TECHNICAL FIELD

[0001] The disclosure relates generally to automatically recommending markdowns, such as markdowns in nascent markets. More particularly, the disclosure relates to automatically recommending markdowns using hierarchical reinforcement learning in markets having limited historical data.

BACKGROUND

[0002] At least some retailers offer markdowns for goods or services that they may offer at retail stores. The markdowns may take the form of reduced product prices, discounts or other incentives that can induce more customers to purchase the goods or services or to induce customers to purchase additional or incremental goods or services in addition to existing purchases. Retailers may offer markdowns on products or services in order to improve the financial performance of the retailer.

[0003] The use of markdowns by retailers can be used in the context of grocery sales. Retailers in the grocery industry can be faced with inventories of products that have a limited shelf life and must typically be sold prior to an expiration date (or expiration time) of the grocery products. In this context, it can be financially advantageous to offer markdowns on grocery products in order to sell such products at a lower price prior to an expiration rather than to maintain a desired full retail price of the grocery products until the expiration. Markdowns can also be used in other industries and contexts in order to improve the financial performance of the retailer. Such other industries can be faced with other constraints in which markdowns can be employed to sell products or services more quickly. Such other constraints can include, for example, capacity constraints, storage constraints, transportation constraints, manpower constraints and the like.

[0004] Many different methods and techniques can be used in order maximize the financial performance of a retailer when it chooses to implement markdowns. In some examples, historical data can be used to determine markdowns that can improve financial performance. Historical data, however, can be limited for some geographic regions, for new retail store locations, for new products, or for other reasons. The use of historical data for determining markdowns can produce undesirable results if the quantity of historical data is limited. There is a need, therefore, for improved methods of determining markdowns by retailers in circumstances in which historical data is limited.

SUMMARY

[0005] The embodiments described herein are directed to a markdown system and related methods. The markdown system can be implemented using one or more computing devices that can include operative elements that can determine and send recommended price markdowns to a retailer. The markdown system can be effectively used in nascent markets in which there may be limited historical markdown data that may otherwise be required by existing models and methods of recommending price markdowns. The markdown system can be implemented to determine optimized price markdowns and experimental price markdowns. The markdown system can probabilistically choose between the optimized price markdowns and the experimental price markdowns and then send final recommended price markdowns to be implemented by the retailer.

[0006] In accordance with various embodiments, exemplary systems may be implemented in any suitable hardware or hardware and software, such as in any suitable computing device. For example, in some embodiments, a computing device is configured to obtain historical markdown data characterizing customer purchasing behavior. The computing device can also determine initial recommended price markdown data for a first period based on the historical markdown data wherein the initial recommended price markdown data for the first period comprises optimized price markdowns and experimental price markdowns. The optimized price markdowns and the experimental price markdowns can be different. The computing device can also determine final recommended price markdown data for the first period based on probabilistically choosing between the optimized price markdowns and the experimental price markdowns of the initial price markdown data. The computing device can also send the final recommended price markdown data to at least one retailer and obtain markdown performance data characterizing customer purchasing behavior of implemented price markdowns based on the final recommended price markdown data.

[0007] In one aspect, the historical markdown data can include actual markdown data and customer purchasing data for one or more periods prior to the first period. [0008] In another aspect, the optimized price markdowns can be determined by calculating a maximum savings for a retailer.

[0009] In another aspect, the experimental price markdowns can be chosen from possible markdowns within a predetermined range of the optimized price markdowns.

[0010] In another aspect, the experimental price markdowns can be chosen based on a trend of savings to a retailer in periods prior to the first period.

[0011] In another aspect, the computing device is further configured to determine second initial recommended price markdown data that includes optimized price markdowns and experimental price markdowns for a second period subsequent to the first period, and to determine second final recommended price markdown data for the second period based on probabilistically choosing between the optimized price markdowns and the experimental price markdowns if savings to the retailer in periods prior to the second period is unstable.

[0012] In another aspect, the computing device can probabilistically choose between the optimized price markdowns and the experimental price markdowns by comparing a current savings of the retailer to a historical savings of the retailer.

[0013] In some embodiments of the present disclosure a method of recommending price markdowns in a nascent market is provided. The method can include obtaining historical markdown data characterizing customer purchasing behavior in a nascent market and determining initial recommended price markdown data for a first period based on the historical markdown data wherein the initial recommended price markdown data for the first period comprises optimized price markdowns and experimental price markdowns. The optimized price markdowns and the experimental price markdowns can be different. The method can also include determining final recommended price markdown data for the first period based on probabilistically choosing between the optimized price markdowns and the experimental price markdowns of the initial price markdown data and sending the final recommended price markdown data to at least one retailer in the nascent market. The method can also include obtaining markdown performance data characterizing customer purchasing behavior of implemented price markdowns based on the final recommended price markdown data. [0014] In yet other embodiments, a non-transitory computer readable medium has instructions stored thereon, where the instructions, when executed by at least one processor, cause a computing device to perform operations that include obtaining historical markdown data characterizing customer purchasing behavior in a nascent market and determining initial recommended price markdown data for a first period based on the historical markdown data wherein the initial recommended price markdown data for the first period comprises optimized price markdowns and experimental price markdowns. The optimized price markdowns and the experimental price markdowns can be different. The instructions, when executed by the at least one processor, can cause the computing device to perform operations that can also include determining final recommended price markdown data for the first period based on probabilistically choosing between the optimized price markdowns and the experimental price markdowns of the initial price markdown data and sending the final recommended price markdown data to at least one retailer in the nascent market. The operations can also include obtaining markdown performance data characterizing customer purchasing behavior of implemented price markdowns based on the final recommended price markdown data.

BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The features and advantages of the present disclosures will be more fully disclosed in, or rendered obvious by the following detailed descriptions of example embodiments. The detailed descriptions of the example embodiments are to be considered together with the accompanying drawings wherein like numbers refer to like parts and further wherein:

[0016] FIG. 1 is a block diagram of a markdown system in accordance with some embodiments;

[0017] FIG. 2 is a block diagram of a markdown computing device of the markdown system of FIG. 1 in accordance with some embodiments;

[0018] FIG. 3 is a block diagram illustrating examples of various portions of the markdown system of FIG. 1 in accordance with some embodiments;

[0019] FIG. 4 is a block diagram illustrating examples of various portions of another markdown system in accordance with some embodiments; [0020] FIG. 5 is a block diagram illustrating examples of various portions of the markdown systems in accordance with some embodiments;

[0021] FIG. 6 is diagram illustrating example processing of historical data to determine recommended price markdowns in accordance with some embodiments;

[0022] FIG. 7 is a diagram illustrating another example of processing of historical data to determine recommended price markdowns in accordance with some embodiments;

[0023] FIG. 8 is a diagram illustrating further processing of the example shown in FIG. 7; and

[0024] FIG. 9 is a flowchart of an example method of determining final recommended price markdowns in accordance with some embodiments.

DETAILED DESCRIPTION

[0025] The description of the preferred embodiments is intended to be read in connection with the accompanying drawings, which are to be considered part of the entire written description of these disclosures. While the present disclosure is susceptible to various modifications and alternative forms, specific embodiments are shown by way of example in the drawings and will be described in detail herein. The objectives and advantages of the claimed subject matter will become more apparent from the following detailed description of these exemplary embodiments in connection with the accompanying drawings.

[0026] It should be understood, however, that the present disclosure is not intended to be limited to the particular forms disclosed. Rather, the present disclosure covers all modifications, equivalents, and alternatives that fall within the spirit and scope of these exemplary embodiments. The terms “couple,” “coupled,” “operatively coupled,” “connected,” “operatively connected,” and the like should be broadly understood to refer to connecting devices or components together either mechanically, electrically, wired, wirelessly, or otherwise, such that the connection allows the pertinent devices or components to operate (e.g., communicate) with each other as intended by virtue of that relationship. [0027] Retailers can offer markdowns on goods or services in order to improve the financial performance of the retailer. The markdowns (MD’s) can take the form of discounts, price reductions, bundling promotions or other offers such as buy one, get one free offers. One example industry that uses markdowns to improve the financial performance of its retail stores is the grocery industry. Retailers in the grocery industry often offer markdowns in order to increase sales of grocery products and/or to minimize waste. The markdowns, in turn, can improve the financial performance of the retailer. Other industries can also use markdowns to improve financial services. Example industries can include industries that may sell goods or services that are subject to spoilage or may have a tendency to reduce in value over time, for example due to seasonality effects or other factors. Such other industries can include fashion retail, commodity sellers/brokerage, farming, and the like.

[0028] In the context of grocery retailers, the methods and systems of the present disclosure address the goal of such retailers to improve financial performance by balancing losses due to markdowns of product price and loss due to waste. Markdown loss is the loss in revenue (LoR) due to a decrease in the price of the product when the product is sold. Loss due to waste is the lost revenue to the retailer when the retailer must dispose of the product due to expiration, spoilage or other reason.

[0029] The methods and systems of the present disclosure can provide dynamic price markdowns to the retailers. The methods and systems are dynamic in that the recommended price markdowns can vary depending on the item quantity that is centered around a base price markdown. As can be appreciated, an increased price markdown may be necessary if the retailer has larger quantities of product on hand that are subject to spoilage. Similarly, the retailer may offer a reduced price markdown if the quantity on hand is less than the base price markdown.

[0030] The methods and systems are dynamic in that the recommended price markdowns can be monitored and adjusted on a periodic basis to ensure an optimal return and/or savings to the retailer. The methods and systems can monitor a trend of loss due to waste and a trend of loss due to loss of revenue to make appropriate adjustments on the periodic basis.

[0031] As described above, recommended price markdowns can be provided based on a base markdown and then adjusted according to the inventory of product on hand at a given time. These base markdowns, however, are often calculated using historical price markdown data. This historical price markdown data can be available for established markets in which a retailer has been operating for a prolonged period of time and has been collecting customer purchasing behavior data and markdown data over this prolonged period of time. It can be more difficult, however, to determine stable and accurate base markdowns in nascent markets in which historical price markdown data is not available. In such nascent markets, therefore, improved methods and systems for determining recommended price markdown data is needed. The methods and systems of the present disclosure provide such improved methods for providing recommended price markdowns in nascent markets.

[0032] The methods and systems of the present disclosure can determine both optimized price markdowns and experimental price markdowns. The methods and systems can then probabilistically choose between the optimized price markdowns and the experimental price markdowns to determine a final recommended price markdown to be implemented at a retail store. The customer’s purchasing behavior based on the implemented price markdowns can be collected and then re-introduced into the markdown system to determine newly determined optimized price markdowns and newly determined experimental price markdowns. The methods and systems of the present disclosure can then continue such process of determining final recommended price markdowns until such time as there is sufficient historical price markdown data to stabilize the optimized price markdowns and the experimental price markdowns.

[0033] Turning to the drawings, FIG. 1 illustrates a block diagram of a markdown selection system 100 that includes a markdown computing device 102 (e.g., a server, such as an application server), a central ordering computing device 124, an information source 126 (e.g., a web server), a retailer inventory management system 112, a database 108, and a customer computing device 104 operatively coupled over network 110. Markdown computing device 102, central ordering computing device 124, information source 126, retailer inventory management system 112, and customer computing device 104 can each be any suitable computing device that includes any hardware or hardware and software combination for processing and handling information. For example, each can include one or more processors, one or more field-programmable gate arrays (FPGAs), one or more application-specific integrated circuits (ASICs), one or more state machines, digital circuitry, or any other suitable circuitry. In addition, each can transmit data to, and receive data from, communication network 110.

[0034] In some examples, markdown computing device 102 can be a computer, a workstation, a laptop, a server such as a cloud-based server, or any other suitable device. In some examples the customer computing device 104 can be a cellular phone, a smart phone, a tablet, a personal assistant device, a voice assistant device, a digital assistant, a laptop, a computer, or any other suitable device. In some examples, markdown computing device 102 is operated by a retailer, and the customer computing devices 104 can be operated by a customer of the retailer.

[0035] The retailer inventory management system 112 can include one or more workstations 116 that can be coupled to a server, communication network or router 114. The retailer inventory management system 112 can, for example, be located at a store, warehouse or other retailer location. The store or warehouse can be operated by the retailer and can be a location at which markdown information is received from the central ordering computing device 124. The central ordering computing device 124 can also send customer ordering information, pricing information and other information to the retailer locations. The markdown data can include price markdown information that can be used at the retail location and in its online ordering systems as will be further described below.

[0036] Markdown computing device 102 can also be operable to communicate with database 108 over the communication network 110. The database 108 can be a remote storage device, such as a cloud-based server, a memory device on another application server, a networked computer, or any other suitable remote storage. Although shown remote to markdown computing device 102, in some examples, database 108 can be a local storage device, such as a hard drive, a non-volatile memory, or a USB stick. Markdown computing device 102 can also be operable to communicate with information source 126 in order to acquire, obtain or otherwise collect information that can be used during the process of determining price markdowns. While only one information source 126 is shown, markdown computing device 102 can be operable to communicate with multiple information sources 126 or other computing devices, servers and retailer information systems. [0037] Communication network 110 can be a WiFi® network, a cellular network such as a 3GPP® network, a Bluetooth® network, a satellite network, a wireless local area network (LAN), a network utilizing radio-frequency (RF) communication protocols, a Near Field Communication (NFC) network, a wireless Metropolitan Area Network (MAN) connecting multiple wireless LANs, a wide area network (WAN), or any other suitable network. Communication network 110 can provide access to, for example, the Internet.

[0038] The customer computing device 104 may communicate with the central ordering computing device 124 over communication network 110. For example, the central ordering computing device 124 may host one or more websites. Each of the customer computing devices 104 (although only one is shown in FIG. 1) may be operable to view, access and interact with the websites hosted by the central ordering computing device 124. In some examples, the central ordering computing device 124 can allow a customer via the customer computing devices 104 to browse, search and/or select products for purchase. The central ordering computing device 124 can also allow the customer to select products for purchase and may also permit the customer to add, delete or revise the ordered items.

[0039] FIG. 2 illustrates an example computing device 200. The markdown computing device 102, the retailer inventory management system 112, the central ordering computing device 124, the information source 126, and/or the customer computing device 104 may include the features shown in FIG. 2. For the sake of brevity, FIG. 2 is described relative to the markdown computing device 102. It should be appreciated, however, that the elements described can be included, as applicable, in the retailer inventory management system 112, the central ordering computing device 124, the information source 126, and/or the customer computing device 104.

[0040] As shown, the markdown computing device 102 can be a computing device 200 that may include one or more processors 202, working memory 204, one or more input/output devices 206, instruction memory 208, a transceiver 212, one or more communication ports 214, and a display 216, all operatively coupled to one or more data buses 210. Data buses 210 allow for communication among the various devices. Data buses 210 can include wired, or wireless, communication channels. [0041] Processors 202 can include one or more distinct processors, each having one or more cores. Each of the distinct processors can have the same or different structure. Processors 202 can include one or more central processing units (CPUs), one or more graphics processing units (GPUs), application specific integrated circuits (ASICs), digital signal processors (DSPs), and the like.

[0042] Processors 202 can be configured to perform a certain function or operation by executing code, stored on instruction memory 208, embodying the function or operation. For example, processors 202 can be configured to perform one or more of any function, method, or operation disclosed herein.

[0043] Instruction memory 208 can store instructions that can be accessed (e.g., read) and executed by processors 202. For example, instruction memory 208 can be a non-transitory, computer-readable storage medium such as a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), flash memory, a removable disk, CD-ROM, any non-volatile memory, or any other suitable memory.

[0044] Processors 202 can store data to, and read data from, working memory 204. For example, processors 202 can store a working set of instructions to working memory 204, such as instructions loaded from instruction memory 208. Processors 202 can also use working memory 204 to store dynamic data created during the operation of the markdown computing device 102. Working memory 204 can be a random access memory (RAM) such as a static random access memory (SRAM) or dynamic random access memory (DRAM), or any other suitable memory.

[0045] Input-output devices 206 can include any suitable device that allows for data input or output. For example, input-output devices 206 can include one or more of a keyboard, a touchpad, a mouse, a stylus, a touchscreen, a physical button, a speaker, a microphone, or any other suitable input or output device.

[0046] Communication port(s) 214 can include, for example, a serial port such as a universal asynchronous receiver/transmitter (UART) connection, a Universal Serial Bus (USB) connection, or any other suitable communication port or connection. In some examples, communication port(s) 214 allows for the programming of executable instructions in instruction memory 208. In some examples, communication port(s) 214 allow for the transfer (e.g., uploading or downloading) of data, such as historical markdown data, optimized price markdowns, experimental price markdowns, final recommended price markdown data, customer purchasing data, historical savings data and other types of data described herein.

[0047] Display 216 can display a user interface 218. User interfaces 218 can enable user interaction with the markdown computing device 102. For example, user interface 218 can be a user interface that allows an operator to interact, communicate, control and/or modify different features or parameters of the markdown computing device 102. The user interface 218 can, for example, display the recommended price markdowns or the performance of the markdown computing device 102 using different textual, graphical or other types of graphs, tables or the like. In some examples, a user can interact with user interface 218 by engaging input-output devices 206. In some examples, display 216 can be a touchscreen, where user interface 218 is displayed on the touchscreen.

[0048] Transceiver 212 allows for communication with a network, such as the communication network 110 of FIG. 1. For example, if communication network 110 of FIG. 1 is a cellular network, transceiver 212 is configured to allow communications with the cellular network. In some examples, transceiver 212 is selected based on the type of communication network 110 markdown computing device 102 will be operating in. Processor(s) 202 is operable to receive data from, or send data to, a network, such as communication network 110 of FIG. 1, via transceiver 212.

[0049] Turning now to FIG. 3, further aspects of the markdown system 100 are shown. In this example of the markdown system 100, the elements described below are shown in communication with each other. It should be appreciated, however, that while not shown, the elements in example 100 can be in communication with each other over a communication network (e.g., communication network 110). The markdown computing device 102 can be in communication with the central ordering computing device 124, the information source 126 and the database 108. A customer can access content in the retailer’s ecommerce marketplace (i.e., a website) hosted on the central ordering computing device 124 by using his or her mobile computing device 104. In other examples, the customer may use a desktop customer computing device or any other suitable computing device. The customer can browse, search, select or otherwise interact with the ecommerce marketplace.

[0050] The markdown computing device 102 can be operable to recommend price markdowns to a retailer. The markdown computing device 102 can, for example, communicate with a retailer by sending recommended price markdown data to the retailer’s inventory management system 112. The recommended price markdown data can be implemented by the retailer in the retailer’s online ordering system and/or at the physical store. Customer purchasing behavior can be collected at the retailer’s stores and communicated back to the retailer via the inventory management system 112 and the central ordering computing device 124. The customer purchasing behavior can, in turn, be stored in the information source 126 or in the database 108. As shown, the database 108 can store the customer purchasing behavior as historical markdown data 310.

[0051] The historical markdown data 310 can include various types of information related to a customer’s purchasing behavior and particularly, customers’ purchasing behavior in view of markdowns that have been made to various products. For example, the historical markdown data 310 can include information regarding the customer (location, demographics, purchase history, etc.), the customer’s purchasing habits in light of markdowns and other environmental or time data (type of product, quantity, date, time of day, etc.). The historical markdown data 310 can then be used by the markdown computing device 102 to recommend markdowns for products at the retailer. In nascent markets, the amount of historical markdown data 310 can be limited such that existing models and methods for determining recommended markdowns do not produce stable, efficient or optimized results. The methods and systems of the present disclosure are particularly suited for such nascent markets that have limited historical markdown data. For purposes of the present disclosure, nascent markets are markets that have limited historical markdown data. Such nascent markets can be characterized using various metrics. In one example, a nascent market is a market in which historical markdown data 310 is limited to sales that have occurred in four or less weeks of sales. In other examples, a nascent market is a market that has six or less months of historical markdown data 310. In still other examples, a nascent market is a market that has three or less months of historical markdown data 310. In yet other examples, nascent markets can be characterized as markets that have historical markdown data 310 that produces unstable results. [0052] The markdown computing device 102 can include various elements that can perform various operations in order to provide markdown recommendations to the retailer. In the example shown, the markdown computing device 102 can include a data extraction engine 302, a markdown optimizer 304 and an automated feedback engine 306. The operations and further details of these components will be further described below. The data extraction engine 302 can obtain data from other components or databases (such as database 108, central ordering computing device 124, retailer inventory management system 112, and/or information source 126) that can be used to recommend markdowns for various products that may be offered for sale by the retailer. The data extraction engine 302 can, for example, use various application protocol interfaces (APIs) to obtain information from such other information sources, databases or memory. In other examples, the data extraction engine 302 can use other methods or processes to obtain information such as accessing look-up tables, scraping data from information sources or the like.

[0053] The markdown optimizer 304 can be implemented in various ways and use various methods or processes to use the data that is obtained by the data extraction engine 302. The markdown optimizer 304 can, for example, include a hierarchical reinforcement learning (HRL) engine that can be implemented using Spark piping. In other examples, other types of deep reinforcement learning (DRL) can be used. In still other examples, the markdown optimizer 304 can be implemented using other tools, algorithms or methodologies.

[0054] The automated feedback engine 306 can be implemented in various ways and use various methods or processes. The automated feedback engine 306 can compare or evaluate the markdowns that are determined by the markdown optimizer 304. The automated feedback engine can evaluate the markdowns using the historical markdown data 310. The automated feeback engine 306 can determine and observe savings that are achieved when markdowns are deployed at the retail stores, for example. The automated feedback engine 306 can observe trends over time and revise and/or adjust the recommended markdowns that are determined by the markdown optimizer 304. In other examples, the automated feedback engine 306 can use other methods or other metrics to adjust the performance of the markdown optimizer 304.

[0055] Turning now to FIG. 4, another example markdown system 400 is shown. In this example, the markdown computing device 102 is shown to include the data extraction engine 302, the markdown optimizer 304 and the automated feedback engine 306. In this example, the data extraction engine 302 can be coupled to store system data 402. The store system data 402 can include the historical markdown data 310 as previously described. The markdown optimizer 304 can be coupled to the data extraction engine 302 to allow the markdown optimizer to access the data that is obtained by the data extraction engine 302. The markdown optimizer 304, in this example is a HRL engine using Spark Piping. In other examples, the markdown optimizer 304 can utilize other tools or other methodology. The markdown optimizer 304 can determine and exploit historical markdowns that resulted in the maximum savings to the retailer. The markdown optimizer 304 can also explore and/or determine new experimental markdowns within a predetermined range of the historical markdowns. The automated feedback module 306 can observe the savings that result from implemented markdowns over time and then provide feedback to indicate whether to continue the process and/or determine whether the recommended markdowns have stabilized.

[0056] The markdown computing device 102 can then push or send the model parameters that are determined by the markdown computing device 102 to the a database or MS SQL server 404. The model parameters or final recommended price markdown data 406 can be sent or otherwise communicated to the retail store. This final recommended price markdown data 406 can be implemented at the retail store. The retail store can collect further information in its local systems and store such data with store system data 402. This data can show the customer’s purchasing behaviors after the final recommended price markdown data 406 was implemented at the retail store. Thus, the retailer can collect the performance and/or change in savings that may result from implementation of the price markdown data. This information is fed back into the markdown computing device by the data extraction engine 302 and/or the automated feedback engine 306. In this manner, the performance of the markdown computing device 102 can be stabilized and improved over time.

[0057] Referring now to FIG. 5, another illustration of the methods and systems of the present disclosure is shown. As shown, the methods and systems of the present disclosure may include the batch processing 500 that may be performed by one or more elements of the markdown system 100. The markdown computing device 102 (FIGs. 3, 4), or various elements thereof, can be implemented, for example, as a Hadoop cluster 502. The Hadoop cluster 502 can include, for example, a Hadoop Distributed File System (HDFS) and one or more programming models for processing of the data extracted and obtained by the system. The Hadoop cluster 502 can process the data as described using hierarchical reinforcement learning to determine recommended price markdown data. The price markdown data can be organized or structured in a structured query language (SQL) database or in the MS-SQL Server 506. The Spark Tables 504 can organize and/or structure the price markdown data before the price markdown data is transferred to the MS-SQL Server 506.

[0058] FIG. 5 also illustrates elements of the real-time store execution process(es) 550 that can be performed to implement marked-down store prices 514 at a retail store. The workstations, or store scanners 116 can be used at the retail store to maintain a real-time inventory of the products that are available at a retail store. Any suitable calculation service 510 can be used to collect the information that is collected using a store scanner 116. For example, the calculation service 510 can be the retailer inventory management system 112 previously described. The markdown computing device 512 can obtain data from the calculation service and from the database or MS- SQL Server 506 to determine final recommended price markdown data. This final recommended price markdown data can then be implemented at the retail store as marked-down store-prices 514. The real-time store execution process 550 can be performed in real-time so that recommended price markdowns can be provided over time and can even be provided at multiple times during a given day as circumstances and inventory of available products may change over time and during the day.

[0059] FIGs. 6 and 7 illustrate example calculations that can be performed by the markdown system 100. In the example shown in FIG. 6, a sample calculation for Period A and a subsequent Period B is shown. In the example, the period is shown as one week but the example method and related calculations can be performed at any suitable intervals, including, for example, hourly, daily, bi-weekly, monthly or the like. The illustrated process 600 can begin with the available historical data (or historical markdown data) 602. This historical data 602 can comprise actual markdown data and customer purchasing data. The actual markdown data can include information about the actual price markdowns that were implemented at one or more retail stores and the customer purchasing data can include information about the number of customer purchases that were made of the products with the actual marked-down prices. [0060] As further shown, the markdown computing device 102 can determine a range of price markdowns 604 and the related savings 606 that were achieved at each respective price markdown 604. The savings 606 can be determined using any suitable savings metric or a combination of savings metrics including, rate of return, actual savings as computed from a prior period, or the like. The markdown computing device 102 can then determine the price markdown 604 at which the savings 606 is the greatest. In the example shown, the price markdown 604 of 40% results in the greatest savings of $100. For this reason, the markdown computing device can select the 40% price markdown as an optimized price markdown 620.

[0061] In other systems that may determine or select price markdowns, the system may only choose the price markdown 604 that optimizes the savings 606. The methods and systems of the present disclosure recognize that in nascent markets the limited availability of historical data 602 can provide only a limited picture of the possibility of greater savings over the optimized price markdown 620 because the retail store may not have implemented other price markdowns in the past and therefore has limited experience with other price markdowns. For this reason, the markdown computing device 102 can also select an experimental price markdown 622. The experimental price markdown 622 can be selected using any suitable methodology. In one example, the experimental price markdown 622 can be selected from a predetermined range of possible price markdowns. The predetermined range of possible price markdowns can have any suitable size. In one example, the predetermined range of possible price markdowns is determined based on the price markdown that was implemented in a prior period.

[0062] It can be desirable to limit the magnitude that price markdowns vary from one period to the next. It can be desirable to limit the change of price markdowns so as to manage the customer’s experience with price markdowns. For example, if customers observe or experience price markdowns that vary greatly from period to period, it can cause customers to be wary of normal retail prices and/or to cause customers to question a retailer’s normal retail prices. In addition, large variations in price markdowns can cause customers to prolong or wait to purchase products with an expectation of a future price markdown. For these reasons and others, the variation of price markdowns from period to period can be limited. With this in mind, the predetermined range of possible price markdowns can be limited so that it can vary only by 3% to 5% over the price markdown of a previous period. In other examples, the predetermined range of possible price markdowns can be limited so that it varies by less than 5% over the previous period. In still other examples, the predetermined range of possible price markdowns can be allowed to vary by other percentages of a previous period price markdown such as by less than 10%, 15% or 20%.

[0063] Referring back to FIG. 6, the experimental price markdown 622 that is selected in this example is 30%. As shown, the markdown computing device 102, in this example, has selected and/or determined both the optimized price markdown 620 and the experimental price markdown 622. The markdown computing device 102 can then choose between the optimized price markdown 620 and the experimental price markdown 622. The markdown computing device 102 can choose using various methods or processes. In some examples, the markdown computing device 102 can probabilistically choose between the optimized price markdown 620 and the experimental price markdown 622. A threshold probability can be set or a probabilistic rule can be applied. When the threshold probability and/or the probabilistic rule is met, the markdown computing device 102 can choose the experimental price markdown 622 rather than the optimized price markdown 620. The threshold probability and/or the probabilistic rule can be created using any suitable level based on the desire to explore new markdowns rather than exploiting the optimized markdowns. Various characteristics can be used to set the threshold probability and/or the probabilistic rules such as the availability of historical data 602 and/or the performance of the price markdowns in achieving savings during prior periods.

[0064] In the example shown in FIG. 6, the markdown computing device 102 has chosen the experimental price markdown 622 over the optimized price markdown 620. This experimental price markdown 622 can be sent to the retail store and then implemented in the store. The customer’s purchasing behavior (i.e., the number of customers that purchase the product(s) with the marked-down price) can be recorded and stored as historical data 602. As shown in Period B, the data set of historical data 602 results in the markdown computing device 102 determining the price markdowns 610 for Period B. The price markdowns 610 now include the price markdown of 30% that was implemented in Period A. The markdown computing device 102 can then determine the savings 612 for the respective price markdowns 610 for Period B. As shown in this example, the price markdown of 30% resulted in savings of $120. With this new data, the markdown computing device now determines that the optimized price markdown 626 (i.e., the price markdown that results in the greatest savings) is 30%.

[0065] Similarly to Period A, the markdown computing device 102 can also determine an experimental price markdown 628 for Period B . In this example, the experimental price markdown is 25%. The price markdown computing device 102 can now probabilistically choose between the optimized price markdown 626 and the experimental price markdown 628. The same methodology as previously described above (or as otherwise described herein) can be applied to choose between the optimized price markdown 626 and the experimental price markdown 628. This process can continue for subsequent periods until the process stabilizes. The markdown computing device 102 can, for example, determine the performance of the price markdowns by comparing savings that result from implementation of the price markdowns. If the markdown computing device 102 determines that implementation of experimental price markdowns does not provide opportunity for improved results, then the process can stop the determination of experimental price markdowns. In other examples, the quantity of historical data 602 can be evaluated to determine whether enough data exists to implement traditional price markdown selection processes.

[0066] Turning now to FIG. 7, another example process 700 for selecting price markdowns is illustrated. The example shown in FIG. 7 is similar in many respects with that of FIG. 6 and illustrates further details of the methods and systems of the present disclosure. The process 700 begins with the collection or acquisition of historical markdown data 702 for a nascent market. As can be seen in this example, the historical markdown data 702 includes actual markdown data for periods or weeks 1 through 6. The markdown computing device 102 (FIGs. 1, 3 and 4) can be used to perform one or more of the various steps of the process 700. As shown, the markdown computing device 102 can use the historical markdown data 702 to determine further other metrics related to the various actual price markdowns 704. For each actual price markdown 704, the markdown computing device 102 can determine a Reward (Rw). The Reward can be calculated using Equations 1, 2 and 3 shown below.

Eq. 1: Thrwaway where R=Average Return % (716), Q=Markdown Quantity (708) and Rf=recency factor (710) that are each aggregated for each of the Markdown %’s.

[0067] The determination of the Reward for each of the markdowns 704 uses Equation 3 to weigh the average return % based on the quantity of products that were offered with the markdowns as well as based on the recency with which the markdown was offered. The markdown computing device 102 can determine the optimized price markdown 716 that is the markdown 704 that corresponds with the greatest Reward 712. In the example shown, the markdown 704 of 30% corresponds with the greatest Reward 712 of 0.08. Thus, the markdown computing device can select the 30% markdown as the optimized price markdown 716.

[0068] Since the historical data 702 for this example nascent market includes only six weeks of historical markdown data, the markdown computing device can also determine an experimental price markdown 718. The markdown computing device 102 can then choose between the optimized price markdown 716 and the experimental price markdown 718.

[0069] Turning to FIG. 8, the process 700 illustrates an example process of determining the experimental price markdown 718. The experimental price markdown 718 (also termed the next week markdown or MDNW) can be determined using Equation 4 below. where MDLW = Last week’s markdown, ALoR = Change in Loss of Revenue % between two prior periods, and A Waste = change in Waste between two prior periods.

[0070] The markdown computing device can determine a random markdown percentage within a predetermined range of the optimized price markdown 716 or, as in the example shown in Equation 4, within a predetermined range of the previous period’s markdown. In this example, the experimental price markdown 718 is selected randomly within 3% to 5% of the previous period’s markdown. In other examples, the predetermined range can have other values as previously described. The experimental price markdown 718 can also be selected to be higher or lower than the optimized price markdown 716 or higher or lower than the previous period’s markdown. In the example shown, the experimental price markdown 718 is selected from the range that is lower than the previous period’s markdown when the change in Loss of Revenue (ALoR)(FIG. 8, 732) is greater than or equal to the change in Waste (AWaste)(FIG. 8, 736). The experimental price markdown 718 that is higher than the previous period’s markdown is selected when the change in Loss of Revenue (ALoR)(FIG. 8, 732) is less than the change in Waste (AWaste)(FIG. 8, 736). In other examples, the experimental price markdown 718 can be determined using other methods such as randomly choosing the experimental price markdown 718 from other ranges, or varying the experimental price markdown 718 from the optimized price markdown 716 or other suitable methods.

[0071] After the optimized price markdown 716 and the experimental price markdown 718 are determined, the markdown computing device 102 can choose between the two price markdowns. In the example shown, the markdown computing device 102 can choose by applying a probabilistic rule. The markdown computing device can choose the experimental price markdown 718 when the probability is less than a probability threshold p. The probability p can be generated randomly or can be determined by the markdown computing device using Equation 5 below. and RetTY = Return in current period (this year), RetLY = Return in previous period (last year), INVTY = inventory in current period (this year)

[0072] The probabilistic rule can take the form of Equation 6 below. where ArgmaxMD = optimized price markdown and Random MD = experimental price markdown

[0073] In other examples, the markdown computing device can use other probabilistic rules or other methods of choosing between the optimized price markdown 716 and the experimental price markdown 718. Once the final recommended price markdown is chosen, the final recommended price markdowns can be sent to the retailer to be implemented at the retail store. When the final recommended price markdowns are implemented at the retail store, the final recommend price markdown can be a base recommended price markdown 740 (i.e., an average price markdown for all quantities in the analysis time period). The retail store can implement variations to the final recommended price markdown based on the quantity of products that may be available at a given time in the period (over the course of a day, or week, for example). The implemented price markdowns may vary for a given period based on the quantity as shown in FIG. 8 in the table 744.

[0074] Any suitable methodology can be used to determine the appropriate implemented price markdown 744 for a given markdown quantity. In one example, the implemented markdown percentage can be determined by the markdown computing device 102 (or by the retailer’s inventory management system 112) using Equation 7 below. where x = the Markdown Quantity

[0075] For example and as shown in FIG. 8, the base markdown can be determined to be 25%. This base recommended markdown is then applied to the individual product quantities that may be available at the retail store. In this example, the implemented markdowns differ for individual markdown quantities (Md qty) of 12, 8, 5. At these markdown quantities, the implemented markdowns can be 30%, 25% and 20% respectively. Various suitable methods, models or other relationships can be used to translate the base final recommended markdown into the implemented markdowns at each markdown quantity. Examples of such relationships include linear regression, interpolation and other learned models.

[0076] The periods used in the above example and in the related equations 1 through 7 may describe the periods as weeks, or years as shown. In other examples, the methodologies, equations and processes can be applied to other periods as may be appropriate given the individual data that may be available and/or as may be appropriate given the industry and purchasing patterns of the related customers.

[0077] Referring now to FIG. 9, another example method of selecting price markdowns 900 is illustrated. The method can be implemented by the markdown systems previously described. In addition, the steps shown and described can incorporate the processing, elements and/or the models, algorithms and equations previously described as well. For the sake of brevity, the method 900 is described with respect to the markdown system 100. It should be understood, however, that the method can be performed by variations of such system and other systems as contemplated and described by the present disclosure.

[0078] At step 902, the markdown computing device 102 can obtain historical markdown data. The markdown computing device 102 can obtain the historical markdown data using any suitable processing. In one example, the markdown computing device 102 can access and obtain the historical markdown data from a retailer’s central database of retail transactions such as from database 108. As previously described, the historical markdown data can comprise actual markdown data of price markdowns that have already been implemented by the retailer and customer purchasing data that characterizes customers’ purchasing activity related to the implemented actual price markdowns. Such historical data can be available for one or more periods prior to the performance of the method 900. As previously described, such historical markdown data can be limited and method 900 is particularly useful in such nascent markets with limited historical markdown data.

[0079] At step 904, the markdown computing device 102 can determine initial recommended price markdown data. The initial price markdown data can include optimized price markdowns and experimental price markdowns. The optimized price markdowns and the experimental price markdowns can be determined using the example processes described with respect to FIGs. 6, 7 and 8. In other examples, other processing as described herein or as known to one of ordinary skill in the art can also be used.

[0080] At step 906, the markdown computing device 102 can determine final recommended price markdown data. The markdown computing device can determine the final recommended price markdown data using any suitable methodology. In one example, the markdown computing device 102 can choose between the optimized price markdowns and the experimental price markdowns that were determined at step 904. The markdown computing device 102 can, for example, probabilistically choose between the optimized price markdowns and the experimental price markdowns using the methodologies described above. In this example, the final recommended price markdown data can be determined (or chosen) by comparing a current savings of the retailer to a historical savings of the retailer in one or more prior periods.

[0081] At step 908, the markdown computing device 102 can send the final recommended price markdown to at least one retailer. The markdown computing device 102 can send the final recommended price markdowns by any suitable communication methods. In one example, the markdown computing device 102 can send the final recommended price markdowns to the retailer inventory management system 112. In other examples, the markdown computing device 102 can upload or otherwise communicate the final recommended price markdown data to a server or other intermediary component that, in turn, can be accessed by the retailer (or pushed to the retailer).

[0082] While not shown, the retailer can choose to implement the price markdowns at the retail store (or online retail environment) based on the final recommended price markdown data. The retailer may use the final price markdown data in combination with local inventory data that may be available regarding available quantities of product at the local retail location. The final recommended price markdown data can be adjusted or interpolated based on the available quantities of product before being implemented. The retailer can then collect and/or store customer purchasing data that describes actual purchases of product by customers of products that have price markdowns that were based on the final recommended price markdown data. This data can be stored as historical markdown data in the database 108, for example. [0083] At step 910, the markdown computing device 102 can obtain markdown performance data. The markdown performance data can be data that characterizes the performance of the retailer after the final recommended price markdown data has been implemented at the retail store. For example, one or more performance metrics can be determined by the markdown computing device 102 and/or other performance engines. The performance metrics can include Loss of Revenue, Profit, Waste, Reward, and the like. These performance metrics can be included in the markdown performance data that is obtained at step 910.

[0084] At step 912, the markdown computing device 102 can determine if the process is stable. If the process is stable, the method 900 can end. If the process is not stable, the method 900 can return to step 902 and continue to determine final recommended price markdown data for the nascent market. The markdown computing device 102 can use any suitable analysis to determine if the process is stable. For example, the markdown computing device 102 can determine if there is enough historical markdown data to determine that the optimized price markdowns are stable. In other examples, the markdown computing device 102 can determine if the process is stable by comparing the performance metrics that may be obtained at step 910 to performance metrics from previous periods. If such performance metrics do not vary by more than a predetermined threshold (e.g., 20%, 10%, or 5%), the markdown computing device 102 can determine that the process has stabilized. In one example, the process can stabilize within six to eight weeks of model deployment. In other examples, the stabilization can take other periods of time.

[0085] As described above, the methods and system of the present disclosure provide many advantages over existing systems that may determine recommended price markdowns and is particularly advantageous when applied in nascent markets in which historical price markdown data is limited. The methods and systems of the present disclosure can provide increased savings to retailers and can translate to increased customer satisfaction.

[0086] Although the methods described above are with reference to the illustrated flowcharts, it will be appreciated that many other ways of performing the acts associated with the methods can be used. For example, the order of some operations may be changed, and some of the operations described may be optional. [0087] In addition, the methods and system described herein can be at least partially embodied in the form of computer-implemented processes and apparatus for practicing those processes. The disclosed methods may also be at least partially embodied in the form of tangible, non-transitory machine -readable storage media encoded with computer program code. For example, the steps of the methods can be embodied in hardware, in executable instructions executed by a processor (e.g., software), or a combination of the two. The media may include, for example, RAMs, ROMs, CD-ROMs, DVD-ROMs, BD-ROMs, hard disk drives, flash memories, or any other non-transitory machine -readable storage medium. When the computer program code is loaded into and executed by a computer, the computer becomes an apparatus for practicing the method. The methods may also be at least partially embodied in the form of a computer into which computer program code is loaded or executed, such that, the computer becomes a special purpose computer for practicing the methods. When implemented on a general-purpose processor, the computer program code segments configure the processor to create specific logic circuits. The methods may alternatively be at least partially embodied in application specific integrated circuits for performing the methods.

[0088] The term model as used in the present disclosure includes data models created using machine learning. Machine learning may involve training a model in a supervised or unsupervised setting. Machine learning can include models that may be trained to learn relationships between various groups of data. Machine learned models may be based on a set of algorithms that are designed to model abstractions in data by using a number of processing layers. The processing layers may be made up of non-linear transformations. The models may include, for example, artificial intelligence, neural networks, deep convolutional and recurrent neural networks. Such neural networks may be made of up of levels of trainable filters, transformations, projections, hashing, pooling and regularization. The models may be used in large-scale relationshiprecognition tasks. The models can be created by using various open-source and proprietary machine learning tools known to those of ordinary skill in the art.

[0089] The foregoing is provided for purposes of illustrating, explaining, and describing embodiments of these disclosures. Modifications and adaptations to these embodiments will be apparent to those skilled in the art and may be made without departing from the scope or spirit of these disclosures.