled “Systems and Methods for Collaborative Offer Generation,” U.S. application Ser. No. 16/216,997, filed Dec. 11, 2018, by Ventrice et al., (Attorney Docket No. EVS-1803), recently allowed.
• The present invention is additionally related to the following applications and patents, all of which are incorporated herein by reference:
• Commonly owned U.S. application Ser. No. 14/231,426, filed on Mar. 31, 2014, entitled “Adaptive Experimentation and Optimization in Automated Promotional Testing,” by Moran et al. (Attorney Docket No. EVS-1402), now U.S. Pat. No. 10,438,230 issued Oct. 8, 2019.
• Commonly owned U.S. application Ser. No. 14/231,432, filed on Mar. 31, 2014, entitled “Automated and Optimal Promotional Experimental Test Designs Incorporating Constraints,” by Moran et al. (Attorney Docket No. EVS-1403), now U.S. Pat. No. 9,940,639 issued Apr. 10, 2018.
• Commonly owned U.S. application Ser. No. 14/231,440, filed on Mar. 31, 2014, entitled “Automatic Offer Generation Using Concept Generator Apparatus and Methods Therefor,” by Moran et al. (Attorney Docket No. EVS-1404)), now U.S. Pat. No. 10,438,231, issued Oct. 8, 2019.
• Commonly owned U.S. application Ser. No. 14/231,442, filed on Mar. 31, 2014, entitled “Automated Event Correlation to Improve Promotional Testing,” by Moran et al. (Attorney Docket No. EVS-1405), now U.S. Pat. No. 9,940,640 issued Apr. 10, 2018.
• Commonly owned U.S. application Ser. No. 14/231,460, filed on Mar. 31, 2014, entitled “Automated Promotion Forecasting and Methods Therefor,” by Moran et al. (Attorney Docket No. EVS-1406)), now U.S. Pat. No. 10,445,763, issued Oct. 15, 2019.
• Commonly owned U.S. application Ser. No. 14/231,555, filed on Mar. 31, 2014, entitled “Automated Behavioral Economics Patterns in Promotion Testing and Methods Therefor,” by Moran et al. (Attorney Docket No. EVS-1407), now U.S. Pat. No. 10,140,629 issued Nov. 27, 2018.
• All the applications/patents listed above are incorporated herein in their entirety by this reference.
BACKGROUND
• The present invention relates generally to offer generation methods and apparatus therefor. More particularly, the present invention relates to computer-implemented methods and computer-implemented apparatus for the generation of promotions leveraging existing contracts a retailer enters into in order to determine the best promotional variable values.
• Promotion refers to various practices designed to increase sales of a particular product or services and/or the profit associated with such sales. Generally speaking, the public often associates promotion with the sale of consumer goods and services, including consumer packaged goods (e.g., food, home and personal care), consumer durables (e.g., consumer appliances, consumer electronics, automotive leasing), consumer services (e.g., retail financial services, health care, insurance, home repair, beauty and personal care), and travel and hospitality (e.g., hotels, airline flights, and restaurants). Promotion is particularly heavily involved in the sale of consumer packaged goods (e.g., consumer goods packaged for sale to an end consumer). However, promotion occurs in almost any industry that offers goods or services to a buyer (whether the buyer is an end consumer or an intermediate entity between the producer and the end consumer).
• The term promotion may refer to, for example, providing discounts (using for example a physical or electronic coupon or code) designed to, for example, promote the sales volume of a particular product or service. One aspect of promotion may also refer to the bundling of goods or services to create a more desirable selling unit such that sales volume may be improved. Another aspect of promotion may also refer to the merchandising design (with respect to looks, weight, design, color, etc.) or displaying of a particular product with a view to increasing its sales volume. It includes calls to action or marketing claims used in-store, on marketing collaterals, or on the package to drive demand. Promotions may be composed of all or some of the following: price based claims, secondary displays or aisle end-caps in a retail store, shelf signage, temporary packaging, placement in a retailer circular/flyer/coupon book, a colored price tag, advertising claims, or other special incentives intended to drive consideration and purchase behavior. These examples are meant to be illustrative and not limiting.
• In addition to promotional activities, it is also desirable to perform optimizations of base pricing (e.g. non-promotional prices). Often retailers rely upon manufacturer's suggested retail pricing (MSRP) for setting of base prices. In other circumstances, base prices are set based upon competitive analysis—a retailer may monitor competitor's and match or beat the competitor's price on some or all the goods in a store. Alternatively, some retailers may set a desired margin, or sales volume, for a good, and set prices accordingly. Generally however, the base prices of goods in a brick-and-mortar store do not vary significantly due to logistical concerns of updating signage and point of sales (POS) databases, consumer expectation of generally consistent base prices, and the tendency that a retailer will continue patterns of behavior (e.g., “this is what we have always done”).
• In discussing various embodiments of the present invention, the sale of consumer packaged goods (hereinafter “CPG”) is employed to facilitate discussion and ease of understanding. It should be kept in mind, however, that the promotion and base pricing optimization methods and apparatuses discussed herein may apply to any industry in which there is any pricing flexibility in the past or may be employed in the future.
• Further, price discount is employed as an example to explain the promotion methods and apparatuses herein. It should be understood, however, that promotion optimization may be employed to manipulate factors other than price discount in order to influence the sales volume. An example of such other factors may include the call to action on a display or on the packaging, the size of the CPG item, the manner in which the item is displayed or promoted or advertised either in the store or in media, etc.
• Generally speaking, it has been estimated that, on average, 17% of the revenue in the consumer packaged goods (CPG) industry is spent to fund various types of promotions, including discounts, designed to entice consumers to try and/or to purchase the packaged goods. In a typical example, the retailer (such as a grocery store) may offer a discount online or via a print circular to consumers. The promotion may be specifically targeted to an individual consumer (based on, for example, that consumer's demographics or past buying behavior). The discount may alternatively be broadly offered to the general public. Examples of promotions offered to general public include for example, a printed or electronic redeemable discount (e.g., coupon or code) for a specific CPG item. Another promotion example may include, for example, general advertising of the reduced price of a CPG item in a particular geographic area. Another promotion example may include in-store marking down of a particular CPG item only for a loyalty card user base.
• In an example, if the consumer redeems the coupon or electronic code, the consumer is entitled to a reduced price for the CPG item. The revenue loss to the retailer due to the redeemed discount may be reimbursed, wholly or partly, by the manufacturer of the CPG item in a separate transaction.
• Because promotion and base price testing is expensive (in terms of, for example, the effort to conduct a promotion campaign, modify display prices and/or the per-unit revenue loss to the retailer/manufacturer when the consumer decides to take advantage of the discount), efforts are continually made to minimize promotion cost while maximizing the return on promotion dollars investment. This effort is known in the industry as promotion optimization.
• For example, a typical promotion optimization method may involve examining the sales volume of a particular CPG item over time (e.g., weeks). The sales volume may be represented by a demand curve as a function of time, for example. A demand curve lift (excess over baseline) or dip (below baseline) for a particular time period would be examined to understand why the sales volume for that CPG item increases or decreases during such time period.
• FIG. 1 shows anexample demand curve102 for Brand X cookies over some period of time. Twolifts110 and114 and onedip112 indemand curve102 are shown in the example ofFIG. 1. Lift110 shows that the demand for Brand X cookies exceeds the baseline at least duringweek 2. By examining the promotion effort that was undertaken at that time (e.g., in the vicinity of weeks 1-4 or week 2) for Brand X cookies, marketers have in the past attempted to judge the effectiveness of the promotion effort on the sales volume. If the sales volume is deemed to have been caused by the promotion effort and delivers certain financial performance metrics, that promotion effort is deemed to have been successful and may be replicated in the future in an attempt to increase the sales volume. On the other hand,dip112 is examined in an attempt to understand why the demand falls off during that time (e.g.,weeks 3 and 4 inFIG. 1). If the decrease in demand was due to the promotion in week 2 (also known as consumer pantry loading or retailer forward-buying, depending on whether the sales volume shown reflects the sales to consumers or the sales to retailers), this decrease inweeks 3 and 4 should be counted against the effectiveness ofweek 2.
• One problem with the approach employed in the prior art has been the fact that the prior art approach is a backward-looking approach based on aggregate historical data. In other words, the prior art approach attempts to ascertain the nature and extent of the relationship between the promotion and the sales volume by examining aggregate data collected in the past. The use of historical data, while having some disadvantages (which are discussed later herein below), is not necessarily a problem. However, when such data is in the form of aggregate data (such as in simple terms of sales volume of Brand X cookies versus time for a particular store or geographic area), it is impossible to extract from such aggregate historical data all of the other factors that may more logically explain a particular lift or dip in the demand curve.
• To elaborate, current promotion and base price optimization approaches tend to evaluate sales lifts or dips as a function of four main factors: discount depth (e.g., how much was the discount on the CPG item), discount duration (e.g., how long did the promotion campaign last), timing (e.g., whether there was any special holidays or event or weather involved), and promotion type when analyzing for promotions (e.g., whether the promotion was a price discount only, whether Brand X cookies were displayed/not displayed prominently, whether Brand X cookies were features/not featured in the promotion literature).
• However, there may exist other factors that contribute to the sales lift or dip, and such factors are often not discoverable by examining, in a backward-looking manner, the historical aggregate sales volume data for Brand X cookies. This is because there is not enough information in the aggregate sales volume data to enable the extraction of information pertaining to unanticipated or seemingly unrelated events that may have happened during the sales lifts and dips and may have actually contributed to the sales lifts and dips.
• Suppose, for example, that there was a discount promotion for Brand X cookies during the time whenlift110 in thedemand curve102 happens. However, during the same time, there was a breakdown in the distribution chain of Brand Y cookies, a competitor's cookies brand which many consumers view to be an equivalent substitute for Brand X cookies. With Brand Y cookies being in short supply in the store, many consumers bought Brand X instead for convenience sake. Aggregate historical sales volume data for Brand X cookies, when examined after the fact in isolation by Brand X marketing department thousands of miles away, would not uncover that fact. As a result, Brand X marketers may make the mistaken assumption that the costly promotion effort of Brand X cookies was solely responsible for the sales lift and should be continued, despite the fact that it was an unrelated event that contributed to most of the lift in the sales volume of Brand X cookies.
• As another example, suppose, for example, that milk produced by a particular unrelated vendor was heavily promoted in the same grocery store or in a different grocery store nearby during the week that Brand X cookies experienced thesales lift110. The milk may have been highlighted in the weekly circular, placed in a highly visible location in the store and/or a milk industry expert may have been present in the store to push buyers to purchase milk, for example. Many consumers ended up buying milk because of this effort whereas some of most of those consumers who bought during the milk promotion may have waited another week or so until they finished consuming the milk they bought in the previous weeks. Further, many of those milk-buying consumers during this period also purchased cookies out of an ingrained milk-and-cookies habit. Aggregate historical sales volume data for Brand X cookies would not uncover that fact unless the person analyzing the historical aggregate sales volume data for Brand X cookies happened to be present in the store during that week and had the insight to note that milk was heavily promoted that week and also the insight that increased milk buying may have an influence on the sales volume of Brand X cookies.
• Software may try to take some of these unanticipated events into account but unless every SKU (stock keeping unit) in that store and in stores within commuting distance and all events, whether seemingly related or unrelated to the sales of Brand X cookies, are modeled, it is impossible to eliminate data noise from the backward-looking analysis based on aggregate historical sales data.
• Even without the presence of unanticipated factors, a marketing person working for Brand X may be interested in knowing whether the relativelymodest sales lift114 comes from purchases made by regular Brand X cookies buyers or by new buyers being enticed by some aspect of the promotion campaign to buy Brand X cookies for the first time. If Brand X marketer can ascertain that most of the lift in sales during the promotion period that spanslift114 comes from new consumers of Brand X cookies, such marketer may be willing to spend more money on the same type of sales promotion, even to the point of tolerating a negative ROI (return on investment) on his promotion dollars for this particular type of promotion since the recruitment of new buyers to a brand is deemed more much valuable to the company in the long run than the temporary increase in sales to existing Brand X buyers. Again, aggregate historical sales volume data for Brand X cookies, when examined in a backward-looking manner, would not provide such information.
• Furthermore, even if all unrelated and related events and factors can be modeled, the fact that the approach is backward-looking means that there is no way to validate the hypothesis about the effect an event has on the sales volume since the event has already occurred in the past. With respect to the example involving the effect of milk promotion on Brand X cookies sales, there is no way to test the theory short of duplicating the milk shortage problem again. Even if the milk shortage problem could be duplicated again for testing purposes, other conditions have changed, including the fact that most consumers who bought milk during that period would not need to or be in a position to buy milk again in a long time. Some factors, such as weather, cannot be duplicated, making theory verification challenging.
• Attempts have been made to employ non-aggregate sales data in promoting products. For example, some companies may employ a loyalty card program (such as the type commonly used in grocery stores or drug stores) to keep track of purchases by individual consumers. If an individual consumer has been buying sugar-free cereal, for example, the manufacturer of a new type of whole grain cereal may wish to offer a discount to that particular consumer to entice that consumer to try out the new whole grain cereal based on the theory that people who bought sugar-free cereal tend to be more health conscious and thus more likely to purchase whole grain cereal than the general cereal-consuming public. Such individualized discount may take the form of, for example, a redeemable discount such as a coupon or a discount code mailed or emailed to that individual.
• Some companies may vary the approach by, for example, ascertaining the items purchased by the consumer at the point of sale terminal and offering a redeemable code on the purchase receipt. Irrespective of the approach taken, the utilization of non-aggregate sales data has typically resulted in individualized offers, and has not been processed or integrated in any meaningful sense into a promotion optimization effort to determine the most cost-efficient, highest-return manner to promote a particular CPG item to the general public.
• Attempts have also been made to obtain from the consumers themselves indications of future buying behavior instead of relying on a backward-looking approach. For example, conjoint studies, one of the stated preference methods, have been attempted in which consumers are asked to state preferences. In an example conjoint study, a consumer may be approached at the store and asked a series of questions designed to uncover the consumer's future shopping behavior when presented with different promotions. Questions may be asked include, for example, “do you prefer Brand X or Brand Y” or “do you spend less than $100 or more than $100 weekly on grocery” or “do you prefer chocolate cookies or oatmeal cookies” or “do you prefer a 50-cent-off coupon or a 2-for-1 deal on cookies”. The consumer may state his preference on each of the questions posed (thus making this study a conjoint study on stated preference).
• However, such conjoint studies have proven to be an expensive way to obtain non-historical data. If the conjoint studies are presented via a computer, most users may ignore the questions and/or refuse to participate. If human field personnel are employed to talk to individual consumers to conduct the conjoint study, the cost of such studies tends to be quite high due to salary cost of the human field personnel and may make the extensive use of such conjoint studies impractical.
• Further and more importantly, it has been known that conjoint studies are somewhat unreliable in gauging actual purchasing behavior by consumers in the future. An individual may state out of guilt and the knowledge that he needs to lose weight that he will not purchase any cookies in the next six months, irrespective of discounts. In actuality, that individual may pick up a package of cookies every week if such package is carried in a certain small size that is less guilt-inducing and/or if the package of cookies is prominently displayed next to the milk refrigerator and/or if a 10% off discount coupon is available. If a promotion effort is based on such flawed stated preference data, discounts may be inefficiently deployed in the future, costing the manufacturer more money than necessary for the promotion.
• Finally, none of the approaches track the long-term impact of a promotion's effect on brand equity for an individual's buying behavior over time. Some promotions, even if deemed a success by traditional short-term measures, could have damaging long-term consequences. Increased price-based discounting, for example, can lead to consumers increasing the weight of price in determining their purchase decisions, making consumers more deal-prone and reluctant to buy at full price, leading to less loyalty to brands and retail outlets.
• It is therefore apparent that an urgent need exists for systems and methods that enable improvements in the generation of offers on products that have been contracted with the retailer.
SUMMARY
• To achieve the foregoing and in accordance with the present invention, systems and methods for contract based offer generation is provided. In some embodiments, a contract for a promotion is received by the retailer. Often these promotion contracts are periodically negotiated between the retailer and a distributor or manufacturer. Transaction logs of a retailer are accessed to determine offer variants, which are stored in an offer bank. These collected offers are then compared against the contracted offer and scored accordingly to select and administer a select set of offers. This causes improved sales (or other objective), and provides additional feedback data to further refine the best possible promotional activities.
• In some embodiments, a contract for a promotional offer on a product is received. Data is extracted from the contract. An offer band is accessed, and a plurality of test offers are selected from the offer bank by scoring each offer in the offer bank against the extracted data. The promotional offer and the selected plurality of test offers are deployed in a plurality of retail locations. This is done by maximizing orthogonality between the following variables: store sales, store out of stock rates, number of relevant SKUs carried in each store, temporal effects, discount depth, buy quantity and offer structure.
• The selecting the number of test offers to run in-market is done using reinforcement learning techniques, and in particular Thompson sampling. The offer bank is populated with forecasted offers, which are based upon transaction logs of a plurality of retailers. The transaction logs are adjusted for compliance by the given retailer, estimated out of stock events, normalized across stores to account for different store attributes, and adjusted for temporal effects. Machine learning is applied to the adjusted transaction logs to determine lift and standard deviation for a given test offer. The forecasts are a baseline function of time from the transaction log data plus elasticity from cross store experiments times a change in price, where in the elasticity is calculated as a function of the lift, and a confidence for the forecast is calculated as a function of the standard deviation.
• Note that the various features of the present invention described above may be practiced alone or in combination. These and other features of the present invention will be described in more detail below in the detailed description of the invention and in conjunction with the following figures.
BRIEF DESCRIPTION OF THE DRAWINGS
• In order that the present invention may be more clearly ascertained, some embodiments will now be described, by way of example, with reference to the accompanying drawings, in which:
• FIG. 1 shows anexample demand curve102 for Brand X cookies over some period of time;
• FIG. 2A shows, in accordance with an embodiment of the invention, a conceptual drawing of the forward-looking promotion optimization method;
• FIG. 2B shows, in accordance with an embodiment of the invention, the steps for generating a general public promotion;
• FIG. 3A shows in greater detail, in accordance with an embodiment of the invention, the administeringstep206 ofFIG. 2 from the user's perspective;
• FIG. 3B shows in greater detail, in accordance with an embodiment of the invention, the administeringstep206 ofFIG. 2 from the forward-looking promotion optimization system perspective;
• FIG. 4 shows various example segmentation criteria that may be employed to generate the purposefully segmented subpopulations;
• FIG. 5 shows various example methods for communicating the test promotions to individuals of the segmented subpopulations being tested;
• FIG. 6 shows, in accordance with some embodiments, various example promotion-significant responses;
• FIG. 7 shows, in accordance with some embodiments, various example test promotion variables affecting various aspects of a typical test promotion;
• FIG. 8 shows, in accordance with some embodiments, a general hardware/network view of a forward-looking promotion optimization system;
• FIG. 9 shows, in accordance with some embodiments, a block diagram of an offer collaboration architecture;
• FIG. 10 shows, in accordance with some embodiments, an example illustration of a contract based offer generation system;
• FIG. 11 shows, in accordance with some embodiments, a flowchart of an example method for the generation and administration of a contract based offer;
• FIG. 12 shows, in accordance with some embodiments, a flowchart of an example method for contract parsing;
• FIG. 13 shows, in accordance with some embodiments, a flowchart of an example method for offer generation;
• FIG. 14 shows, in accordance with some embodiments, an example screenshot of a contract based offer management system at the category level;
• FIG. 15 shows, in accordance with some embodiments, an example screenshot of a contract based offer management system at the product level;
• FIG. 16 shows, in accordance with some embodiments, an example screenshot of a contract based offer management system at a particularly promotional event;
• FIG. 17 shows, in accordance with some embodiments, an example screenshot of a contract based offer management system at for the results of execution of a particular promotion event;
• FIG. 18 shows, in accordance with some embodiments, an example screenshot of a contract based offer management system for editing a promotional event;
• FIG. 19 shows, in accordance with some embodiments, an example screenshot of a contract based offer management system illustrating the offer bank; and
• FIGS. 20A and 20B are example computer systems capable of implementing the systems for offer generation and administration.
DETAILED DESCRIPTION
• The present invention will now be described in detail with reference to several embodiments thereof as illustrated in the accompanying drawings. In the following description, numerous specific details are set forth in order to provide a thorough understanding of embodiments of the present invention. It will be apparent, however, to one skilled in the art, that embodiments may be practiced without some or all of these specific details. In other instances, well known process steps and/or structures have not been described in detail in order to not unnecessarily obscure the present invention. The features and advantages of embodiments may be better understood with reference to the drawings and discussions that follow.
• Aspects, features and advantages of exemplary embodiments of the present invention will become better understood with regard to the following description in connection with the accompanying drawing(s). It should be apparent to those skilled in the art that the described embodiments of the present invention provided herein are illustrative only and not limiting, having been presented by way of example only. All features disclosed in this description may be replaced by alternative features serving the same or similar purpose, unless expressly stated otherwise. Therefore, numerous other embodiments of the modifications thereof are contemplated as falling within the scope of the present invention as defined herein and equivalents thereto. Hence, use of absolute and/or sequential terms, such as, for example, “will,” “will not,” “shall,” “shall not,” “must,” “must not,” “first,” “initially,” “next,” “subsequently,” “before,” “after,” “lastly,” and “finally,” are not meant to limit the scope of the present invention as the embodiments disclosed herein are merely exemplary.
• The present invention relates to the generation of offers based upon contracts that are agreed upon between the retailer and another party. Often the other party is a distributor or the manufacturer of the product. In this application the term “manufacturer” may include the actual producer of a good, or could include resellers or intermediate branding entities. The term “retailer” refers to a business entity that is offering the good or service to consumers directly, or less commonly to yet another downstream business entity. Examples of manufacturers could include, for example, a soda bottling plant, a consumer package good (CPG) producer, or a wholesale retailer. Examples of retailers, in contrast include main-street stores, such as Target, Safeway, Walmart and the like.
• Historically, when a retailer and another party would contract for a given promotion, the retailer would implement the contracted offer across all applicable retail spaces. The present systems and methods provide an alternative structure, whereby offers that are likely to be more successful, yet are similar to the contracted offer, and deployed in some subgroup of the retail locations, allowing for better testing of offer formats and types, as well as maximization of a retailer's business goals.
• The following description of some embodiments will be provided in relation to numerous subsections. The use of subsections, with headings, is intended to provide greater clarity and structure to the present invention. In no way are the subsections intended to limit or constrain the disclosure contained therein. Thus, disclosures in any one section are intended to apply to all other sections, as is applicable.
I. Forward Looking Promotion Optimization
• Within the forward-looking promotion optimization, revealed preferences are obtained when the individual consumers respond to specifically designed actual test promotions. The revealed preferences may be tracked in individual computer-implemented accounts (which may, for example, be implemented via a record in a centralized database and rendered accessible to the merchant or the consumer via a computer network such as the internet) associated with individual consumers, or may be collected at a physical retailer based upon transaction records. For example, when a consumer responds, using his smart phone, web browser, or in a physical store through completion of a transaction, to a test promotion that offers 20% off a particular consumer packaged goods (CPG) item, that response is tracked in his individual computer-implemented account, or in a transaction record. Such computer-implemented accounts may be implemented via, for example, a loyalty card program, apps on a smart phone, computerized records, social media news feed, etc.
• In one or more embodiments, a plurality of test promotions may be designed and tested on a plurality of groups of consumers (the groups of consumers are referred to herein as “subpopulations”). The responses by the consumers are recorded and analyzed, with the analysis result employed to generate additional test promotions or to formulate the general population promotion. In the event of physical testing in a retailer space, it may be possible to segment the consumer base using a variety of collected demographic and activity data. This may include a loyalty care program, pharmacy ID, and information collected from public data sources. Such information may be correlated to credit card (or debit card, electronic wallet, etc.) information, and stored as a profile for the individual consumer and aggregated into consumer households. Information about the consumer, or the household, such as purchasing behaviors, income levels, ethnicity, age(s), number of persons in the household, gender, political affiliations, geography, medical conditions, etc., may be used to categorize the consumers into like subpopulations using neural network techniques and known clustering algorithms.
• As will be discussed later herein, if the consumer actually redeems the offer, one type of response is recorded and noted in the computer-implemented account of that consumer. Even if an action by the consumer does not involve actually redeeming or actually taking advantage of the promotional offer right away, an action by that consumer may, however, constitute a response that indicates a level of interest or lack of interest and may still be useful in revealing the consumer preference (or lack thereof). For example, if a consumer saves an electronic coupon (offered as part of a test promotion) in his electronic coupon folder or forwards that coupon to a friend via an email or a social website, that action may indicate a certain level of interest and may be useful in determining the effectiveness of a given test promotion. In the physical retailer space, if a consumer stops to look at a product, or even pick up the product but chooses not to purchase it at the register, such activity, to the extent it is reliably measured, may indicate interest in the promotion despite the lack of a transaction being completed. Different types of responses/actions by the consumers may be accorded different weights, in one or more embodiments.
• As noted, the groups of consumers involved in promotion testing represent segments of the public that have been purposefully segmented in accordance with segmenting criteria specifically designed for the purpose of testing the test promotions. As the term is employed herein, a subpopulation is deemed purposefully segmented when its members are selected based on criteria other than merely to make up a given number of members in the subpopulation. Demographics, buying behavior, behavioral economics, geography (e.g., purchasing at a certain brick and mortar retailer) are example criteria that may be employed to purposefully segment a population into subpopulations for promotion testing. In an example, a segmented population may number in the tens or hundreds or even thousands of individuals. In contrast, the general public may involve tens of thousands, hundreds of thousands, or millions of potential customers.
• By purposefully segmenting the public into small subpopulations for promotion testing, embodiments of the invention can exert control over variables such as demographics (e.g., age, income, sex, marriage status, address, etc.), buying behavior (e.g., regular purchaser of Brand X cookies, consumer of premium food, frequent traveler, etc.), weather, shopping habits, life style, and/or any other criteria suitable for use in creating the subpopulations. More importantly, the subpopulations are kept small such that multiple test promotions may be executed on different subpopulations, either simultaneously or at different times, without undue cost or delay in order to obtain data pertaining to the test promotion response behavior. The low cost/low delay aspect of creating and executing test promotions on purposefully segmented subpopulations permits, for example, what-if testing, testing in statistically significant numbers of tests, and/or iterative testing to isolate winning features in test promotions.
• Generally speaking, each individual test promotion may be designed to test one or more test promotion variables. These test promotions variables may relate to, for example, the size, shape, color, manner of display, manner of discount, manner of publicizing, manner of dissemination pertaining to the goods/services being promoted.
• As a very simple example, one test promotion may involve 12-oz packages of fancy-cut potato chips with medium salt and a discount of 30% off the regular price. This test promotion may be tested on a purposefully segmented subpopulation of 35-40 years old professionals in the $30,000-$50,000 annual income range. Another test promotion may involve the same 30% discount 12-oz packages of fancy-cut potato chips with medium salt on a different purposefully segmented subpopulation of 35-40 years old professionals in the higher $100,000-$150,000 annual income range. By controlling all variables except for income range, the responses of these two test promotions, if repeated in statistically significant numbers, would likely yield fairly accurate information regarding the relationship between income for 35-40 years old professionals and their actual preference for 12-oz packages of fancy cut potato chips with medium salt.
• In designing different test promotions, one or more of the test promotions variables may vary or one or more of the segmenting criteria employed to create the purposefully segmented subpopulations may vary. The test promotion responses from individuals in the subpopulations are then collected and analyzed to ascertain which test promotion or test promotion variable(s) yields/yield the most desirable response (based on some predefined success criteria, for example).
• Further, the test promotions can also reveal insights regarding which subpopulation performs the best, or well, with respect to test promotion responses. In this manner, test promotion response analysis provides insights not only regarding the relative performance of the test promotion and/or test promotion variable but also insights regarding population segmentation and/or segmentation criteria. In an embodiment, it is contemplated that the segments may be arbitrarily or randomly segmented into groups and test promotions may be executed against these arbitrarily segmented groups in order to obtain insights regarding personal characteristics that respond well to a particular type of promotion.
• In an embodiment, the identified test promotion variable(s) that yield the most desirable responses may then be employed to formulate a general public promotion (GPP), which may then be offered to the larger public. A general public promotion is different from a test promotion in that a general public promotion is a promotion designed to be offered to members of the public to increase or maximize sales or profit whereas a test promotion is designed to be targeted to a small group of individuals fitting a specific segmentation criteria for the purpose of promotion testing. Examples of general public promotions include (but not limited to) advertisement printed in newspapers, release in public forums and websites, flyers for general distribution, announcement on radios or television, promotion broadly transmitted or made available to members of the public, and/or promotions that are rolled out to a wider set of physical retailer locations. The general public promotion may take the form of a paper or electronic circular that offers the same promotion to the larger public, for example.
• Alternatively or additionally, promotion testing may be iterated over and over with different subpopulations (segmented using the same or different segmenting criteria) and different test promotions (devised using the same or different combinations of test promotion variables) in order to validate one or more the test promotion response analysis result(s) prior to the formation of the generalized public promotion. In this manner, “false positives” may be reduced.
• Since a test promotion may involve many test promotion variables, iterative test promotion testing, as mentioned, may help pin-point a variable (e.g., promotion feature) that yields the most desirable test promotion response to a particular subpopulation or to the general public.
• Suppose, for example, that a manufacturer wishes to find out the most effective test promotion for packaged potato chips. One test promotion may reveal that consumers tend to buy a greater quantity of potato chips when packaged in brown paper bags versus green paper bags. That “winning” test promotion variable value (e.g., brown paper bag packaging) may be retested in another set of test promotions using different combinations of test promotion variables (such as for example with different prices, different display options, etc.) on the same or different purposefully segmented subpopulations. The follow-up test promotions may be iterated multiple times in different test promotion variable combinations and/or with different test subpopulations to validate that there is, for example, a significant consumer preference for brown paper bag packaging over other types of packaging for potato chips.
• Further, individual “winning” test promotion variable values from different test promotions may be combined to enhance the efficacy of the general public promotion to be created. For example, if a 2-for-1 discount is found to be another winning variable value (e.g., consumers tend to buy a greater quantity of potato chips when offered a 2-for-1 discount), that winning test promotion variable value (e.g., the aforementioned 2-for-1 discount) of the winning test promotion variable (e.g., discount depth) may be combined with the brown paper packaging winning variable value to yield a promotion that involves discounting 2-for-1 potato chips in brown paper bag packaging.
• The promotion involving discounting 2-for-1 potato chips in brown paper bag packaging may be tested further to validate the hypothesis that such a combination elicits a more desirable response than the response from test promotions using only brown paper bag packaging or from test promotions using only 2-for-1 discounts. As many of the “winning” test promotion variable values may be identified and combined in a single promotion as desired. At some point, a combination of “winning” test promotion variables (involving one, two, three, or more “winning” test promotion variables) may be employed to create the general public promotion, in one or more embodiments.
• In one or more embodiments, test promotions may be executed iteratively and/or in a continual fashion on different purposefully segmented subpopulations using different combinations of test promotion variables to continue to obtain insights into consumer actual revealed preferences, even as those preferences change over time. Note that the consumer responses that are obtained from the test promotions are actual revealed preferences instead of stated preferences. In other words, the data obtained from the test promotions administered in accordance with embodiments of the invention pertains to what individual consumers actually do when presented with the actual promotions. The data is tracked and available for analysis and/or verification in individual computer-implemented accounts of individual consumers involved in the test promotions. This revealed preference approach is opposed to a stated preference approach, which stated preference data is obtained when the consumer states what he would hypothetically do in response to, for example, a hypothetically posed conjoint test question.
• As such, the actual preference test promotion response data obtained in accordance with embodiments of the present invention is a more reliable indicator of what a general population member may be expected to behave when presented with the same or a similar promotion in a general public promotion. Accordingly, there is a closer relationship between the test promotion response behavior (obtained in response to the test promotions) and the general public response behavior when a general public promotion is generated based on such test promotion response data.
• Also, the lower face validity of a stated preference test, even if the insights have statistical relevance, poses a practical challenge; CPG manufacturers who conduct such tests have to then communicate the insights to a retailer in order to drive real-world behavior, and convincing retailers of the validity of these tests after the fact can lead to lower credibility and lower adoption, or “signal loss” as the top concepts from these tests get re-interpreted by a third party, the retailer, who wasn't involved in the original test design.
• It should be pointed out that embodiments of the inventive test promotion optimization methods and apparatuses disclosed herein operate on a forward-looking basis in that the plurality of test promotions are generated and tested on segmented subpopulations in advance of the formulation of a general public promotion. In other words, the analysis results from executing the plurality of test promotions on different purposefully segmented subpopulations are employed to generate future general public promotions. In this manner, data regarding the “expected” efficacy of the proposed general public promotion is obtained even before the proposed general public promotion is released to the public. This is one key driver in obtaining highly effective general public promotions at low cost.
• Furthermore, the subpopulations can be generated with highly granular segmenting criteria, allowing for control of data noise that may arise due to a number of factors, some of which may be out of the control of the manufacturer or the merchant. This is in contrast to the aggregated data approach of the prior art.
• For example, if two different test promotions are executed on two subpopulations shopping at the same merchant on the same date, variations in the response behavior due to time of day or traffic condition are essentially eliminated or substantially minimized in the results (since the time or day or traffic condition would affect the two subpopulations being tested in substantially the same way).
• The test promotions themselves may be formulated to isolate specific test promotion variables (such as the aforementioned potato chip brown paper packaging or the 16-oz size packaging). This is also in contrast to the aggregated data approach of the prior art.
• Accordingly, individual winning promotion variables may be isolated and combined to result in a more effective promotion campaign in one or more embodiments. Further, the test promotion response data may be analyzed to answer questions related to specific subpopulation attribute(s) or specific test promotion variable(s). With embodiments of the invention, it is now possible to answer, from the test subpopulation response data, questions such as “How deep of a discount is required to increase by 10% the volume of potato chip purchased by buyers who are 18-25 year-old male shopping on a Monday?” or to generate test promotions specifically designed to answer such a question. Such data granularity and analysis result would have been impossible to achieve using the backward-looking, aggregate historical data approach of the prior art.
• In one or more embodiments, there is provided a promotional idea module for generating ideas for promotional concepts to test. The promotional idea generation module relies on a series of pre-constructed sentence structures that outline typical promotional constructs. For example, Buy X, get Y for $Z price would be one sentence structure, whereas Get Y for $Z when you buy X would be a second. It's important to differentiate that the consumer call to action in those two examples is materially different, and one cannot assume the promotional response will be the same when using one sentence structure vs. another. The solution is flexible and dynamic, so once X, Y, and Z are identified, multiple valid sentence structures can be tested. Additionally, other variables in the sentence could be changed, such as replacing “buy” with “hurry up and buy” or “act now” or “rush to your local store to find”. The solution delivers a platform where multiple products, offers, and different ways of articulating such offers can be easily generated by a lay user. The amount of combinations to test can be infinite. Further, the generation may be automated, saving time and effort in generating promotional concepts. In following sections one mechanism, the design matrix, for the automation of promotional generation will be provided in greater detail.
• In one or more embodiments, once a set of concepts is developed, the technology advantageously a) will constrain offers to only test “viable promotions”, e.g., those that don't violate local laws, conflict with branding guidelines, lead to unprofitable concepts that wouldn't be practically relevant, can be executed on a retailers' system, etc., and/or b) link to the design of experiments for micro-testing to determine which combinations of variables to test at any given time.
• In one or more embodiments, there is provided an offer selection module for enabling a non-technical audience to select viable offers for the purpose of planning traditional promotions (such as general population promotion, for example) outside the test environment. By using filters and advanced consumer-quality graphics, the offer selection module will be constrained to only show top performing concepts from the tests, with production-ready artwork wherever possible. By doing so, the offer selection module renders irrelevant the traditional, Excel-based or heavily numbers-oriented performance reports from traditional analytic tools. The user can have “freedom within a framework” by selecting any of the pre-scanned promotions for inclusion in an offer to the general public, but value is delivered to the retailer or manufacturer because the offers are constrained to only include the best performing concepts. Deviation from the top concepts can be accomplished, but only once the specific changes are run through the testing process and emerge in the offer selection windows.
• In an embodiment, it is expressly contemplated that the general population and/or subpopulations may be chosen from social media site (e.g., Facebook™, Twitter™, Google+™, etc.) participants. Social media offers a large population of active participants and often provide various communication tools (e.g., email, chat, conversation streams, running posts, etc.) which makes it efficient to offer promotions and to receive responses to the promotions. Various tools and data sources exist to uncover characteristics of social media site members, which characteristics (e.g., age, sex, preferences, attitude about a particular topic, etc.) may be employed as highly granular segmentation criteria, thereby simplifying segmentation planning.
• Although grocery stores and other brick-and-mortar businesses are discussed in various examples herein, it is expressly contemplated that embodiments of the invention apply also to online shopping and online advertising/promotion and online members/customers.
• These and other features and advantages of embodiments of the invention may be better understood with reference to the figures and discussions that follow.
• FIG. 2A shows, in accordance with an embodiment of the invention, a conceptual drawing of the forward-looking promotion optimization method. As shown in
• FIG. 2A, a plurality of test promotions102a,102b,102c,102d, and102eare administered to purposefully segmented subpopulations104a,104b,104c,104d, and104erespectively. As mentioned, each of the test promotions (102a-102e) may be designed to test one or more test promotion variables.
• In the example ofFIG. 2A,test promotions102a-102dare shown testing three test promotion variables X, Y, and Z, which may represent for example the size of the packaging (e.g., 12 oz versus 16 oz), the manner of display (e.g., at the end of the aisle versus on the shelf), and the discount (e.g., 10% off versus 2-for-1). These promotion variables are of course only illustrative and almost any variable involved in producing, packaging, displaying, promoting, discounting, etc. of the packaged product may be deemed a test promotion variable if there is an interest in determining how the consumer would respond to variations of one or more of the test promotion variables. Further, although only a few test promotion variables are shown in the example ofFIG. 2A, a test promotion may involve as many or as few of the test promotion variables as desired. For example, test promotion102eis shown testing four test promotion variables (X, Y, Z, and T).
• One or more of the test promotion variables may vary from test promotion to test promotion. In the example ofFIG. 2A, test promotion102ainvolves test variable X1 (representing a given value or attribute for test variable X) while test promotion102binvolves test variable X2 (representing a different value or attribute for test variable X). A test promotion may vary, relative to another test promotion, one test promotion variable (as can be seen in the comparison between test promotions102aand102b) or many of the test promotion variables (as can be seen in the comparison between test promotions102aand102d). Also, there are no requirements that all test promotions must have the same number of test promotion variables (as can be seen in the comparison between test promotions102aand102e) although for the purpose of validating the effect of a single variable, it may be useful to keep the number and values of other variables (e.g., the control variables) relatively constant from test to test (as can be seen in the comparison between test promotions102aand102b).
• Generally speaking, the test promotions may be generated using automated testpromotion generation software110, which varies for example the test promotion variables and/or the values of the test promotion variables and/or the number of the test promotion variables to come up with different test promotions.
• In the example ofFIG. 2A, purposefully segmented subpopulations104a-104dare shown segmented using four segmentation criteria A, B, C, D, which may represent for example the age of the consumer, the household income, the zip code, group of consumers shopping at a particular physical retailer, and whether the person is known from past purchasing behavior to be a luxury item buyer or a value item buyer. These segmentation criteria are of course only illustrative and almost any demographics, behavioral, attitudinal, whether self-described, objective, interpolated from data sources (including past purchase or current purchase data), etc. may be used as segmentation criteria if there is an interest in determining how a particular subpopulation would likely respond to a test promotion. Further, although only a few segmentation criteria are shown in connection with subpopulations104a-104din the example ofFIG. 2A, segmentation may involve as many or as few of the segmentation criteria as desired. For example, purposefully segmented subpopulation104eis shown segmented using five segmentation criteria (A, B, C, D, and E).
• In the present disclosure, a distinction is made between a purposefully segmented subpopulation and a randomly segmented subpopulation. The former denotes a conscious effort to group individuals based on one or more segmentation criteria or attributes. The latter denotes a random grouping for the purpose of forming a group irrespective of the attributes of the individuals. Randomly segmented subpopulations are useful in some cases; however they are distinguishable from purposefully segmented subpopulations when the differences are called out.
• One or more of the segmentation criteria may vary from purposefully segmented subpopulation to purposefully segmented subpopulation. In the example ofFIG. 2A, purposefully segmented subpopulation104ainvolves segmentation criterion value A1 (representing a given attribute or range of attributes for segmentation criterion A) while purposefully segmented subpopulation104cinvolves segmentation criterion value A2 (representing a different attribute or set of attributes for the same segmentation criterion A).
• As can be seen, different purposefully segmented subpopulation may have different numbers of individuals. As an example, purposefully segmented subpopulation104ahas four individuals (P1-P4) whereas purposefully segmented subpopulation104ehas six individuals (P17-P22). A purposefully segmented subpopulation may differ from another purposefully segmented subpopulation in the value of a single segmentation criterion (as can be seen in the comparison between purposefully segmented subpopulation104aand purposefully segmented subpopulation104cwherein the attribute A changes from A1 to A2) or in the values of many segmentation criteria simultaneously (as can be seen in the comparison between purposefully segmented subpopulation104aand purposefully segmented subpopulation104dwherein the values for attributes A, B, C, and D are all different). Two purposefully segmented subpopulations may also be segmented identically (e.g., using the same segmentation criteria and the same values for those criteria) as can be seen in the comparison between purposefully segmented subpopulation104aand purposefully segmented subpopulation104b.
• Also, there are no requirements that all purposefully segmented subpopulations must be segmented using the same number of segmentation criteria (as can be seen in the comparison between purposefully segmented subpopulation104aand104ewherein purposefully segmented subpopulation104eis segmented using five criteria and purposefully segmented subpopulation104ais segmented using only four criteria) although for the purpose of validating the effect of a single criterion, it may be useful to keep the number and values of other segmentation criteria (e.g., the control criteria) relatively constant from purposefully segmented subpopulation to purposefully segmented subpopulation.
• Generally speaking, the purposefully segmented subpopulations may be generated usingautomated segmentation software112, which varies for example the segmentation criteria and/or the values of the segmentation criteria and/or the number of the segmentation criteria to come up with different purposefully segmented subpopulations.
• In one or more embodiments, the test promotions are administered to individual users in the purposefully segmented subpopulations in such a way that the responses of the individual users in that purposefully segmented subpopulation can be recorded for later analysis. As an example, an electronic coupon may be presented in an individual user's computer-implemented account (e.g., shopping account or loyalty account), or emailed or otherwise transmitted to the smart phone of the individual. In an example, the user may be provided with an electronic coupon on his smart phone that is redeemable at the merchant. InFIG. 2A, this administering is represented by the lines that extend from test promotion102ato each of individuals P1-P4 in purposefully segmented subpopulation104a. If the user (such as user P1) makes a promotion-significant response, the response is noted indatabase130.
• A promotion-significant response is defined as a response that is indicative of some level of interest or disinterest in the goods/service being promoted. In the aforementioned example, if the user P1 redeems the electronic coupon at the store, the redemption is strongly indicative of user P1's interest in the offered goods. However, responses falling short of actual redemption or actual purchase may still be significant for promotion analysis purposes. For example, if the user saves the electronic coupon in his electronic coupon folder on his smart phone, such action may be deemed to indicate a certain level of interest in the promoted goods. As another example, if the user forwards the electronic coupon to his friend or to a social network site, such forwarding may also be deemed to indicate another level of interest in the promoted goods. As another example, if the user quickly moves the coupon to trash, this action may also indicate a level of strong disinterest in the promoted goods. In one or more embodiments, weights may be accorded to various user responses to reflect the level of interest/disinterest associated with the user's responses to a test promotion. For example, actual redemption may be given a weight of 1, whereas saving to an electronic folder would be given a weight of only 0.6 and whereas an immediate deletion of the electronic coupon would be given a weight of −0.5.
• Analysis engine132 represents a software engine for analyzing the consumer responses to the test promotions. Response analysis may employ any analysis technique (including statistical analysis) that may reveal the type and degree of correlation between test promotion variables, subpopulation attributes, and promotion responses.Analysis engine132 may, for example, ascertain that a certain test promotion variable value (such as 2-for-1 discount) may be more effective than another test promotion variable (such as 25% off) for 32-oz soft drinks if presented as an electronic coupon right before Monday Night Football. Such correlation may be employed to formulate a general population promotion (150) by a general promotion generator software (160). As can be appreciated from this discussion sequence, the optimization is a forward-looking optimization in that the results from test promotions administered in advance to purposefully segmented subpopulations are employed to generate a general promotion to be released to the public at a later date.
• In one or more embodiments, the correlations ascertained byanalysis engine132 may be employed to generate additional test promotions (arrows172,174, and176) to administer to the same or a different set of purposefully segmented subpopulations. The iterative testing may be employed to verify the consistency and/or strength of a correlation (by administering the same test promotion to a different purposefully segmented subpopulation or by combining the “winning” test promotion value with other test promotion variables and administering the re-formulated test promotion to the same or a different set of purposefully segmented subpopulations).
• In one or more embodiments, a “winning” test promotion value (e.g., 20% off listed price) from one test promotion may be combined with another “winning” test promotion value (e.g., packaged in plain brown paper bags) from another test promotion to generate yet another test promotion. The test promotion that is formed from multiple “winning” test promotion values may be administered to different purposefully segmented subpopulations to ascertain if such combination would elicit even more desirable responses from the test subjects.
• Since the purposefully segmented subpopulations are small and may be segmented with highly granular segmentation criteria, a large number of test promotions may be generated (also with highly granular test promotion variables) and a large number of combinations of test promotions/purposefully segmented subpopulations can be executed quickly and at a relatively low cost. The same number of promotions offered as general public promotions would have been prohibitively expensive to implement, and the large number of failed public promotions would have been costly for the manufacturers/retailers. In contrast, if a test promotion fails, the fact that the test promotion was offered to only a small number of consumers in one or more segmented subpopulations, or a limited number of physical locations for a limited time, would limit the cost of failure. Thus, even if a large number of these test promotions “fail” to elicit the desired responses, the cost of conducting these small test promotions would still be quite small.
• In an embodiment, it is envisioned that dozens, hundreds, or even thousands of these test promotions may be administered concurrently or staggered in time to the dozens, hundreds or thousands of segmented subpopulations. Further, the large number of test promotions executed (or iteratively executed) improves the statistical validity of the correlations ascertained by analysis engine. This is because the number of variations in test promotion variable values, subpopulation attributes, etc. can be large, thus yielding rich and granulated result data. The data-rich results enable the analysis engine to generate highly granular correlations between test promotion variables, subpopulation attributes, and type/degree of responses, as well as track changes over time. In turn, these more accurate/granular correlations help improve the probability that a general public promotion created from these correlations would likely elicit the desired response from the general public. It would also, over, time, create promotional profiles for specific categories, brands, retailers, and individual shoppers where, e.g.,shopper1 prefers contests andshopper2 prefers instant financial savings.
• FIG. 2B shows, in accordance with an embodiment of the invention, the steps for generating a general public promotion. In one or more embodiments, each, some, or all the steps ofFIG. 2B may be automated via software to automate the forward-looking promotion optimization process. Instep202, the plurality of test promotions are generated. These test promotions have been discussed in connection withtest promotions102a-102eofFIG. 2A and represent the plurality of actual promotions administered to small purposefully segmented subpopulations to allow the analysis engine to uncover highly accurate/granular correlations between test promotion variables, subpopulation attributes, and type/degree of responses in an embodiment, these test promotions may be generated using automated test promotion generation software that varies one or more of the test promotion variables, either randomly, according to heuristics, and/or responsive to hypotheses regarding correlations fromanalysis engine132 for example.
• Instep204, the segmented subpopulations are generated. In an embodiment, the segmented subpopulations represent randomly segmented subpopulations. In another embodiment, the segmented subpopulations represent purposefully segmented subpopulations. In another embodiment, the segmented subpopulations may represent a combination of randomly segmented subpopulations and purposefully segmented subpopulations. In an embodiment, these segmented subpopulations may be generated using automated subpopulation segmentation software that varies one or more of the segmentation criteria, either randomly, according to heuristics, and/or responsive to hypotheses regarding correlations fromanalysis engine132, for example.
• Instep206, the plurality of test promotions generated instep202 are administered to the plurality of segmented subpopulations generated instep204. In an embodiment, the test promotions are administered to individuals within the segmented subpopulation and the individual responses are obtained and recorded in a database (step208).
• In an embodiment, automated test promotion software automatically administers the test promotions to the segmented subpopulations using electronic contact data that may be obtained in advance from, for example, social media sites, a loyalty card program, previous contact with individual consumers, or potential consumer data purchased from a third party, etc. In some alternate embodiments, as will be discussed in greater detail below, the test promotions may be administered via electronic pricing tags displayed within a physical retail location. Such physical test promotions may be constricted by deployment time due to logistic considerations. The responses may be obtained at the point of sale terminal, or via a website or program, via social media, or via an app implemented on smart phones used by the individuals, for example.
• Instep210, the responses are analyzed to uncover correlations between test promotion variables, subpopulation attributes, and type/degree of responses.
• Instep212, the general public promotion is formulated from the correlation data, which is uncovered by the analysis engine from data obtained via subpopulation test promotions. In an embodiment, the general public promotion may be generated automatically using public promotion generation software which utilizes at least the test promotion variables and/or subpopulation segmentation criteria and/or test subject responses and/or the analysis provided byanalysis engine132.
• Instep214, the general public promotion is released to the general public to promote the goods/services.
• In one or more embodiments, promotion testing using the test promotions on the segmented subpopulations occurs in parallel to the release of a general public promotion and may continue in a continual fashion to validate correlation hypotheses and/or to derive new general public promotions based on the same or different analysis results. If iterative promotion testing involving correlation hypotheses uncovered byanalysis engine132 is desired, the same test promotions or new test promotions may be generated and executed against the same segmented subpopulations or different segmented subpopulations as needed (paths216/222/226 or216/224/226 or216/222/224/226). As mentioned, iterative promotion testing may validate the correlation hypotheses, serve to eliminate “false positives” and/or uncover combinations of test promotion variables that may elicit even more favorable or different responses from the test subjects.
• Promotion testing may be performed on an on-going basis using the same or different sets of test promotions on the same or different sets of segmented subpopulations as mentioned (paths218/222/226 or218/224/226 or218/222/224/226 or220/222/226 or220/224/226 or220/222/224/226).
• FIG. 3A shows in greater detail, in accordance with an embodiment of the invention, the administeringstep206 ofFIG. 2 from the user's perspective. Instep302, the test promotion is received from the test promotion generation server (which executes the software employed to generate the test promotion). As examples, the test promotion may be received at a user's smart phone or tablet (such as in the case of an electronic coupon or a discount code, along with the associated promotional information pertaining to the product, place of sale, time of sale, etc.), in a computer-implemented account (such as a loyalty program account) associated with the user that is a member of the segmented subpopulation to be tested, via one or more social media sites, or displayed on electronic pricing tags within a retailer's physical store. Instep304, the test promotion is presented to the user. Instep306, the user's response to the test promotion is obtained and transmitted to a database for analysis.
• FIG. 3B shows in greater detail, in accordance with an embodiment of the invention, the administeringstep206 ofFIG. 2 from the forward-looking promotion optimization system perspective. Instep312, the test promotions are generated using the test promotion generation server (which executes the software employed to generate the test promotion). Instep314, the test promotions are provided to the users (e.g., transmitted or emailed to the user's smart phone or tablet or computer, shared with the user using the user's loyalty account, displayed in the physical retailer). Instep316, the system receives the user's responses and stores the user's responses in the database for later analysis.
• FIG. 4 shows various example segmentation criteria that may be employed to generate the purposefully segmented subpopulations. As show inFIG. 4, demographics criteria (e.g., sex, location, household size, household income, etc.), buying behavior (category purchase index, most frequent shopping hours, value versus premium shopper, etc.), past/current purchase history, channel (e.g., stores frequently shopped at, competitive catchment of stores within driving distance), behavioral economics factors, etc. can all be used to generate with a high degree of granularity the segmented subpopulations. The examples ofFIG. 4 are meant to be illustrative and not meant to be exhaustive or limiting. As mentioned, one or more embodiments of the invention generate the segmented subpopulations automatically using automated population segmentation software that generates the segmented subpopulations based on values of segmentation criteria.
• FIG. 5 shows various example methods for communicating the test promotions to individuals of the segmented subpopulations being tested. As shown inFIG. 5, the test promotions may be mailed to the individuals, emailed in the form of text or electronic flyer or coupon or discount code, displayed on a webpage when the individual accesses his shopping or loyalty account via a computer or smart phone or tablet, and lastly display on an electronic pricing tag within a retailer's store. Redemption may take place using, for example, a printed coupon (which may be mailed or may be printed from an electronic version of the coupon) at the point of sale terminal, an electronic version of the coupon (e.g., a screen image or QR code), the verbal providing or manual entry of a discount code into a terminal at the store or at the point of sale, or purchase of an item in a physical location that has the promotion displayed. The examples ofFIG. 5 are meant to be illustrative and not meant to be exhaustive or limiting. One or more embodiments of the invention automatically communicate the test promotions to individuals in the segmented subpopulations using software that communicates/email/mail/administer the test promotions automatically. In this manner, subpopulation test promotions may be administered automatically, which gives manufacturers and retailers the ability to generate and administer a large number of test promotions with low cost/delay.
• FIG. 6 shows, in accordance with an embodiment, various example promotion-significant responses. As mentioned, redemption of the test offer is one strong indication of interest in the promotion. However, other consumer actions responsive to the receipt of a promotion may also reveal the level of interest/disinterest and may be employed by the analysis engine to ascertain which test promotion variable is likely or unlikely to elicit the desired response. Examples shown inFIG. 6 include redemption (strong interest), deletion of the promotion offer (low interest), save to electronic coupon folder (mild to strong interest), clicked to read further (mild interest), forwarding to self or others or social media sites (mild to strong interest), stopping to look at an item within the store (mild interest), and picking up the item in a physical store but ultimately not purchasing the item (strong interest). As mentioned, weights may be accorded to various consumer responses to allow the analysis engine to assign scores and provide user-interest data for use in formulating follow-up test promotions and/or in formulating the general public promotion. For example, low interest may be afforded a score of −0.75 to −0.25, mild interest could be afforded a score weight of 0.1-0.5, strong interest may be afforded a score of 0.5-0.8, and purchase of the product may be afforded a score of 1. The examples ofFIG. 6 are meant to be illustrative and not meant to be exhaustive or limiting.
• FIG. 7 shows, in accordance with an embodiment of the invention, various example test promotion variables affecting various aspects of a typical test promotion. As shown inFIG. 7, example test promotion variables include price, discount action (e.g., save 10%, save $1, 2-for-1 offer, etc.), artwork (e.g., the images used in the test promotion to draw interest), brand (e.g., brand X potato chips versus brand Y potato chips), pricing tier (e.g., premium, value, economy), size (e.g., 32 oz, 16 oz, 8 oz), packaging (e.g., single, 6-pack, 12-pack, paper, can, etc.), channel (e.g., email versus paper coupon versus notification in loyalty account). The examples ofFIG. 7 are meant to be illustrative and not meant to be exhaustive or limiting. As mentioned, one or more embodiments of the invention involve generating the test promotions automatically using automated test promotion generation software by varying one or more of the test promotion variables, either randomly or based on feedback from the analysis of other test promotions or from the analysis of the general public promotion.
• FIG. 8 shows, in accordance with an embodiment of the invention, a general hardware/network view of the forward-lookingpromotion optimization system800. In general, the various functions discussed may be implemented as software modules, which may be implemented in one or more servers (including actual and/or virtual servers). InFIG. 8, there is shown a testpromotion generation module802 for generating the test promotions in accordance with test promotion variables. There is also shown apopulation segmentation module804 for generating the segmented subpopulations in accordance with segmentation criteria. There is also shown a testpromotion administration module806 for administering the plurality of test promotions to the plurality of segmented subpopulations. There is also shown ananalysis module808 for analyzing the responses to the test promotions as discussed earlier. There is also shown a general populationpromotion generation module810 for generating the general population promotion using the analysis result of the data from the test promotions. There is also shown amodule812, representing the software/hardware module for receiving the responses.Module812 may represent, for example, the point of sale terminal in a store, a shopping basket on an online shopping website, an app on a smart phone, a webpage displayed on a computer, a social media news feed, etc. where user responses can be received.
• One or more of modules802-812 may be implemented on one or more servers, as mentioned. Adatabase814 is shown, representing the data store for user data and/or test promotion and/or general public promotion data and/or response data.Database814 may be implemented by a single database or by multiple databases. The servers and database(s) may be coupled together using a local area network, an intranet, the interne, or any combination thereof (shown by reference number830).
• User interaction for test promotion administration and/or acquiring user responses may take place via one or more of user interaction devices. Examples of such user interaction devices are wiredlaptop840,wired computer844,wireless laptop846, wireless smart phone ortablet848. Test promotions may also be administered via printing/mailing module850, which communicates the test promotions to the users viamailings852 or printed circular854. The example components ofFIG. 8 are only illustrative and are not meant to be limiting of the scope of the invention. The general public promotion, once generated, may also be communicated to the public using some or all of the user interaction devices/methods discussed herein.
• As can be appreciated by those skilled in the art, providing a result-effective set of recommendations for a generalized public promotion is one of the more important tasks in test promotion optimization.
• In one or more embodiments, there are provided adaptive experimentation and optimization processes for automated promotion testing. Testing is said to be automated when the test promotions are generated in the manner that is likely produce the desired response consistent with the goal of the generalized public promotion.
• For example, if the goal is to maximize profit for the sale of a certain newly created brand of potato chips, embodiments of the invention optimally and adaptively, without using required human intervention, plan the test promotions, iterate through the test promotions to test the test promotion variables in the most optimal way, learn and validate such that the most result-effective set of test promotions can be derived, and provide such result-effective set of test promotions as recommendations for generalized public promotion to achieve the goal of maximizing profit for the sale of the newly created brand of potato chips.
• The term “without required human intervention” does not denote zero human intervention. The term however denotes that the adaptive experimentation and optimization processes for automated promotion testing can be executed without human intervention if desired. However, embodiments of the invention do not exclude the optional participation of humans, especially experts, in various phases of the adaptive experimentation and optimization processes for automated promotion testing if such participation is desired at various points to inject human intelligence or experience or timing or judgment in the adaptive experimentation and optimization processes for automated promotion testing process. Further, the term does not exclude the optional nonessential ancillary human activities that can otherwise also be automated (such as issuing the “run” command to begin generating test promotions or issuing the “send” command to send recommendations obtained).
II. Offer Generation
• As noted previously, offers may be presented by the retailer effectively as they have control over the retail space, and being furthest along in the supply chain, typically have more margin available to make meaningful discounts. Likewise, they can collect transaction logs for downstream analysis of the offer effectiveness. However, while retailers are excellent at selling goods, they typically lack the backend infrastructure and expertise to properly generate, administer and analyze a comprehensive promotional campaign.
• Additionally, manufacturers and distributors themselves have a strong interest in having their specific products promoted (especially over their competitor's products). To this end, manufacturers often engage retailers in promotion agreements. These contracts are usually for a given time period, designate one or more products for discount, and the terms of the offer. Historically, the retailers compile these contracted promotions and roll them out, carte blanche, to all applicable retail outlets. While this may result in increased sales (or other objective), it typically is suboptimal, and could be improved upon by applying improved promotion strategies.
• In order to improve the prior methods of contracted offer delivery and redemption, systems and methods of an offer generation tool based upon contracted offers, and associated backend systems, are provided. To facilitate the discussion,FIG. 9 provides an example block diagram of anoffer generation architecture900. A plurality of retailers910a-xmay interact with a plurality of manufacturers (or distributors)920a-yusing anoffer system960 intermediary. Thisoffer system960 replaces the activities typically performed by a third party consultant, but delivers more refined and accurate results at a much lower cost than traditional methods. Theoffer system960 may access one ormore data stores980, which may be populated with offer structures from an offer bank from theoffer system960, transaction logs from the retailers910a-x,contracts that are active, product details from the manufacturers920a-yand additional third party data (such as trade group analytical data).
• The various components of the offer generation architecture communicate with one another via anetwork950. Thisnetwork950 may include the internet, cellular network(s), private or municipal local and wide area networks, or any combination thereof. In some particular embodiments, persons employed at the retailer and manufacturer may access the offer system on a web based application via the internet and a web browser. Theoffer system960 received inputs from the retailer regarding the contracts for product promotions. Data from the contracts, including contract start and duration, product(s) implicated, and offer structure are gleaned from the contracts.
• The system can then access an offer bank and select offers, using machine learning algorithms, that are designed to best address the objective for the promotional campaign. The objective may be a default of profit or may be configured by the retailer directly. Objectives typically include profitability, gross revenue, product margin, or sales volumes.
• The optimal offers are then scored against the contracted offer details. One example of possible scoring metrics are provided in greater detail in relation to A attached hereto. Essentially, the scoring algorithm looks at various features of the proposed offers, and penalizes then as they migrate too far from the contracted offer. In this manner a small select (typically 2-5) test offers are selected from the larger listing of optimized offers. These test offers, plus the contracted offer, are all deployed at various retailer locations, and transaction logs from the offer duration period are collected for further refinement of the offer bank.
• Theoffer system960 is presented in greater detail atFIG. 10. As noted, the data store(s)980 includes, at a minimum, transaction logs981 collected from one or more retailers, a database of the contracts the retailer is engaged in983,product data985 from the manufacturers, and a repository of prior offers987. Theoffer repository987 is an offer bank of existing offers and analytics regarding offer performance. Analytics may be stored along with the offer for a given population base, and if already deployed previously, the actual response results. This saves significant computational resources, which improves the server's operation by delivering meaningful results with minimal additional computational load.
• The offer system itself960 includes acontract parser963 that uses the contracts database to select contracts for offer testing. The parser may leverage optical character recognition (OCR) techniques, natural language processing (NLP) and trained machine learning models to extract relevant information from the contracts directly. Alternatively, the contract parser may include a human interface that allows an operator to extract the pertinent information from the contract directly. In yet other embodiments, the system may utilize a hybrid approach, whereby confidence of the automated data extraction is measured, and if below a threshold, a human is requested to intervene and process the relevant contract. Information typically extracted from the contract includes the start date, duration of the offer, product(s) implicated by the offer, and offer structure and amounts.
• Theoffer optimizer967 leverages the old transaction logs981,product data985 and objectives, to generate optimal offers (as discussed in the above section in considerable detail). These optimal offers are stored in theoffer database987 as an offer bank for later usage. In some embodiments, the offer optimizations leverage one or more machine learning models to generate the optimal models.
• Theoffer manager965 works in conjunction with the retailer to define an objective of the offer, and ultimately assists in the administration of the offer via theoffer portal961. As the objective is received theoffer optimizer967 may select offers from the offer bank that are best suited to implement along with the contracted offer. Offers are allocated according to their probability of outperforming other offers on the customers objective using reinforced learning techniques. The number of each offer to run in the market is chosen using reinforced learning techniques, such as Thompson sampling, and then offers are then allocated to maximize orthogonality between a set of variables. These variables may include store sales, store stock-out rates, number of relevant SKUs carried in each store, temporal effects and offer features. Offer features can include discount depth, buy quantity and offer structure. The machine learning model performs an optimization problem in which model predictions and uncertainty are computed for each offer for each store for each period and offers are allocated to maximize customer objectives while reducing model uncertainty. The statistical/machine learning model shares information collected about each offer, so that offers are not considered independently. Offers may likewise be scored against the contracted offer to select the set of final test offers that are to be deployed. The scoring process looks at various components of the proposed offers against the contracted offer. As the proposed offers deviate from the contracted offer, they may be penalized.
• Using the offer scores, and anticipated forecast of completing the objective, a select few (configurable number generally between 2-5) offers are selected for administration (along with the contracted offer). Selection may be performed in a number of ways. For example, the top N offers, by forecast, may be first selected, and then ranked by score. From these the top ranked offers, the final selection of offers may be made. In contrast, in some other embodiments, offers with a score above a set threshold are first selected, and these may be ranked by forecast. From this forecast ranking the final offer selection may be made. In yet another embodiment, a weighted composite metric for each offer may be generated. The composite score may be a set weight times the score, plus a normalized value of the forecast times a second configured weight. The offers with the top composite metric are then selected for deployment.
• Forecasts are generated as a base forecast for the retail pricing of the product, plus an expected lift for the offer promotion. Base forecasts are collected from historical transaction log (t-log) data. Lift is a measure of elasticity collected from cross store experimentation times the changed price of the promotion. Baseline t-log data is generally collected from the prior two years, with seasonality, stock, store and growth trends adjusted for in an accounting model. Elasticities from experimentation are calculated and refined over time. The elasticities are utilized in conjunction with either current price or optimal price (depending on the type of forecast desired) to give an additive or subtractive effect to the overall forecast.
• The calculation of elasticities is generated by collecting daily feeds of transactions from retailers. These transaction logs are analyzed and adjusted by seasonality, out of stock events, and store specific adjustments. The data is further determined to be relevant based upon indications if the stores were compliant with a given offer. It is common for a store to adjust an offer unilaterally, or even discarding the offer entirely, in some cases. This generally occurs when a store manager believes she “knows better” than the system, there are insufficient resources to update signage and inventory pricing software, or merely due to miscommunications between the offer generation system and the retailer implementing the promotion. Data is further adjusted to account for out-of-stock events, and normalized across stores to account for different store attributes that may alter the normal velocity of sales for the given product. Lastly, lifts and standard deviations are modeled, and these are used to determine the top performing offers, and confidence levels that they are “better” than the other offers.
• Turning toFIG. 11, a flow chart of an example method for contract based offer generation is provided, as shown at1100. Initially, one or mode contracts are received (at1110) for offers of given products. The contracts are parsed (at1120) to determine the important elements for the contract. Contract parsing is provided in greater detail in relation toFIG. 12. Here the contracts are cross referenced for similarities in products that are subject to offers (at1210). If there are multiple contracts that apply to a single product (at1220) the contract with the longer duration is selected for testing (at1230). This is because longer test periods tend to minimize isolated events that may skew testing results. Once the contract that is applicable is selected, the important information in the contract is extracted. This includes identifying product(s) implicated by the contract (at1240), the offer type/structure and values of the offer (at1250), as well as the dates of the offer (at1260) including the promotional start date and duration. Additional information that may be extracted may include Contract ID, Promotion Type, SKU, Promo Quantity, Promo Price, Number of Stores, Description, Customer Name, Manufacturer ID, Manufacturer Name, Merchant ID, Merchant Name, Deal ID, Deal Type, Deal Calculation Type, Deal Amount, Base Price, Base Cost. This information from the contract offer may be extracted by the system automatically, via human intervention, or by a hybrid approach, as previously discussed in greater detail. The final contract details are output for later usage (at1270).
• Returning toFIG. 11, the next step is to generate test offers (at1130) that are to be tested alongside the contracted offers.FIG. 13 provides greater detail of this offer generation process. Again, initially it is assumed that a number of test promotions have already, and are continually, being run in the retailers. As discussed above, these promotion activities are recorded as transaction logs from the retailers, and are collected by the offer generation system (at1310). The transaction logs are then conditioned, as previously discussed, to take into account store/product variations, out of stock events, and compliance adjustments. This test data is leveraged to generated forecasts for different test offers (not illustrated) as discussed in considerable detail previously, preferably leveraging reinforced learning models. These forecasts for the various offers are used to populate an offer bank for later review and selection of particular offers for testing.
• Subsequently, for the present test, a number of the test offers to be deployed along with the contracted offer is determined (at1330). As discussed, this number typically varies between 2-5 additional test promotions (not including the contract offer). However, the number may dynamically vary based upon the number of retail locations involved in the promotion test, sales volume for the product involved, contract length, or by user configuration. Likewise, the objective of the offers can be set (at1340). Offer objective may include hitting a target (or maximizing) margin, profit maximization, sales volume, revenue maximization, or the like. The possible offers are generated (or selected from the prepopulated offer bank) using reinforced learning techniques (at1350) as previously discussed. As noted, offer selection is an optimization problem for the given objective. The offers are then scored against the contracted offer (at1360). Examples of the algorithms leveraged for offer scoring are provided in Appendix A, as previously noted. This Appendix is entirely for illustrative purposes, and any similar scoring algorithms may be employed. Generally, the further away from the contract offer, or from user's expectations, the larger a penalty is applied to the given test offer. The “best” test offers are then selected (at1370) based upon their score, forecasts or by some hybrid approach, as previously covered in some detail. In some embodiments, Thompson sampling may be leveraged to select the test offers. In some select embodiments, scoring of the test offers isn't performed at all, and the objective optimization alone is leveraged to decide which offers to utilize in the promotion test.
• Returning toFIG. 11, after the selection of the contract based test offers, they may be assigned to retail locations for administration (at1140). The offer administration allocates offers among retailer locations and/or in time periods designed to maximize orthogonality between the following variables: Store sales, Store stock-out rates, Number of relevant SKUs carried in each store, Temporal effects, Discount depth, Buy quantity, and Offer structure. The offer allocation may be done by using reinforcement learning techniques to their fullest, e.g., by an optimization problem in which model predictions and uncertainty are computed for each offer for each store for each period, and offers are allocated to maximize customer objectives while reducing model uncertainty.
• After the offers are thus administered, users in the store redeem the offers (at1150) which becomes reflective in the transaction logs of the retailers. This transaction log data is then leveraged to perform further analysis and refine the machine learning models (at1160). As noted before, the transaction logs are processed and adjusted to account for temporal effects, store variations, out of stock events, and compliance issues. Outliers and stock-outs are detected at the same time and rely on iterating through records for every product-store combination two times. ForPass 1, the steps include: set starting inactivity threshold to 20 days (configurable), mark an item as being out of stock if the inactivity threshold is exceeded, record the average quantity per day when the item was in stock, use the Poisson distribution to get upper and lower bounds for outliers using the average quantity (in some embodiments the system may use 99.9% (configurable) for the threshold. This threshold is desired to be high to make sure it is detecting true outliers and not responses to price changes.), and pick a better day threshold for detecting stock-outs: 1+(int) (−Math.log(0.05)/(average quantity)). The stock out threshold is also configurable. For the Pass 2: mark an item as being out of stock if the new inactivity threshold is exceeded (note, it is not assumed that an item remains out of stock until the next item is sold. Instead, the system designates that it came back into stock for the number of days equal to the threshold before the next item is sold), and mark an item as being an outlier if the quantity is outside of the bounds that was calculated in the previous pass.
• After processing the transaction log data in this manner, lifts and standard deviations are calculated, and these form the basis of the forecasts and confidence intervals of the models.
• Moving forward,FIGS. 14-19 provide illustrations of example dashboards for the disclosed systems. It should be noted that these figures are exemplary only, and are not intended to limit the scope of the given systems. InFIG. 14 categories of products are illustrated at1400. The categories are for promotions that are in development or are being administered. The business objective for the given promotions are provided, including the number of promotion events under the given product category. Attributes being tested are listed, as is the data scientist or managing individual associated with the promotions.
• When a product category is selected, the dashboard goes to a listing of the given promotion events that are associated with the given product category, as is seen in relation toFIG. 15 at1500. Here is individual promotions are listed, along with their statuses, the objective of the given promotion, the contract ID the promotion is associated with, start and end dates, and when applicable a listing of the top offer associated with the promotion event. Attribute tags associated with the promotion are also listed, along with the base price range (or singe number) associated with the product.
• When a specific promotion is selected, the system progresses to a dashboard for the given offer, as seen atFIG. 16 at1600. Here the given offers that have been selected for the promotion (including the control/contract offer) are listed. Each offer includes a rational for it being included, and the percent discount associated with the offer. As noted before, scoring of the offers tends to place a penalty for offers that deviate from the contracted offer. As such, these discount percentages are often the same as the control contract offer, or very similar, as seen here.
• In the promotion screen, the user may also wish to view the execution of the promotion, as seen in relation toFIG. 17 at1700. Here the stores are listed, along with their physical locations, if they are test or control stores, and the offer ID that is deployed in the given store. The percentage of stores that the offers are each deployed at are also seen in the bar graph. As noted, deployment is selected to orthogonally account for a number of variables, as previously touched upon.
• Any given promotion event has the ability to be edited (at least before its deployment). An illustration for this offer editing dashboard is provided atFIG. 18 at1800. Here the control/contracted offer is seen, which cannot be edited. However, additional offers are presented based upon how suited they are for deployment. Offers that scored the highest are listed at the top, and less attractive offers are presented in descending order. For each offer, the type, the rationale for the offer, and discount percentage are all presented. A user is able to manually override the automated offer selection in this editing dashboard, and select different, more or fewer offers for testing.
• InFIG. 19, the offer bank dashboard for all offers is presented, at1900. The category, product, discount, offer type, driver for the offer, and projected sales, unit volumes and margin are all presented for the given offers. As noted, forecasts for each offer are based upon transaction logs previously collected that have been adjusted and processed for lifts and standard deviations based upon the variables associated with the offers using reinforced learning techniques.
IV. System Embodiments
• Now that the systems and methods for the generation and administration of contract based offer testing have been provided, attention shall now be focused upon apparatuses capable of executing the above functions in real-time. To facilitate this discussion,FIGS. 20A and 20B illustrate aComputer System2000, which is suitable for implementing embodiments of the present invention.FIG. 20A shows one possible physical form of theComputer System2000. Of course, theComputer System2000 may have many physical forms ranging from a printed circuit board, an integrated circuit, and a small handheld device up to a huge super computer.Computer system2000 may include aMonitor2002, aDisplay2004, aHousing2006, a server blades including one ormore storage Drives2008, aKeyboard2010, and aMouse2012.Medium2014 is a computer-readable medium used to transfer data to and fromComputer System2000.
• FIG. 20B is an example of a block diagram forComputer System2000. Attached toSystem Bus2020 are a wide variety of subsystems. Processor(s)2022 (also referred to as central processing units, or CPUs) are coupled to storage devices, includingMemory2024.Memory2024 includes random access memory (RAM) and read-only memory (ROM). As is well known in the art, ROM acts to transfer data and instructions uni-directionally to the CPU and RAM is used typically to transfer data and instructions in a bi-directional manner. Both of these types of memories may include any suitable of the computer-readable media described below. AFixed Medium2026 may also be coupled bi-directionally to theProcessor2022; it provides additional data storage capacity and may also include any of the computer-readable media described below. FixedMedium2026 may be used to store programs, data, and the like and is typically a secondary storage medium (such as a hard disk) that is slower than primary storage. It will be appreciated that the information retained within FixedMedium2026 may, in appropriate cases, be incorporated in standard fashion as virtual memory inMemory2024.Removable Medium2014 may take the form of any of the computer-readable media described below.
• Processor2022 is also coupled to a variety of input/output devices, such asDisplay2004,Keyboard2010,Mouse2012 andSpeakers2030. In general, an input/output device may be any of: video displays, track balls, mice, keyboards, microphones, touch-sensitive displays, transducer card readers, magnetic or paper tape readers, tablets, styluses, voice or handwriting recognizers, biometrics readers, motion sensors, brain wave readers, or other computers.Processor2022 optionally may be coupled to another computer or telecommunications network usingNetwork Interface2040. With such aNetwork Interface2040, it is contemplated that theProcessor2022 might receive information from the network, or might output information to the network in the course of performing the above-described promotion offer generation and redemption. Furthermore, method embodiments of the present invention may execute solely uponProcessor2022 or may execute over a network such as the Internet in conjunction with a remote CPU that shares a portion of the processing.
• Software is typically stored in the non-volatile memory and/or the drive unit. Indeed, for large programs, it may not even be possible to store the entire program in the memory. Nevertheless, it should be understood that for software to run, if necessary, it is moved to a computer readable location appropriate for processing, and for illustrative purposes, that location is referred to as the memory in this disclosure. Even when software is moved to the memory for execution, the processor will typically make use of hardware registers to store values associated with the software, and local cache that, ideally, serves to speed up execution. As used herein, a software program is assumed to be stored at any known or convenient location (from non-volatile storage to hardware registers) when the software program is referred to as “implemented in a computer-readable medium.” A processor is considered to be “configured to execute a program” when at least one value associated with the program is stored in a register readable by the processor.
• In operation, thecomputer system2000 can be controlled by operating system software that includes a file management system, such as a medium operating system. One example of operating system software with associated file management system software is the family of operating systems known as Windows® from Microsoft Corporation of Redmond, Washington, and their associated file management systems. Another example of operating system software with its associated file management system software is the Linux operating system and its associated file management system. The file management system is typically stored in the non-volatile memory and/or drive unit and causes the processor to execute the various acts required by the operating system to input and output data and to store data in the memory, including storing files on the non-volatile memory and/or drive unit.
• Some portions of the detailed description may be presented in terms of algorithms and symbolic representations of operations on data bits within a computer memory. These algorithmic descriptions and representations are the means used by those skilled in the data processing arts to most effectively convey the substance of their work to others skilled in the art. An algorithm is, here and generally, conceived to be a self-consistent sequence of operations leading to a desired result. The operations are those requiring physical manipulations of physical quantities. Usually, though not necessarily, these quantities take the form of electrical or magnetic signals capable of being stored, transferred, combined, compared, and otherwise manipulated. It has proven convenient at times, principally for reasons of common usage, to refer to these signals as bits, values, elements, symbols, characters, terms, numbers, or the like.
• The algorithms and displays presented herein are not inherently related to any particular computer or other apparatus. Various general purpose systems may be used with programs in accordance with the teachings herein, or it may prove convenient to construct more specialized apparatus to perform the methods of some embodiments. The required structure for a variety of these systems will appear from the description below. In addition, the techniques are not described with reference to any particular programming language, and various embodiments may, thus, be implemented using a variety of programming languages.
• In alternative embodiments, the machine operates as a standalone device or may be connected (e.g., networked) to other machines. In a networked deployment, the machine may operate in the capacity of a server or a client machine in a client-server network environment or as a peer machine in a peer-to-peer (or distributed) network environment.
• The machine may be a server computer, a client computer, a personal computer (PC), a tablet PC, a laptop computer, a set-top box (STB), a personal digital assistant (PDA), a cellular telephone, an iPhone, a Blackberry, a processor, a telephone, a web appliance, a network router, switch or bridge, or any machine capable of executing a set of instructions (sequential or otherwise) that specify actions to be taken by that machine.
• While the machine-readable medium or machine-readable storage medium is shown in an exemplary embodiment to be a single medium, the term “machine-readable medium” and “machine-readable storage medium” should be taken to include a single medium or multiple media (e.g., a centralized or distributed database, and/or associated caches and servers) that store the one or more sets of instructions. The term “machine-readable medium” and “machine-readable storage medium” shall also be taken to include any medium that is capable of storing, encoding or carrying a set of instructions for execution by the machine and that cause the machine to perform any one or more of the methodologies of the presently disclosed technique and innovation.
• In general, the routines executed to implement the embodiments of the disclosure may be implemented as part of an operating system or a specific application, component, program, object, module or sequence of instructions referred to as “computer programs.” The computer programs typically comprise one or more instructions set at various times in various memory and storage devices in a computer, and when read and executed by one or more processing units or processors in a computer, cause the computer to perform operations to execute elements involving the various aspects of the disclosure.
• Moreover, while embodiments have been described in the context of fully functioning computers and computer systems, those skilled in the art will appreciate that the various embodiments are capable of being distributed as a program product in a variety of forms, and that the disclosure applies equally regardless of the particular type of machine or computer-readable media used to actually effect the distribution
• While this invention has been described in terms of several embodiments, there are alterations, modifications, permutations, and substitute equivalents, which fall within the scope of this invention. Although sub-section titles have been provided to aid in the description of the invention, these titles are merely illustrative and are not intended to limit the scope of the present invention. It should also be noted that there are many alternative ways of implementing the methods and apparatuses of the present invention. It is therefore intended that the following appended claims be interpreted as including all such alterations, modifications, permutations, and substitute equivalents as fall within the true spirit and scope of the present invention.

Claims (20)

What is claimed is:

1. A computer implemented method for generating contract based offers comprising:

receiving a contract for a promotional offer on a product;

extracting data from the contract;

accessing an offer bank; and

selecting a plurality of test offers from the offer bank by scoring each offer in the offer bank against the extracted data.

2. The method ofclaim 1, further comprising deploying the promotional offer and the selected plurality of test offers in a plurality of retail locations.

3. The method ofclaim 2, wherein the deploying is performed to maximize orthogonality between the following variables: store sales, store out of stock rates, number of relevant SKUs carried in each store, temporal effects, discount depth, buy quantity and offer structure.

4. The method ofclaim 1, further comprising selecting a number of test offers to run in-market using reinforcement learning techniques.

5. The method ofclaim 4, wherein Thompson sampling is used to select the number.

6. The method ofclaim 1, wherein the offer bank is populated with forecasted offers.

7. The method ofclaim 6, wherein the forecasts are based upon transaction logs of a plurality of retailers.

8. The method ofclaim 7, wherein the transaction logs are adjusted for compliance by the given retailer, estimated out of stock events, normalized across stores to account for different store attributes, and adjusted for temporal effects.

9. The method ofclaim 8, further comprising applying machine learning to the adjusted transaction logs to determine lift and standard deviation for a given test offer.

10. The method ofclaim 9, wherein the forecasts are a baseline function of time from the transaction log data plus elasticity from cross store experiments times a change in price, where in the elasticity is calculated as a function of the lift, and a confidence for the forecast is calculated as a function of the standard deviation.

11. A computer product comprising non-transitory computer readable medium, which when executed on a computer system causes the computer system to perform the steps of:

receiving a contract for a promotional offer on a product;

extracting data from the contract;

accessing an offer bank; and

selecting a plurality of test offers from the offer bank by scoring each offer in the offer bank against the extracted data.

12. The computer product ofclaim 11, further comprising deploying the promotional offer and the selected plurality of test offers in a plurality of retail locations.

13. The computer product ofclaim 12, wherein the deploying is performed to maximize orthogonality between the following variables: store sales, store out of stock rates, number of relevant SKUs carried in each store, temporal effects, discount depth, buy quantity and offer structure.

14. The computer product ofclaim 11, wherein when the computer readable product when executed further performs the step of selecting a number of test offers to run in-market using reinforcement learning techniques.

15. The computer product ofclaim 14, wherein Thompson sampling is used to select the number.

16. The computer product ofclaim 11, wherein the offer bank is populated with forecasted offers.

17. The computer product ofclaim 16, wherein the forecasts are based upon transaction logs of a plurality of retailers.

18. The computer product ofclaim 17, wherein the transaction logs are adjusted for compliance by the given retailer, estimated out of stock events, normalized across stores to account for different store attributes, and adjusted for temporal effects.

19. The computer product ofclaim 18, wherein when the computer readable product when executed further performs the step of applying machine learning to the adjusted transaction logs to determine lift and standard deviation for a given test offer.

20. The computer product ofclaim 19, wherein the forecasts are a baseline function of time from the transaction log data plus elasticity from cross store experiments times a change in price, where in the elasticity is calculated as a function of the lift, and a confidence for the forecast is calculated as a function of the standard deviation.

Images