System and method for real-time bidding

Abstract

A real-time bidding method includes receiving user input data, generating a first machine learning model that generates a predicted expected performance based on the user input data, and adjusting at least one bid for at least one of at least one product and at least one keyword associated with at least one marketplace based on the predicted expected performance.

Description

[0001] Cross-references to related applications

[0002] This PCT patent application claims the benefit and priority of U.S. Provisional Patent Application Serial No. 63 / 563,090, filed March 8, 2024, the entire disclosure of which is incorporated herein by reference. Technical Field

[0003] This disclosure generally relates to business systems and methods, and more specifically, to systems and methods for real-time bidding. Background Technology

[0004] A commercial system is well-known in the field and is an effective means of facilitating the exchange of products, goods, and services from one party to another. Typically, a commercial system is embodied in a market, where many products are offered for sale, and people as customers can shop or browse products and select items to purchase. Such a market can be represented by eBay. Amazon Wayfair Costco Walmart and Target This involves company management, etc. With the advent of digital marketplaces, sellers can list products for purchase to anyone with an internet connection. Often, many sellers will offer the same or similar products. Shoppers (e.g., users accessing digital marketplaces via the internet) can categorize and browse all these products to find what they are looking for. Summary of the Invention

[0005] The various systems and methods disclosed herein were developed in response to the current state of technology, and in particular to problems and needs in the field that have not yet been fully addressed by the currently available digital markets.

[0006] One aspect of the disclosed implementation includes a real-time bidding system comprising a computing device and a memory. The memory includes instructions that, when executed by the computing device, cause the real-time bidding system to: receive user input data; generate at least one model for a target expenditure value based on the user input data; and adjust at least one bid for at least one of at least one keyword and at least one product based on the output from the at least one model.

[0007] Another aspect of the disclosed implementation includes a real-time bidding system. The system includes a processor and a memory. The memory includes instructions that, when executed by the processor, cause the processor to: receive user input data; generate a first machine learning model based on the user input data to generate predicted expected performance; and adjust at least one bid for at least one of at least one keyword and at least one product associated with at least one market based on the predicted expected performance.

[0008] Another aspect of the disclosed implementation includes a real-time bidding method. The method includes: receiving user input data; generating a first machine learning model, the first machine learning model generating predicted expected performance based on the user input data; and adjusting at least one bid for at least one of at least one keyword and at least one product associated with at least one market, based on the predicted expected performance.

[0009] These and other aspects of this disclosure are disclosed in the following detailed description of embodiments, the appended claims, and the accompanying drawings. Attached Figure Description

[0010] This disclosure can be best understood by reading in conjunction with the accompanying drawings and the following detailed description. It should be emphasized that, by convention, the various features in the drawings are not drawn to scale. Instead, for clarity, the dimensions of the various features have been arbitrarily enlarged or reduced.

[0011] Figure 1 A schematic block diagram of a system based on the principles of this disclosure is shown in general.

[0012] Figure 2A This illustrates the principles based on this disclosure. Figure 1 A schematic block diagram of a computing device in the form of a smartphone.

[0013] Figure 2B This illustrates the principles based on this disclosure. Figure 1 Desktop computer-type computing devices and Figure 1 A schematic block diagram of a server in the first server form.

[0014] Figure 3 This is a schematic block diagram illustrating computing devices and servers used in operating a digital marketplace in accordance with the principles of this disclosure.

[0015] Figure 4 This is a schematic block diagram illustrating computing devices and servers used when hosting a digital marketplace that includes attributes of both target and competing products, in accordance with the principles of this disclosure.

[0016] Figure 5This is a schematic block diagram illustrating a computing device including a graphical user interface according to the principles of this disclosure.

[0017] Figure 6 This is a flowchart illustrating a method for evaluating a product based on the principles of this disclosure.

[0018] Figure 7 This is a flowchart illustrating a method for providing a competitive evaluation of a target product in the market, based on the principles of this disclosure.

[0019] Figure 8 This is a schematic block diagram illustrating computing devices and servers used in operating a digital marketplace in accordance with the principles of this disclosure.

[0020] Figure 9 It is a graphical representation of multiple search terms plotted at points according to the principles of this disclosure, which represent the frequency and similarity of search terms associated with the target product relative to competing products.

[0021] Figure 10 It is a graphical representation of multiple search terms plotted at the points described below, based on the principles of this disclosure, representing the relevance and search volume of search terms associated with the target product relative to competing products.

[0022] Figure 11 It is a graphical representation of the model output based on the principles of this disclosure.

[0023] Figure 12 This is a flowchart that generally illustrates a real-time bidding method based on the principles of this disclosure. Detailed Implementation

[0024] The following discussion relates to various embodiments of this disclosure. While one or more of these embodiments may be preferred, the disclosed embodiments should not be construed as or otherwise used to limit the scope of this disclosure, including the claims. Furthermore, those skilled in the art will understand that the following description has broad application, and the discussion of any embodiment is intended only as an illustrative example of that embodiment and is not intended to imply that the scope of this disclosure, including the claims, is limited to that embodiment.

[0025] In commercial systems, pay-per-click advertising operates on a cost-per-click auction basis, where bids are set at the target level (e.g., keywords or products). Using this auction, bids are set to be competitive, and ads are served until the campaign budget cap is reached. Once the campaign budget cap is reached, the ads are turned off and can no longer be served.

[0026] Typically, industry standard practice is to increase bids to spend more and decrease bids to spend less, usually using bidding algorithms adjusted based on efficiency goals. Such bid adjustments are made without considering the campaign budget and require manual input to determine the desired amount to spend.

[0027] Therefore, systems and methods configured to improve real-time bidding, as described herein, may be desirable. In some implementations, the systems and methods described herein can be configured to provide optimal cost-per-click (CPC) bids that simultaneously or substantially simultaneously consider expected spending and target efficiency. The systems and methods described herein can be configured to use machine learning, best practices in data extraction, and data monitoring to create optimal CPC systems.

[0028] In some implementations, the systems and methods described herein can be configured to maximize user performance or success criteria while spending a user-defined budget, provided that the expenditure exceeds a minimum threshold.

[0029] The systems and methods described herein can be configured to receive user input from a user computing device. For example, a user can interact with the user computing device to provide input at a user interface, including desired spending and a time period for that desired spending amount. For instance, a user can specify for a single day, a single week, a single month, etc. 100. User input may also include product and / or product information associated with the advertising budget.

[0030] The systems and methods described herein can be configured to receive input indicating budget, time period, and applicable market for the product via user selections made using a user interface. The systems and methods described herein can also be configured to receive user selections indicating the minimum performance framework to be achieved per unit of spending. For example, per unit of spending... 1 Minimum Acquisition 5% income, per expenditure 1. Acquire at least 2 new customers, per cost 1. Achieve 300 user impressions, etc. It should be understood that user input may include any suitable input for fine-tuning or further guiding the systems and methods described herein, such as specifying which target groups should receive different minimum performance requirements or budget expectations, or other suitable information.

[0031] The systems and methods described in this paper can be configured to estimate the distribution of relative expected ad spend based on user input for a specified time period, product, and market to determine whether the spend is above, below, or in line with targets. For example, if the time period (e.g., a user-specified time period) is a single day, 3 PM will typically have higher expected spend than 3 AM, and the systems and methods described in this paper can be configured to backtrack relative spend throughout the day from a machine learning model. The output of the model is a predicted distribution of user demand for each sub-time period (e.g., hour) in the market for a specified duration.

[0032] To predict relative spending for each sub-period, machine learning models use tree-based, neural network, or Bayesian architectures to ingest data such as historical ad spending by sub-period, date and hour-related elements, product and brand categories, demand behavior for similar products, and advertising behavior for similar products. For example, if a user expects to spend on a set of products in the market on a single day... If the value is 100, then given the product's history, category, brand, and other relevant information, the model expects a value of 5.3. Advertising spending will occur in a one-hour sub-period starting at 9:00 AM. As real-time data arrives, the model continues to update and adjust to adapt to changing conditions.

[0033] The systems and methods described in this paper can be configured to generate a set of potential targets to be considered for each product. The systems and methods described in this paper can be configured to generate a large number of potential targets and their associated data and correlations, which satisfy minimum filtering correlation for consideration in the process.

[0034] The systems and methods described in this paper can be configured to generate two models. The systems and methods described in this paper can be configured to use a first model of the two generated models to estimate a performance metric given different potential bid levels. The systems and methods described in this paper can be configured to use a second model of the two generated models to estimate the relevant cost of each objective given different potential bid levels. Additionally or alternatively, the two models can be combined into a single model to generate an output indicating performance per unit of currency.

[0035] Each performance metric can include a ratio of performance per unit of currency spent. The systems and methods described in this paper can be configured to maximize the performance ratio while meeting expected spending in response to actual spending exceeding a minimum threshold ratio. For example, the systems and methods described in this paper can be configured to... Maximize the revenue from every dollar spent while achieving 100.

[0036] The first model (e.g., the performance model) can output the expected performance for any given bid and target (e.g., (50% of revenue). The first model can use tree-based, neural network, Bayesian, or other architectures. The input to the first model may include: output metrics, historical performance and bids of the target, historical performance of similar targets, product information, competitor information, category information, and other target attributes.

[0037] The systems and methods described in this paper can be configured to fine-tune a first model using training inputs to satisfy a custom loss function based on user preferences and historical behavior captured through the interface. Specifically, the risk preference of the advertising system for performance outcomes may be both asymmetric and variable with magnitude. For example, users may not treat ROI above or below the target by 0.1 indiscriminately, and a deviation of 0.5 from the target may not simply be 5 times worse than a deviation of 0.1. Therefore, the first model is trained and fine-tuned based on user preferences for performance. User preferences can be explicitly captured by directly asking questions for each user, or inferred from the preferences of a general user group (e.g., average behavior).

[0038] The second model (e.g., the cost model) can output the expected cost in a specified currency unit for any given bid and target (e.g., (Cost of 10). Unlike the performance model, this is always cost, not specified user input. The second model can use tree-based, neural network, Bayesian, or other architectures. The input to the second model can include: output metric, historical cost and bid of the target, historical performance of similar targets, product information, competitor information, category information, other target attributes, etc.

[0039] The systems and methods described in this paper can be configured to fine-tune a second model using training inputs to satisfy a custom loss function based on user preferences and historical behavior captured through the interface. Specifically, the risk preference of the advertising system for spending outcomes may be both asymmetric and variable with size. For example, users may not be willing to spend more than their target. 100 and below target expenditure 100% indiscriminate treatment, and deviation from the target. The result of 500 may not be simply more than deviation. The result for 100 is 5 times worse. Therefore, the second model can be trained and fine-tuned based on users' preferences for spending on both size and direction.

[0040] The system and method described in this paper can be configured to combine these two models to generate both the ratio and the expected expenditure. For example, bidding on product ABC for the target keyword "example". 0.50 was estimated to produce 3.00 revenue and A cost of 1.00 results in a revenue-to-cost ratio of 3. If the user requires a performance ratio of at least 4, only the lower bid and its associated lower revenue, lower cost, but higher ratio are considered. As described, depending on conditions and accuracy, a single model that directly estimates this ratio can also be used. The final output of this model is a downward-sloping curve (e.g., as...). Figure 11 (as shown in the diagram), where total expenditure is the input across all targets, and the performance ratio per monetary unit serves as the output, as depicted below. The systems and methods described herein can be configured to generate this output for any combination of desired time periods or sub-periods, markets, products, product groups, etc. Performance results below the desired performance threshold can be discarded.

[0041] In some implementations, the systems and methods described herein can be configured to: select all targets and bids associated with the expected expenditure and performance, given the expected expenditure for each sub-period and the ability to select a set of bids and targets that will achieve that expected expenditure with the highest performance, wherein the estimated expenditures will be aggregated to the expected expenditure. This set of targets and bids is then sent to the market for each sub-period.

[0042] The systems and methods described in this paper can be configured to provide real-time spending and performance monitoring throughout the entire time period to achieve expected user spending that is both above minimum and maximized performance. The systems and methods described in this paper can be configured to incorporate the latest performance and spending data into the model to generate updated estimates of performance and costs. As the real-time performance of both spending and performance is generated, the latest data can be fed back into the model for reinforcement learning, adjustments to the latest market conditions, and improvements to future forecasts. The updated model can then make forecasts for the remaining sub-periods of the day.

[0043] The systems and methods described in this paper can be configured to adjust bids and target aggregations up or down throughout the time period based on model estimates of the target spending pacing, to achieve spending that is as close as possible to the expected spending. For example, the user's target spending over a 24-hour period is... 100, Minimum performance metric is 4, 12 hours have elapsed, Expenditure has been incurred. 55. The performance metric is 6, and the expected expenditure from the second-step model is 44. (or In scenario 44), the systems and methods described herein may exceed the intended target expenditure. 11. First, the latest performance and expenditure can be incorporated into the model. Second, since expenditure is exceeding the estimated target pace, it is possible to choose to adjust the estimated... A set of lower bids corresponding to 45 expenditures is used to complete the day and achieve the desired expenditure. 55 (to date)+ 45 (Given the updated target, bid, and estimated new estimate) = 100. This could lead to higher estimation performance, as bidding has become more conservative and performance per currency unit has improved.

[0044] Reference Figure 1 The schematic block diagram illustrates a system 100 based on the principles of this disclosure. Figure 1 As shown, system 100 can be used to provide benefits to one or more users 110, which may include a first user 112, a second user 114, a third user 116, and a fourth user 118. Each of the users 110 may use one of a variety of computing devices 120, which may include any of a variety of devices that perform computing steps, including but not limited to a desktop computer 122 used by the first user 112, a laptop computer 124 used by the second user 114, a smartphone 126 used by the third user 116, a camera device 128 used by the fourth user 118, etc. The systems and methods presented herein can be performed on any type of computing device.

[0045] Computing devices 120 may optionally connect to each other and / or other resources. Such connections may be wired or wireless and may be achieved using any known wired or wireless communication standards, including but not limited to Ethernet, 802.11a, 802.11b, 802.11g, and 802.11n, Universal Serial Bus (USB), Bluetooth, cellular, Near Field Communication (NFC), Bluetooth Smart, ZigBee, etc. Figure 1 In the illustration, wired communication is shown with solid lines and wireless communication is shown with dashed lines, as an example.

[0046] Routing can be facilitated by using router 130 and / or in other ways. Figure 1Communication between various components. Router 130 can be of any type known in the art and can be designed for wired and / or wireless communication via any known communication standard, including but not limited to those listed herein. Router 130 may include, for example, a first router 132 facilitating communication to and / or from desktop computer 122; a second router 134 facilitating communication to and / or from laptop computer 124; a third router 136 facilitating communication to and / or from smartphone 126; and a fourth router 138 facilitating communication to and / or from camera device 128.

[0047] Router 130 facilitates communication between computing device 120 and one or more networks 140, which can include any type of network, including but not limited to local area networks (LANs) (e.g., LAN 142) and wide area networks (e.g., WAN 144). In some embodiments, LAN 142 may be a network serving entities such as businesses, non-profit entities, government organizations, etc. WAN 144 can provide communication for multiple entities and / or individuals, and in some embodiments, may be the Internet. LAN 142 can communicate with WAN 144. If needed, one or more routers or other devices may be used to facilitate such communication.

[0048] Network 140 may store information on server 150 or other information storage devices. As shown, first server 152 may be connected to local area network 142 and therefore may communicate with devices connected to local area network 142, such as desktop computer 122 and laptop computer 124. Second server 154 may be connected to wide area network 144 and therefore may communicate with devices connected to wide area network 144, such as smartphone 126 and camera device 128. If needed, second server 154 may be a web server providing web pages, web connectivity services, executable code designed to operate on the Internet, and / or other functions that facilitate the provision of information and / or services on wide area network 144.

[0049] Reference Figure 2A The schematic block diagram illustrates an exemplary computing device 120, which enables the implementation of the present disclosure in a standalone computing environment. The computing device may be, for example... Figure 1 The smartphone 126. However, this specification envisions that the computing device 120 may include... Figure 1 Any one or any other type of computing device among those computing devices 120 described herein.

[0050] As shown, smartphone 126 may include processor 210 designed to execute instructions on data. Processor 210 may be of a variety of types, including microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), etc., having an x86-based architecture or other architectures known in the art. Processor 210 may optionally include multiple processing elements or "cores." Processor 210 may include a cache providing temporary storage for data generated incidentally as a result of the operation of processor 210.

[0051] The smartphone 126 may also include memory 220, which may be volatile memory, such as random access memory (RAM). Memory 220 may include one or more memory modules. Memory 220 may include executable instructions, data referenced by such executable instructions, and / or any other data that may advantageously make the processor 210 readily accessible.

[0052] The smartphone 126 may also include a data storage 230, which may be a non-volatile memory such as a hard disk drive or flash memory. The data storage 230 may include one or more data storage elements. The data storage 230 may store executable code, such as an operating system and / or various programs to run on the smartphone 126. The data storage 230 may also store data to be used by such programs. For the systems and methods of this disclosure, the data storage 230 may store computer-executable code associated with the evaluation module 232, text analysis module 238, filtering module 235, comparison module 234, recommendation module 236, and competitiveness score generation module 233. The data storage 230 may also include data associated with: descriptive terms 241 related to the target product and / or competing products, relevant descriptive terms 242 associated with either the target product or competing products, competitiveness scores 239, and actionable reports 237. This data stored by the data storage 230 may be retained on the data storage 230 for any length of time, and some data may be created or overwritten at any time to facilitate the methods described herein.

[0053] Smartphone 126 may also include one or more wired transmitters / receivers 240, which can facilitate the connection of smartphone 126 with any other device (e.g., Figure 1 Wired communication between other computing devices 120, servers 150, and / or routers 130. The wired transmitter / receiver 240 can communicate via any known wired protocol (including but not limited to...). Figure 1 The wired transmitter / receiver 240 may communicate with any of the wired protocols described herein. In some implementations, the wired transmitter / receiver 240 may include an Ethernet adapter, a Universal Serial Bus (USB) adapter, etc.

[0054] Smartphone 126 may also include one or more wireless transmitters / receivers 250, which facilitates communication between Smartphone 126 and any other device (e.g., Figure 1 Wireless communication between other computing devices 120, servers 150, and / or routers 130. The wireless transmitter / receiver 250 can communicate via any known wireless protocol (including but not limited to...). Figure 1 The wireless transmitter / receiver 250 may communicate with any of the wireless protocols described herein. In some implementations, the wireless transmitter / receiver 250 may include a Wi-Fi adapter, a Bluetooth adapter, a cellular adapter, etc. Either the wired transmitter / receiver 240 or the wireless transmitter / receiver 250 may be associated with the network interface device (NID) 280. The network interface device 280 may provide connectivity to any network via the Internet, such as a wide area network (WAN), a local area network (LAN), a wireless local area network (WLAN), a wireless personal area network (WPAN), a wireless wide area network (WWAN), or other networks.

[0055] Smartphone 126 may also include one or more user inputs 260, said one or more user inputs 260 from a user (e.g. Figure 1 Input is received by any user in user 110. User 110 described herein may be referred to as the seller of the target product. User input 260 may be integrated into smartphone 126, or may be separate from smartphone 126 and connected to smartphone 126 via a wired or wireless connection that may operate via wired transmitter / receiver 240 and / or wireless transmitter / receiver 250. User input 260 may include elements such as touch screen, buttons, keyboard, mouse, trackball, trackpad, stylus, digitizer, digital camera, microphone, and / or other user input devices known in the art.

[0056] Smartphone 126 may also include one or more user outputs 270, which output to a user (e.g. Figure 1The user output 270 provides output to any user in user 110. The user output 270 may be integrated into the smartphone 126, or it may be separate from the smartphone 126 and connected to the smartphone 126 via a wired or wireless connection, which may operate via a wired transmitter / receiver 240 and / or a wireless transmitter / receiver 250. The user output 270 may include elements such as a display screen, speaker, vibration device, LED or other light, and / or other output devices known in the art. In some embodiments, one or more of the user inputs 260 may be combined with one or more of the user outputs 270, as is the case with a touchscreen. In some embodiments, the user output 270 may present a graphical user interface to a user through which the user interacts with the smartphone 126 to influence the methods and processes described herein.

[0057] The smartphone 126 may include various other components not shown or described herein. Those skilled in the art will recognize, with the aid of this disclosure, that in addition to combining… Figure 2A Any such component, other than those shown and described or as an alternative to these components, may be used to perform the contents of this disclosure.

[0058] The smartphone 126 can execute the contents of this disclosure in an independent computing environment, i.e., without relying on communication with other devices such as other computing devices 120 or server 150. This specification also envisions that any one of the evaluation module 232, competitiveness score generation module 233, comparison module 234, filtering module 235, recommendation module 236, and text analysis module 238 can be distributed across multiple computing devices (e.g., ...). Figure 1 Between computing devices 120 and / or any server (e.g., Figure 1 Between 150 and 150). In other embodiments, the present disclosure can be utilized in different computing environments. (The following is a continuation of the previous sentence, but the context is unclear: "will be combined with...") Figure 2B An example of a client / server environment is shown and described.

[0059] Reference Figure 2B The schematic block diagram shows Figure 1 Desktop computer 122 form computing device and Figure 1 The first server 152 is a server in the form of a computing device that can cooperate to implement the practices of this disclosure having a client / server architecture. As shown, the desktop computer 122 can be a "dumb terminal" that is configured to operate together with the first server 152.

[0060] Therefore, desktop computer 122 may only have the ability to communicate with users (e.g., Figure 1The desktop computer 122 provides the hardware required for the first user interface 112 and communication with the first server 152. Therefore, the desktop computer 122 may include one or more user inputs 260, one or more user outputs 270, one or more wired transmitters / receivers 240, and / or one or more wireless transmitters / receivers 250. Similarly, any of the wired transmitters / receivers 240 or wireless transmitters / receivers 250 may be associated with the NID 280a. The NID 280a may provide connectivity to any network via the Internet, such as a wide area network (WAN), local area network (LAN), wireless local area network (WLAN), wireless personal area network (WPAN), wireless wide area network (WWAN), or other network in which the first server 152 is part. These components can be combined as follows... Figure 2A As described.

[0061] Computational functions (excluding those events that receive input from the user and deliver output to the user) can be performed wholly or partially at the first server 152. Therefore, the processor 210, memory 220, data storage 230, wired transmitter / receiver 240, and wireless transmitter / receiver 250 can be housed in the first server 152. These components can also be combined as follows. Figure 1 As described by A.

[0062] In operation, desktop computer 122 can receive input from a user via user input 260. The user input can be delivered to first server 152 via wired transmitter / receiver 240 and / or wireless transmitter / receiver 250. The user input can be further transmitted by any intermediate device, such as first router 132 and any other device in local area network 142 that needs to transmit user input from first router 132 to first server 152.

[0063] The first server 152 can perform any necessary processing steps in response to receiving user input. The first server 152 can then send user output to the user via wired transmitter / receiver 240 and / or wireless transmitter / receiver 250. This user output can be further transmitted by any intermediate device, such as the first router 132 and any other device in the local area network 142 (or alternatively, the wide area network 144) that needs to transmit the user output from the first server 152 to the first router 132. The user output can then be provided to the user via user output 270. In some embodiments, user output 270 can present a graphical user interface to the user, displaying a list of relevant descriptive terms 242 for the target product and competing products, and displaying an actionable report describing the expected performance of the target product in the computer network market relative to at least one naturally ranked competing product also presented in the computer network market, according to the methods described herein.

[0064] Reference Figure 3 The schematic block diagram illustrates a computing device 322 (similar to) operating a digital market according to the principles of this disclosure. Figure 1 Any of the computing devices shown) and server 350 (similar to Figure 1 (Any of the servers shown), the computing device and the server can collaborate to implement the practices of this disclosure with a client / server architecture. As shown, computing device 322 can be operatively coupled to server 350 via NID 380 as described herein. This operative coupling allows computing device 322 to access, when appropriate, a digital marketplace 382 on which target and competing products are sold. Digital marketplace 382 can be any web-accessible website listing multiple products, allowing users to view products, rate products, purchase products, and perform other tasks related to digital commerce when they visit such websites. Digital marketplace 382 can be, for example, eBay. Amazon Wayfair Costco Walmart and Target The company manages the product. After purchasing the product, the consumer can have it delivered to their home or workplace for consumption. In some implementations, the digital marketplace 382 can be any one of multiple websites for which the server 350 provides storage and processing resources.

[0065] As described herein, computing device 322 may include processor 310, memory 320, user input 360, user output 370, and data storage 330, and operates similarly to a combination of Figure 2A and Figure 2B The similar elements described herein. Data storage 330 may include the modules described herein, including evaluation module 332, competitiveness score generation module 333, comparison module 334, filtering module 335, recommendation module 336, and text analysis module 338.

[0066] During operation, the evaluation module 332 can evaluate certain attributes of the target product. As described herein, the target product is a specific product that a user (e.g., a seller) of the computing device 322 is seeking to discover its competitiveness in a particular market. For example, the target product could be a product that a user is selling or wants to sell on a digital marketplace 382 hosted by server 350. To determine the competitiveness of the target product, the evaluation module 332 can access certain data about the target product present on server 350. The evaluation module 332 can access this data by sending a data request via NID 380 through a wired connection (e.g., via wired transmitter / receiver 340) or a wireless connection (e.g., via wireless transmitter / receiver 350).

[0067] Data requests can be requests for attributes of the target product. While any number of attributes about the target product can be requested, evaluation module 332 can request specific attributes that will be used to develop an actionable report 337 regarding the target's competitiveness in digital marketplace 382. A first attribute can describe a rating provided by at least one buyer of the target product on digital marketplace 382. Typically, digital marketplace 382 provides consumers with a graphical user interface (GUI) that allows those consumers to rate products they purchase on digital marketplace 382. In a particular implementation, consumers / buyers of the target product can use a 5-star rating system to rate the target product. A one-star rating would indicate that the consumer / buyer rated the target product poorly, while a five-star rating would indicate that the consumer / buyer rated the target product very well. Therefore, evaluation module 332 can use each star rating, or the average of those star ratings, as input to create actionable report 337.

[0068] The second attribute may include reviews and / or user-generated content associated with the target product. Similarly, the digital marketplace 382 typically provides a GUI that allows consumers of the target product to input text describing their experience with the product. This text may include specific positive or negative keywords describing the consumer's experience with the target product. In some implementations, using this data, the evaluation module 332 may enable the text analysis module 338 to parse each review against these keywords describing the target product. Furthermore, the text analysis module 338 may also extract keywords describing certain features of the target product. As an example, the text analysis module 338 may extract the phrase "ergonomic handle," which not only describes that the target product includes a handle but also that the handle is an "ergonomic" handle, giving the consumer who gave the review a perception that they liked the fit or feel of the target product.

[0069] The third attribute can be similar to the second attribute, in that evaluation module 332 determines the number of reviews associated with the target product presented on the digital marketplace 382. The number of reviews can indicate the level of engagement with the target product, whether it is derogatory or positive. Together with the text content of these reviews, the number of reviews associated with the target product can be used to help create actionable reports based on engagement with the target product within the digital marketplace 382.

[0070] The fourth attribute can include the listed price of the target product. While the amount charged for purchasing the product may not indicate its value, the amount charged relative to other similar competing products may indicate its value or current price point (whether accurate or inaccurate).

[0071] The fifth attribute may also include the target product's ranking relative to at least one organic ranking competitor. This ranking can be the result of the target product's average or cumulative rating relative to organic ranking competitors. Typically, digital marketplaces 382 allow buyers to list organic ranking competitors and the target product by average rating. By doing so, the evaluation module 332 can understand the target product's ranking relative to at least one organic ranking competitor and use this information to develop an actionable report 337.

[0072] Evaluation module 332 can also determine at least one similar attribute of an organically ranked competing product, said attribute being similar to those attributes that evaluation module 332 has found for the target product. In the context of this specification, the term "organically ranked competing product" is intended to be understood as any product ranked on the digital marketplace 382 based on consumer reviews. Therefore, an "organically ranked" competing product is one that has naturally ranked based on reviews provided by past consumers, rather than those that may have gained such a position by being given "top shelf" priority after payment. This organic ranking nature of products on the digital marketplace 382 is generally intended to provide potential consumers with evidence that others favor the product. A "competing" product is any product similar to the target product but sold by a seller other than the seller of the target product. The "similarity" of the target product relative to at least one organically ranked competing product depends on the data obtained by text analysis module 338, and particularly on the analysis of descriptive terms 341 associated with each of these types of products. In a particular embodiment, text analysis module 338 can also obtain descriptive data associated with each of the target product and organically ranked competing products based on a list of them. Similarly, the digital marketplace 382 allows for the posting of product descriptions alongside each product, outlining its features, physical characteristics, and perceived advantages as a premium product. All of this is presented as text information on the GUI to potential consumers to entice them to purchase. The text analytics module 338 can analyze this text and use parsing to extract keywords for comparing text associated with the target product with text associated with competing products in the organic rankings.

[0073] When computing device 322 obtains attributes associated with the target product and at least one organically ranked competitor product via evaluation module 332, descriptive terms 341 describing these attributes may be listed for use by filtering module 335 in some embodiments. Filtering module 335 may be used to filter descriptive terms 341 to only those relevant to the purchase of the target product in digital marketplace 382. For example, some descriptive terms 341 may correctly or incorrectly include the color or color scheme of the target product or organically ranked competitor product. While some consumers may appreciate a particular color of a product, these may not be the determining factors for attracting consumers to purchase the target product or organically ranked competitor product. This may be especially true when purchase history associated with the target product or organically ranked competitor product indicates that no product of a particular color has significantly outsold another color. In this specific example, although the color of the product is a descriptive term 341 that the text analysis module 338 has already parsed from the product, it may not necessarily be a relevant descriptive term 341, and such information can be filtered out by the filtering module 335 to obtain only those relevant descriptive terms 342 that are associated with any target product or organic ranking competing products.

[0074] In a more general example, the filtering module 335 can narrow down the range of descriptive terms 341 of interest by analyzing metrics collected on sufficiently "mature" keywords (e.g., sales > 2), as emerging keywords may lack sufficient data to influence predictions of purchases of the target product or organically ranked competing products. Based on the keywords used to search for the product, click-through rates and conversion rates (clicks leading to purchases) associated with any given product can be considered. In these examples, the lack of data on a particular descriptive term 341 can also filter out that particular descriptive term 341 to obtain relevant descriptive terms 342 as described herein. It should also be understood that descriptive terms 341 can be filtered by the filtering module 335 for any other reason to obtain relevant descriptive terms 342, and these other reasons are considered in this specification.

[0075] Once the relevant descriptive terms 342 are determined, these relevant descriptive terms 342 can be sent to the comparison module 334 to compare those relevant descriptive terms 342 of the target product with those relevant descriptive terms 342 associated with at least one organic ranking competitor product. Although this specification describes the comparison process as being performed between a single organic ranking competitor product (e.g., "at least one") and the target product, any number of organic ranking competitors products can be compared with the target product. In a particular example, the comparison module 334 can compare the top 10 organic ranking competitors products with the target product.

[0076] During the execution of comparison module 334 by processor 310, descriptive terms 341 may be compared to generate a competitiveness score 339 using competitiveness score generation module 333 executed by processor 310. In some embodiments, the competitiveness score may use any processing or algorithm used to define the degree to which a target product can or cannot compete with any discovered naturally ranked competing products.

[0077] During operation, the recommendation module 336 may receive the competitiveness score 339 and other data from the digital marketplace 382 hosted by server 350. This other data may include revenue data associated with organically ranked competing products and the target product (if available). For example, if a click-through rate for any given product (e.g., the target product or an organically ranked competing product) leads to a purchase, this conversion rate data, along with the product's pricing data, may be passed to the recommendation module 336. The recommendation module 336 can then provide a recommendation describing whether the target product can (or cannot) compete with at least one organically ranked competing product. In some implementations, a threshold competitiveness score may be set such that the report provided by the recommendation module 336 instructs the seller of the target product whether to continue selling the product on the digital marketplace 382. Alternatively, if the competitiveness score does not meet the threshold, the competitiveness score generation module 333 may not forward the competitiveness score to the recommendation module 336 to generate an actionable report 337. Alternatively or additionally, if the competitiveness score does not meet the threshold, the competitiveness score generation module 333 can transmit a threshold failure signal to the recommendation module 336, which indicates the non-competitive status of the target product. When the threshold competitiveness score is not met, the recommendation module 336 can provide an instruction to the seller that it is not recommended that the seller start or continue selling the target product on the digital marketplace 382.

[0078] Upon reaching a threshold competitiveness score, the recommendation module 336 can provide additional economic data describing the price point and ACoS statistics to increase revenue. Similarly, the seller of the target product may not know the appropriate target advertising sales cost (ACoS) to meet or at what price point to sell the target product to see long-term revenue rather than short-term profit. The recommendation module 336 provides this information based on the competitiveness score 339 generated by the competitiveness score generation module 333 and the revenue data received from the digital marketplace 382. In a specific example, the revenue potential of the target product can be determined by the recommendation module 336 by calculating the advertising spend profit margin, advertising spend potential, and revenue potential. The advertising spend profit margin can be calculated by multiplying the target ACoS by the price of the target product. The target ACoS can be determined and set by the seller based on available capital, or it can be set by the seller based on the share of revenue received from sales of the target product from the digital marketplace 382 to date and manufacturing costs. The advertising spend potential can then be calculated by multiplying the monthly opportunity units (OU) by the spending profit margin. The monthly OU can be calculated as a result of the conversion rate of clicks to the target product resulting from sales after a buyer views the product. Revenue potential can then be calculated by multiplying OU by the price of the target product. The revenue potential of this target product and each of its organically ranked competitors can be ranked to determine the target product's placement within the digital marketplace 382.

[0079] In some implementations, the recommendations presented by recommendation module 336 (e.g., actionable report 337) can be refined by inputting an estimated bid amount from digital marketplace 382 to “win” ad placement for the target product. Digital marketplace 382, ​​while selling products, may also engage in presenting ads to potential buyers of one or more products. These ads may appear as banners or other sub-sections of the GUI presented to the buyer, or as pop-up ads. These forms of advertising present alternative products that potential buyers are seeking to purchase in real time. These ads may present the target product and persuade buyers to buy the target product rather than competitors' products. Therefore, investment may be required to increase purchase instances for the target product. The system and method may also present the target product to sellers on actionable report 337: based on keywords associated with the target product and entered in searches by potential users, how much additional investment might be needed to win ad placement. For example, the required investment can be calculated by multiplying the estimated bid amount by the target product's click-through rate and the exposure (e.g., usage) of specific keywords used to search for those products that are associated with the target product and competing products in organic rankings. Return on investment (ROI) can then be calculated by subtracting the required investment from the payback period and multiplying it by the advertising expenditure potential. Products with no (or low) potential will receive suggested outputs through the recommendation module 336 and its actionable report 337, explaining why they are not competitive or have poor conversion rates, so that these attributes of the target product can be improved for future potential, or the spending on selling the target product can be reallocated for other purposes.

[0080] Figure 4 This is a schematic block diagram illustrating computing device 420 and server 452 hosting attributes of a target product and competing products in a digital marketplace 482, which can collaborate to implement the practices of this disclosure with a client / server architecture. As described herein, evaluation module 432 can evaluate certain attributes of the target product. The target product, as described herein, is a specific objective that a user (e.g., a seller) of computing device 420 is seeking to discover the competitiveness of a product in a particular market. For example, the target product could be a product that a user is selling or wants to sell on the digital marketplace 482 hosted by server 452. To know the competitiveness of the target product, evaluation module 432 can access certain data about the target product present on server 452. Evaluation module 432 can access the data by sending a data request via NID 480 via a wired (e.g., via wired transmitter / receiver 440) connection or a wireless (e.g., via wireless transmitter / receiver 450) connection.

[0081] Data requests can be requests for attributes of the target product. While any number of attributes about the target product can be requested, evaluation module 432 can request specific attributes that will be used to develop an actionable report on the target's competitiveness in digital marketplace 482. A first attribute can describe a rating 483 provided by at least one buyer of the target product on digital marketplace 482. Typically, digital marketplace 482 provides consumers with a graphical user interface (GUI) that allows those consumers to rate the products they purchase. In a particular implementation, consumers / buyers of the target product can use a 5-star rating system to rate the target product. A one-star rating would indicate that the consumer / buyer rated the target product poorly, while a five-star rating would indicate that the consumer / buyer rated the target product very well. Therefore, evaluation module 432 can use each star rating, or the average of those star ratings, as input to create an actionable report.

[0082] The second attribute may include content 486 of reviews and descriptions associated with the target product. Similarly, the digital marketplace 482 typically provides a GUI that allows consumers of the target product to input text describing their experience with the product. This text may include specific positive or negative keywords describing the consumer's experience with the target product. Using this data, the evaluation module 432 may enable the text analysis module 438, in some implementations, to parse each review describing the target product. Furthermore, the text analysis module 438 may also extract keywords describing certain features of the target product. As an example, the text analysis module 438 may extract the phrase "ergonomic handle," which not only describes that the target product includes a handle but also that the handle is an "ergonomic" handle, giving the consumer who gave the review a perception of how well the target product fits.

[0083] The third attribute could be the number of reviews (484) associated with the target product presented on the digital marketplace (482). The number of reviews (482) can indicate the level of engagement with the target product, whether it is negative or positive. Together with the text content of these reviews, the number of reviews associated with the target product can be used to help create actionable reports based on engagement with the target product within the digital marketplace (482).

[0084] The fourth attribute may include the listed price of the target product, 485. While the amount charged for purchasing the product may not indicate the value of the target product, the amount charged relative to other similar competing products may indicate its value or current price point (whether accurate or inaccurate).

[0085] The fifth attribute may also include the target product's ranking relative to at least one organic ranking competitor product 487. This ranking may be the result of the target product's average or cumulative rating relative to organic ranking competitors. Typically, the digital marketplace 382 allows buyers to list organic ranking competitors and the target product by average rating. By doing so, the evaluation module 432 can understand the target product's ranking relative to at least one organic ranking competitor product and use this information to develop actionable reports.

[0086] Each of these target product attributes can be requested by computing device 420 and its evaluation module 432, and delivered by server 452 upon request. Furthermore, similar attributes associated with at least one organically ranked competing product can also be requested and sent to computing device 420. These organically ranked product attributes may include competing product ratings 488, competing product review counts 489, competing product prices 490, competing product content 491, and competing product rankings 492. Each of these competing product attributes may be similar to those attributes associated with and described herein in conjunction with the target product.

[0087] Figure 5 This is a schematic block diagram illustrating a computing device 520 including a graphical user interface 522 for enabling the implementation of the present disclosure within a client / server architecture. The graphical user interface 522 can be used by a seller of a target product to evaluate the competitiveness of the target product as described herein. As described herein, the computing device 520 includes a filtering module 535. The filtering module 535 can be used to filter descriptive terms 541 to only those relevant descriptive terms 542 that would lead to the purchase of the target product in a digital marketplace.

[0088] Filtering module 535 may include various types of filters to filter descriptive terms 541 into relevant descriptive terms 542. These filters may include exposure filters 524, click-through rate filters 526, and conversion rate filters 528, each of which may result in the removal of those descriptive terms 541 that did not lead to a purchase of the target product or any organically ranked comparison product. As described herein, exposure filter 524 may be provided with the number of times that an ad (whether a banner, button, or text link) associated with the target product or competing products has been (or will be) exposed to potential buyers and has led to a purchase of that product. Therefore, exposure filter 524 may filter out instances where potential buyers did not see or were not shown the ad but did lead to a purchase. Click-through rate filter 526 may filter out those descriptive terms that, despite the wording of the ad, did not lead to an ad selection or product purchase. Conversion rate filter 528 may filter out those descriptive terms that, despite the wording of the ad and potential buyers' selection of the ad, did not lead to a product purchase.

[0089] By filtering descriptive terms via filter module 535 and its associated filters 524, 526, and 528, GUI 522 can display to the seller of the target product those relevant descriptive terms 542 suitable for analyzing the competitiveness of the target product. Although Figure 5 The use of specific filters 524, 526, and 528 to filter descriptive term 541 is shown, but this specification contemplates that any criteria may be used to filter descriptive term 541.

[0090] In some implementations, system 100 and / or any computing device or other device or system described herein may perform the methods described herein. However, the methods described herein are not intended to be limiting, and any type of software executing on a controller or processor may perform the methods described herein without departing from the scope of this disclosure.

[0091] Figure 6 This is a flowchart illustrating a method 600 for evaluating a product according to the principles of this disclosure. Method 600 may begin at block 605, utilizing an evaluation module executed by a processor to evaluate attributes of a target product. As described herein, the evaluation of a target product (or any other competing product) may indicate certain attributes of the target product. Although any number of attributes regarding the target product may be requested, the evaluation module may request specific attributes that will be used to develop an actionable report on the target's competitiveness in the digital market.

[0092] At box 610, method 600 may further include listing relevant descriptive terms for the target product that describe the attributes of the target product. This listing of relevant descriptive terms may also be performed by an evaluation module being executed by the processor of a computing device. In some embodiments, the list of relevant descriptive terms may be generated based on a filter of all descriptive terms generated for the target product as described herein. There may be some irrelevant information that can be filtered from the descriptive terms generated from the attributes of the target product, and this irrelevant information does not need to appear in the actionable report.

[0093] Method 600 may continue at block 615, accessing the computer network marketplace via NID and identifying at least one organic ranking competitor product that matches at least one descriptive term. This identification may enable an evaluation module to compare the descriptive terms associated with the target product with any generated descriptive terms associated with any organic ranking competitor product. In some implementations, this matching process of descriptive terms associated with the target product and descriptive terms associated with organic ranking competitor products may be performed before or after filtering the descriptive terms by a filtering module as described herein. When performed before filtering, more organic ranking competitor products may be matched, while when performed after filtering, relatively fewer organic ranking competitor products may be matched due to a smaller list of relevant descriptive terms.

[0094] Method 600 may further include, at block 620, comparing descriptive terms of the target product with descriptive terms associated with at least one organically ranked competing product to generate a competitiveness score. This may be performed via execution of a comparison module 620 executed by a processor. During the execution of the comparison module by the processor, descriptive terms may be compared to generate a competitiveness score using a competitiveness score generation module executed by the processor. In some embodiments, the competitiveness score may use any processing or algorithm used to define the degree to which the target product can or cannot compete with any discovered organically ranked competing products.

[0095] At box 625, method 600 may further include generating an actionable report describing the projected performance of the target product relative to at least one organically ranked competitor in the computer network marketplace. The actionable report may be generated via a recommendation module executed by a processor. During operation, the recommendation module may receive the competitiveness score, along with other data, from the server-hosted digital marketplace. This other data may include revenue data associated with the organically ranked competitor and the target product (if available). For example, conversion rate data, along with pricing data, may be passed to the recommendation module if a click-through rate for any given product (e.g., the target product or an organically ranked competitor) leads to a purchase. The recommendation module can then provide a recommendation describing whether the target product can (or cannot) compete with at least one organically ranked competitor. In some implementations, a threshold competitiveness score may be set such that the report provided by recommendation module 336 instructs the seller of the target product whether to continue selling the product in the digital marketplace. Alternatively, if the competitiveness score does not meet the threshold, the competitiveness score generation module may not forward the competitiveness score to the recommendation module to generate an actionable report. When the competitiveness score does not reach the threshold, the recommendation module only provides instructions to the seller, namely, it does not recommend that the seller start or continue selling the target product in the digital marketplace.

[0096] Upon reaching a threshold competitiveness score, the recommendation module can provide additional economic data describing the price point and ACoS statistics to increase revenue. Similarly, the seller of a target product may not know the appropriate target ACoS to meet or at what price point to sell the target product to see long-term gains rather than short-term profits. The recommendation module provides this information based on the competitiveness score generated by the competitiveness score generation module and revenue data received from the digital marketplace. In a specific example, the revenue potential of the target product can be determined by the recommendation module by calculating the ad spend profit margin, ad spend potential, and revenue potential. The ad spend profit margin can be calculated by multiplying the target ACoS by the price of the target product. The target ACoS can be determined and set by the seller based on available capital, or it can be set by the seller based on the share of revenue received from sales of the target product from the digital marketplace to date and manufacturing costs. The ad spend potential can then be calculated by multiplying the monthly opportunity units (OU) by the ad spend profit margin. The monthly OU can be calculated as the conversion rate of clicks to the target product resulting from sales after a buyer views the product. The revenue potential can then be calculated by multiplying the OU by the price of the target product. The revenue potential of each of the target product and its organically ranked competitors can be ranked to determine the target product's visibility in the digital marketplace.

[0097] At this point, method 600 can end.

[0098] Figure 7 This is a flowchart illustrating a method 700 for providing a competitive evaluation of a target product in the market, based on the principles of this disclosure. Here, method 700 may begin at block 705, evaluating the target product to determine its attributes. In some embodiments, this evaluation may be performed via the execution of an evaluation module. In some embodiments, the evaluation may be performed by requesting descriptive terms about the target product at a GUI. Additionally or alternatively, the evaluation may be performed by the evaluation module by accessing a digital marketplace to retrieve descriptive terms via a text analysis module as described herein. Additionally or alternatively, certain input devices, such as digital camera devices, may be used to image the target product and infer certain characteristics of the product, such as size, color, texture, etc.

[0099] Method 700 can continue at block 710, accessing the digital marketplace during processor execution to determine at least one organically ranked competing product for the target product. In this implementation, the evaluation module can access certain data about the target product, such as descriptive terms, and cross-reference those descriptive terms to determine whether at least one descriptive term matches any competing product listed on the digital marketplace.

[0100] At box 715, method 700 may include calculating a competitiveness score relating to the target product’s ability to compete with at least one organically ranked competitor. This process may be performed while a competitiveness score generator is executed by a processor of a computing device accessing a digital marketplace. In some implementations, the competitiveness score may use any processing or algorithm used to define the degree to which the target product can or cannot compete with any discovered organically ranked competitor.

[0101] Method 700 may further include generating an actionable report at box 720 based on the target product's ability to compete with at least one organically ranked competitor. During operation, a recommendation module executed by the processor may receive a competitiveness score and other data from the server-hosted digital marketplace. This other data may include revenue data associated with organically ranked competitors and the target product (if available). For example, if a click-through rate for any given product (e.g., the target product or an organically ranked competitor) leads to a purchase, this conversion rate data, along with the product's pricing data, may be passed to the recommendation module. The recommendation module can then provide a recommendation describing whether the target product can (or cannot) compete with at least one organically ranked competitor. In some implementations, a threshold competitiveness score may be set such that the report provided by the recommendation module instructs the seller of the target product whether to continue selling the product on the digital marketplace. Alternatively, if the competitiveness score does not meet the threshold, the competitiveness score generation module may not forward the competitiveness score to the recommendation module to generate an actionable report. When the threshold competitiveness score is not met, the recommendation module only instructs the seller not to start or continue selling the target product on the digital marketplace. At this point, method 700 may terminate.

[0102] Figure 8 This is a schematic block diagram illustrating computing device 822 and server 852 when operating digital marketplace 882, which can cooperate to implement the practices of this disclosure having a client / server architecture. In addition to providing, as in combination Figures 1 to 7 An actionable report describing the target's competitiveness in the digital market 882 ( Figure 3 In addition to (337), this computing device 822 may also describe actionable reports 837 on product-level sustainable and feasible growth over time on an e-commerce platform (e.g., digital marketplace 882), and provide a winability report 804 describing the probability of winning each search term (e.g., associating the target product with the search term) at any given point in time, as well as the estimated cost of winning those search terms. In some implementations, actionable reports 837 and winability reports 804 may provide users with instructions on how to optimize advertising and search engine implementation to increase revenue.

[0103] As described herein, computing device 822 may include processor 810, memory 820, user input 860, user output 870, and data storage 830, and its operation is similar to, for example, combining Figure 2A and Figure 2B The data storage 830 may include those modules described herein, including comparison module 834 and revenue module 899.

[0104] The described computing device 822 may include, as described herein, the computing device 822. Figure 3 The modules described herein, data storage 830, or any modules, data storage 830, or data maintained on a computer. In the embodiments described herein, similar combinations can be used. Figure 3 The comparison module 334 described is used to provide an actionable report 837. Although these modules (e.g., comparison module 834) can be similar to... Figure 3 The modules described in [the document], but Figure 8 The modules within can perform additional and different processes as described herein in order to provide actionable reports 837 that indicate optimized advertising and search engine implementations.

[0105] In some implementations, computing device 822 can initially identify any competing products that compete with the target product at any given time. Computing device 822 can perform this operation by accessing a search engine 894 associated with digital marketplace 882 via processor 810 and NID 880 of computing device 822. When accessing search engine 894, processor 810 can retrieve data describing the frequency of occurrence of one or more search terms associated with the target product. Additionally, processor 810 can obtain data related to the ranking of those search terms. This data can describe the co-occurrence of the target product and any competing products associated with the same search terms. Furthermore, this data can describe how similar the search terms associated with the target product and each competing product are in their rankings. For example, in the case of the target product being athletic shoes, some relevant search terms could include running, hiking, basketball, tennis, sole, shoelaces, and marathon, as well as other potential terms associated with the target product, athletic shoes. This data may also include which competing products also rank similarly with these terms. For example, a competing product that matches 9 out of 10 search terms with the target product ranks higher than a competing product that matches 4 out of 10 search terms.

[0106] In a particular embodiment, processor 810 may use a search query website, such as Google® Trends®, to access the data. Processor 810 may use these types of websites to access multiple search queries for a specific term associated with any one of the target product and any number of competing products. Processor 810 may access search query websites to automatically access search query queries in order to obtain the data used herein by computing device 822. While specific search query websites are contemplated herein, this specification also contemplates that processor 810 may access other search query databases, regardless of whether those databases are accessible to a user via a website.

[0107] The computing device 822 also includes a machine learning module 896 (e.g., which may be referred to herein as machine learning model 896). The machine learning module 896 can construct multiple mathematical models that provide a competition set report 898 describing the competition set of products competing with the target product. Like each machine learning module 896, the machine learning module 896 can be "taught" using multiple sets of target product search terms and rankings, as well as multiple sets of competing product search terms and rankings, as input. Similarly, the multiple sets of target product search terms and rankings, as well as the multiple sets of competing product search terms and rankings, can be accessed by the processor 810 via a specific search query website or database.

[0108] In some embodiments, when executed by processor 810, machine learning module 896 may determine such relevance based on any machine learning or neural network methods known in the art or developed in the future. In certain embodiments, machine learning module 896 may implement unsupervised learning clustering techniques. For example, in some embodiments, the machine learning module may use a hierarchical neural network topology to model the relationship between multiple sets of target product search terms and rankings and each of multiple sets of competing product search terms and rankings. In some embodiments, such a neural network may include: an input layer (e.g., multiple sets of target product search terms and rankings and multiple sets of competing product search terms and rankings) comprising a known, recorded set of values ​​for each of these parameters, settings, metrics, and data metrics used; and an output layer based on the known, recorded set of values ​​in the input layer, including the predicted optimal competitive set report 898. In some embodiments, machine learning module 896 may propagate the input through the layers of the neural network to predict or forecast the optimal competitive set report 898 based on known and recorded search term metrics, and compare these predicted values ​​with the optimal search terms to be presented in the competitive set report 898. In some implementations, using a backpropagation method, the machine learning module 896 can then adjust the weight matrix of the neural network using the difference between the predicted value and the known optimal value. The weight matrix describes how changes in each of the search term data metrics may affect the optimal search term to be presented in the competitive set report 898.

[0109] The computing device 822 can provide learned competing search terms through its output layer. If any of these search terms has been specified, and based on similar and frequent search terms detected at the search engine 894 of the digital marketplace 882 during use of the computing device 822, these learned competing search terms are determined to be optimal search terms. These obtained optimal search terms can be suggested to the user or implemented automatically. The suggestions may include indicators and may be graphically displayed at a user interface (e.g., a graphical user interface) for user approval in some implementations prior to the implementation of other processes executed by the processor 810 of the computing device 822.

[0110] Figure 9 The image shows an example representation of a graph. This example graph can indicate the position of each search term for competing products relative to the target product based on frequency. Figure 9 Each point (e.g., a circle) on the graph represents a search term. Each representative search term is... Figure 9The graph is arranged at points defining the frequency with which a term appears alongside search terms related to the target product, and at positions indicating whether a search term is similar or dissimilar to search terms associated with the target product. In this example graph, the further right any given search term is, the more similar the competing product's search terms are to the target product's. Additionally, the further left any given search term is, the less similar the competing product's search terms are to the target product's. Furthermore, the closer any given search term is to the top of the graph, the more generic it is compared to the target product, while the closer any given search term is to the bottom of the graph, the less niche it is compared to the target product. In some implementations, it may be most desirable to have a target product with associated search terms that are more generic and similar to those of competing products. This indicates that the target product is competing with relatively well-known competing products. The process described herein can help provide users with reports indicating how to adjust advertising revenue to focus on search terms that are more generic and similar to those of competing products.

[0111] In some implementations, the machine learning module 896 can perform forward and backward propagation, repeatedly using different input node values ​​to fine-tune any weighted or unweighted matrix. In this way, in some implementations, the machine learning module 896 can adaptively learn how changes in multiple sets of target product search terms and rankings, as well as multiple sets of competing product search terms and rankings, might affect the data reflected in the competition set report 898. In such implementations, the weight matrices associated with the layers of the neural network model can mathematically describe these relevances for a single target product. In some implementations, the neural network model (including the specification of node values ​​in the input layers and the number of layers) along with the weight matrices associated with each layer can form a trained machine learning classifier, algorithm, or mathematical model for generating any competition set report 898 as described herein.

[0112] As described herein, the output from the now-trained machine learning module 896 is a competition set report 898. Using the competition set report 898, computing device 822 can utilize processor 810 and NID 880 to determine the current performance of search terms most relevant to the target product and the competition set defined in the competition set report 898. Two variables identified in this process are the frequency with which the term generally appears in searches (e.g., general search term search volume, or how many times people search for the term per day) and the frequency with which the term appears in searches associated with the competition set report 898. More specifically, in some embodiments, those search terms found to be the most general and similar to the target product and each competing product are provided to comparison module 834, which searches for those search terms defined in the competition set report 898 via execution at search engine 894 by processor 810. During this process, processor 810 can access search engine 894 or any other search engine at digital marketplace 882 and obtain search term metadata describing the current performance of each of the search terms most relevant to the target product and the competition set defined in the competition set report 898. The comparison module 834 can compare these most relevant search terms from the competition set report 898 and provide this data to the user in the form of an actionable report 837. In some implementations, data describing the most relevant search terms in the actionable report 837 related to the target product and associated with the competition set can be provided to the user in a graphical representation.

[0113] Figure 10 The image shows an example graphical representation of the current performance for search terms related to the target product. For example... Figure 10 As shown, the further right any search term (e.g., represented by a circle) is on the graph, the higher its search volume or the more frequently it appears compared to other search terms, indicating a relatively higher relevance to competing products. Additionally, the further left any search term is on the graph, the lower its search volume or the less frequently it appears compared to other search terms, indicating a relatively lower relevance to competing products. Furthermore, the closer any search term is to the top of the graph, the higher its relevance compared to other search terms, indicating a relatively higher relevance to competing products. Moreover, the lower a search term is placed on the graph, the lower its relevance compared to other search terms, indicating a relatively lower relevance to competing products. The most frequent and most relevant terms can also be provided to the comparison module 834 and used to further define the sustainability and viable growth of the target product over time, for example, in the digital marketplace 882.

[0114] In such Figure 10If the most relevant and frequent search terms indicated are found and presented in the actionable report 837, the computing device 822 can also quantify the opportunity of those search terms to increase revenue and profit margin of the target product when associated with it. In some implementations, the processor 810 can execute the revenue module 899 to receive those relevant and most frequent search terms from the actionable report 837 and provide output to the user in the form of an increased revenue metric. The increased revenue can be calculated by the revenue module when the processor 810 executes using the following formula:

[0115] Increased revenue = Exposure Click-through rate Conversion rate Shopping basket size price

[0116] Equation 1

[0117] In the context of Equation 1, in some implementations, exposure can be defined as the search volume for each of the most relevant and most frequent search terms. In some implementations, the amount of exposure can be measured by the number of times an advertisement associated with the target product is presented to any given user during or after the most relevant and most frequent search terms are entered into search engine 894. This data can be retrieved by processor 810 by accessing a specific database or a search query website as described herein.

[0118] In some implementations, the click-through rate in Equation 1 can be defined as an estimate of the probability of receiving a click along the curve associated with its ranking for each of the most relevant and frequent search terms provided by Actionable Report 837. For example, the ranking can be set to include the click-through rate of the first term (e.g., 20). (14 clicks), second place click rate (14) (clicks), until the tenth place click rate (6) (Clicks), and more than any number of the most relevant and frequent search terms ranked. This data can be retrieved by processor 810 by accessing specific databases or search query websites as described in this article.

[0119] In some implementations, the conversion rate in Equation 1 can be defined as the percentage of the most relevant and frequent search terms that are clicked, associated with the target product, and converted into sales (e.g., resulting in a sale of the target product). This data can be retrieved by the processor 810 by accessing a specific database or a search query website as described herein.

[0120] In some implementations, the shopping basket size can be defined as the number of units purchased per conversion. In some implementations, this number can be averaged over multiple purchases. For example, if multiple conversions have been detected, processor 810 can calculate how many units of the target product were purchased in any given transaction (e.g., units placed in the "cart" for purchase at digital marketplace 882). This value can be at least equal to 1 or more. Similarly, this data can be retrieved by processor 810 by accessing a specific database or a search query website as described herein.

[0121] In some implementations, the price of the target product may be a manufacturer-suggested retail price. In some implementations, the quantitative value of the price in Equation 1 is the net price of the target product or other branded products involved in the purchase across any plurality of digital marketplaces 882, less any discounts or promotions associated with those sales. This data may be retrieved by processor 810 by accessing a specific database, accessing a search query website as described herein, or accessing sales data from a database maintained by the manufacturer of the target product.

[0122] In some implementations, any one of the exposure, click-through rate, conversion rate, basket size, and price values ​​in Equation 1 can be enhanced by a weight value. In this implementation, the weight value can either enhance or diminish the impact of any of these values ​​in Equation 1 to better determine the increased revenue value or opportunity for the seller of the target product to increase that revenue. Because actual real-time data is used in Equation 1, the seller of the target product or the user of computing device 822 can know in real time whether to leverage any instances of increased page views or sales to increase interest in the target product compared to any competitor's product.

[0123] In some implementations, the value associated with the click-through rate (CTR) in Equation 1 can significantly alter a user's decision regarding whether to take actions such as providing more advertising support for the target product. In some implementations, the CTR associated with improving the target product's search ranking from its current position on the search term to its potential ranking position on the search phrase can be weighted to accommodate an increase in the value's importance. For example, for a given search term that could improve the organic search ranking of any search term from 20th to 5th, the CTR would increase by an estimated 3-fold. Some ranking improvements may also stem from increased exposure, particularly in cases where a target product not ranked on the search term reaches a search ranking of 10th. In this example, this would increase clicks from zero (due to zero exposure) to an estimated 10th place on the search term. As output, processor 810 can provide an increased revenue report 802 via revenue module 899, describing how, if any, revenue related to the sale of the target product will increase.

[0124] In some cases, certain search terms may not be applicable to the target product, but could potentially increase revenue if applicable. In the context of advertising, these currently unapplicable search terms may be referred to as "unavailable." These unavailable search terms may be those that are at least initially irrelevant to the target product for some reason, or those that have not yet been associated with the target product due to a lack of data associated with the target product in the platform data associated with the digital marketplace 882. In some implementations, the machine learning module 896 may also be trained and used to receive data related to the characteristics of the target product, the target product's current competitors, and the current state of the e-commerce search term algorithm to determine the "winnability" of search terms. The winnability of search terms can be defined as the probability of winning each search term (e.g., associating the target product with the search term) at any given point in time, and the estimated cost of winning those search terms.

[0125] Machine learning module 896 can be trained using winability inputs as described herein to provide winability report 804. Some inputs to the model include any number of inputs, and the description of certain types of inputs is not intended to limit the breadth of inputs fed into machine learning module 896 to obtain a winability report, and this specification envisions these additional and different inputs. As an example, inputs may include the current and historical prices of both the target product and competing products. This historical pricing can be retrieved from one or more digital marketplaces 882 via the execution of processor 810 and NID 880 as described herein. In this specific example, processor 810 may enable NID 880 to access one or more digital marketplaces 882 via a wired (wired transmitter / receiver 840) connection or a wireless (wireless transmitter / receiver 850) connection, find instances where the target product and competing products are sold, and retrieve their historical pricing values.

[0126] Another input to the machine learning module 896 may include current and historical review ratings and review counts associated with the target and competing products. This review rating and review count data can be retrieved from one or more digital marketplaces 882 via the execution of processor 810 and NID 880 as described herein. Digital marketplaces 882 typically provide a GUI that allows consumers of the target and competing products to input text describing their experience with the target and competing products, as well as ranking assessments of those products in the form of a digital rating system or a star rating system. In this specific example, processor 810 may enable NID 880 to access one or more digital marketplaces 882 via a wired or wireless connection, find review ratings and review counts associated with the sale of the target and competing products, and provide this review rating and review count data to the machine learning module 896.

[0127] Another input to the machine learning module 896 can include content similarity scores for any search terms related to the target product and competing products. In some implementations, these scores can be generated based on the provided data. For example, in Figure 9 In, any given search term in Figure 9 The further to the right on the graph, the more similar the search terms for competing products are to the search terms for the target product. In a specific example, Figure 9 The x-axis (bottom) of the graph or its associated data can be used to assign the similarity score. For example... Figure 9 As shown, according to Figure 9 The similarity score can be a positive or negative score, representing a numerical ranking on the x-axis. In this example, the similarity score could be a positive or negative weight reflected in the winability report 804 provided by the processor 810 when the machine learning module 896 is executed. In this implementation, the processor 810 can again enable the NID 880 to access one or more digital marketplaces 882 via a wired or wireless connection and retrieve multiple sets of target product search terms and rankings, as well as multiple sets of competitor product search terms and rankings, associated with the sale of the target product and competing products. This data is then provided to the machine learning module 896.

[0128] Furthermore, other inputs to the machine learning module 896 may include platform-specific information, such as the average best-selling rank (BSR) of any given digital marketplace 882 associated with the target product and any number of competing products. The BSR may differ across any given digital marketplace 882, but these rankings can be averaged across multiple digital marketplaces 882 to obtain the value. In this implementation, the processor 810 may again enable the NID 880 to access one or more digital marketplaces 882 via a wired or wireless connection and retrieve the BSR data. This data is then provided to the machine learning module 896.

[0129] Other inputs to the machine learning module 896 may include projected search term volume and click distribution. Regarding this type of data provided to the machine learning module 896, the projected search term volume can be obtained from the data used to create... Figure 10 The data is retrieved from the graph. This data, describing the frequency of any given search term appearing in the search results associated with the target product and competing products, can be accessed by processor 810 and provided as input to machine learning module 896. Additionally, any click distribution describing how many clicks any given search term receives can be accessed by processor 810 and NID 880 at search engine 894 in digital marketplace 882.

[0130] Other inputs to the machine learning module 896 may include historical changes in the ranking of search terms related to the target product and the search phrase product. At any given time, the content searched by the search engine 894 on the internet may fluctuate. These search terms can be ranked, and their historical rankings may change over time based on numerous social, political, environmental, and economic factors. This historical data can be retrieved from the search engine 894 by the processor 810 and NID 880 and provided to the machine learning module 896.

[0131] Another example input to the machine learning module 896 may include targeted advertising spending associated with search terms related to the target product. This data can be maintained on any database accessible to the processor 810 of the computing device 822. In a particular embodiment, this data describing targeted advertising spending associated with search terms related to the target product may be maintained by the seller of the target product on a private database, and the user of the computing device 822 may be granted secure access to that database. This type of data may also be provided to the machine learning module 896.

[0132] With all these different types of data available to the processor 810 via the NID 880, the machine learning module 896 can construct multiple mathematical models that provide a winability report 804 describing the probability of winning each search term (e.g., associating the target product with the search term) at any given time and the estimated cost of winning those search terms. Like each machine learning module 896, the machine learning module 896 can be “taught” using the winability factors described herein. In some embodiments, the machine learning module 896 can implement one or more non-parametric and parametric learning techniques. For example, in some embodiments, the machine learning module can model the relationships between each of multiple sets of winability factors using a hierarchical neural network topology. In some embodiments, such a neural network may include: an input layer (e.g., winability factors) comprising a known, recorded set of values ​​for each of these parameters, settings, metrics, and data metrics used; and an output layer based on the known, recorded set of values ​​in the input layer, including the predicted winability report 804. In some implementations, the machine learning module 896 may propagate input through the layers of a neural network to predict or forecast the optimal winability of a search term based on known and recorded search term metrics, and compare these predictions with the optimal search term to be presented in the winability report 804. In some implementations, using a backpropagation approach, the machine learning module 896 may then use the difference between the predicted values ​​and the known optimal values ​​to adjust the weight matrix of the neural network, which describes how changes in each of the search term data metrics might affect the optimal search term to be presented in the winability report 804.

[0133] The computing device 822 can provide learned competing search terms through its output layer. If any of these search terms has been specified, and based on winning search terms detected at the search engine 894 of the digital marketplace 882 or other databases during the use of the computing device 822, these learned competing search terms are determined to be optimal search terms. These obtained optimal search terms can be suggested to the user or implemented automatically. The suggestions may include indicators and may be graphically displayed at the user interface for user approval in some implementations before implementing other processes executed by the processor 810 of the computing device 822.

[0134] In some implementations, the machine learning module 896 can perform forward and backward propagation, repeatedly using different input node values ​​to fine-tune any weighted or unweighted matrix. In this way, in some implementations, the machine learning module 896 can adaptively learn how changes in the winability factor might affect the data reflected in the winability report 804. In such implementations, the weight matrices associated with the layers of the neural network model can mathematically describe these correlations for a single target product. In some implementations, the neural network model (including the specification of node values ​​in the input layers and the number of layers) along with the weight matrices associated with each layer can form a trained machine learning classifier, algorithm, or mathematical model for generating any winability report 804 as described herein.

[0135] As described herein, the output from the now-trained machine learning module 896 is a winability report 804. Using the winability report 804, computing device 822 can utilize processor 810 and NID 880 to determine the probability of achieving the expected change in revenue based on the required investment. In some implementations, the required investment can be calculated using the following equation:

[0136] Required investment = Expected bid (exposure Click pass rate

[0137] Equation 2

[0138] The return on investment (ROI) can then be calculated using the following equation:

[0139] ROI = Increased Revenue (Stay within the expected timeframe for the required investment)

[0140] Equation 3

[0141] Using equations 2 and 3, target products for search terms with high ROI can then be prioritized by users for both advertising and search engine optimization (SEO) actions. In this way, computing device 822 can execute machine learning module 896 for the secondary purpose of determining the "winnability" of search terms, where additional funding is applied to both advertising and SEO.

[0142] In some implementations, the calculations of ad spend profit margin, ad spend potential, and revenue potential performed by processor 810 can also be performed to specifically determine how much additional advertising funding to apply to the target product. Similarly, the ad spend profit margin can be calculated by multiplying the target ACoS by the price of the target product. The target ACoS can be determined and set by the seller based on available capital, or it can be set by the seller based on the share of revenue received to date from sales of the target product from digital marketplace 382 and manufacturing costs. The ad spend potential can then be calculated by multiplying the monthly opportunity unit (OU) by the ad spend profit margin. The monthly OU can be calculated as the conversion rate of clicks to the target product resulting from sales after a buyer views the product. The revenue potential can then be calculated by multiplying the OU by the price of the target product. The revenue potential of each of the target products can be ranked to determine the target product's placement within digital marketplace 882. The search terms presented in the winability report 804 can be sorted by revenue potential to determine the best opportunity for the target product to generate revenue growth. To refine the recommendations, the process can continue by inputting estimated bid amounts from digital marketplace 882 for winning ad placements for these keywords. In this way, the execution of processor 810 can initiate these calculations to predict the number of clicks and costs necessary to achieve potential growth. The equation for this calculation is shown in Equation 2 in this document.

[0143] ROI can also be calculated using the following equation:

[0144] ROI = Advertising Spending Potential (Investment payback period - required investment)

[0145] Equation 4

[0146] Because high-winnability terms are targeted through both advertising and search engine optimization (SEO) techniques, increasing associated exposure, clicks, and conversions, the processing applied to the target product can be continuously adjusted. As the target product succeeds on new search terms, the set of competing products as defined in Competitive Set Report 898 shifts to compare against larger and less niche competing products. As the set of competing products as defined in Competitive Set Report 898 shifts, the set of competing terms also shifts. With changes in reviews, terms, seller rankings, and other attributes, the winnability of each term and the associated required investment also change. As winnability shifts, new terms are prioritized, and this cycle continues iteratively to proportionally increase revenue associated with the target product.

[0147] In some implementations, computing device 882 (e.g., and / or system 100, computing device 420, computing device 520, and / or computing device 322) can be configured to generate an optimal cost-per-click real-time bidding system that balances target spending and efficiency. For example, computing device 882 can lower bids for bidding campaigns that would otherwise exhaust the budget or exceed expected spending. Additionally or alternatively, computing device 882 can increase bids for advertising campaigns that are underspending but still have room given efficiency targets.

[0148] The computing device 882 can adjust bids to achieve a return on advertising expenditure over a period of time. Campaigns that exhaust their budget or exceed their spending targets before the end of a time period may fail to realize all potential revenue due to spending more than necessary in an undesirably short period. Alternatively, campaigns that do not spend sufficient budget sacrifice sales volume. Therefore, the computing device 882 can provide automated calculations that adjust bids to maximize efficiency while maintaining target spending. The computing device 882 can use one or more key models and / or data processes to automatically adjust bids. These one or more key models and / or data processes may include and / or utilize efficiency targets, estimated conversion rates, estimated order values, diminishing returns on incremental spending, and / or other suitable rates, values, data, etc.

[0149] In some implementations, computing device 882 uses one or more of the following: minimum efficiency and strategic requirements provided by the user; and expected expenditure within the strategic requirements and minimum efficiency level provided by the user. Computing device 882 may generate one or more models for various objectives, including: a model for estimating conversion rates; a model for estimating order value; a model for estimating cost per click; and / or one or more other suitable models.

[0150] Computing device 882 can use the output from one or more models to generate another model to predict the trade-off between bid, cost per click, and revenue, thereby generating incremental revenue and incremental cost trade-offs. Computing device 882 can use data streams to ingest data and leverage these models to predict target bids in near real-time. Computing device 882 can receive user preferences, constraints, and spending targets, and can optimally adjust bids based on model outputs to achieve desired spending, making the desired spending achievable given constraints.

[0151] In some embodiments, the computing device 882 may receive user input data. In some embodiments, the user input data includes the expected total bid value and bidding period. In some embodiments, the user input data includes information associated with at least one product. In some embodiments, the user input includes information associated with at least one market.

[0152] The computing device 882 can generate a first machine learning model that produces predicted expected performance based on user input data. The first machine learning model may include a tree-based machine learning model, a Bayesian machine learning model, or other suitable machine learning model.

[0153] The computing device 882 can adjust at least one bid for at least one of at least one keyword and at least one product associated with at least one market based on predicted expected performance.

[0154] In some implementations, computing device 882 may generate a second machine learning model that produces predicted expected costs based on user input. The second machine learning model may include a tree-based machine learning model, a Bayesian machine learning model, or other suitable machine learning model. In some implementations, computing device 882 may further adjust at least one bid for at least one of at least one keyword and at least one product associated with at least one market based on the predicted expected costs.

[0155] Figure 12 This is a flowchart illustrating a real-time bidding method 1200 based on the principles of this disclosure. At 1202, method 1200 receives user input data.

[0156] At 1204, method 1200 generates a first machine learning model that generates the expected performance of the prediction based on user input data.

[0157] At 1206, method 1200 adjusts at least one bid for at least one of at least one keyword and at least one product associated with at least one market, based on the predicted expected performance.

[0158] In some implementations, a real-time bidding system includes a processor and a memory. The memory includes instructions that, when executed by the processor, cause the processor to: receive user input data; generate a first machine learning model based on the user input data to generate predicted expected performance; and adjust at least one bid for at least one of at least one keyword and at least one product associated with at least one market based on the predicted expected performance.

[0159] In some implementations, the user input data includes the expected total bid value and bidding period. In some implementations, the user input data includes information associated with at least one product. In some implementations, the user input includes information associated with at least one market. In some implementations, the first machine learning model includes a tree-based machine learning model. In some implementations, the first machine learning model includes a Bayesian machine learning model. In some implementations, the instructions further cause the processor to generate a second machine learning model that generates a predicted expected cost based on the user input. In some implementations, the second machine learning model includes a tree-based machine learning model. In some implementations, the second machine learning model includes a Bayesian machine learning model. In some implementations, the instructions further cause the processor to further adjust the at least one bid for at least one of the at least one keyword and at least one product associated with the at least one market based on the predicted expected cost.

[0160] In some implementations, a real-time bidding method includes: receiving user input data; generating a first machine learning model, the first machine learning model generating predicted expected performance based on the user input data; and adjusting at least one bid for at least one of at least one keyword and at least one product associated with at least one market, based on the predicted expected performance.

[0161] In some implementations, the user input data includes the expected total bid value and bidding period. In some implementations, the user input data includes information associated with at least one product. In some implementations, the user input includes information associated with at least one market. In some implementations, the first machine learning model includes a tree-based machine learning model. In some implementations, the first machine learning model includes a Bayesian machine learning model. In some implementations, the method further includes generating a second machine learning model that generates a predicted expected cost based on the user input. In some implementations, the second machine learning model includes a tree-based machine learning model. In some implementations, the second machine learning model includes a Bayesian machine learning model. In some implementations, the method further includes adjusting the at least one bid of at least one of the at least one keyword and at least one product associated with the at least one market based on the predicted expected cost.

[0162] Any method disclosed herein includes one or more steps or actions for performing the described method. Method steps and / or actions may be interchanged with each other. In other words, the order and / or use of specific steps and / or actions may be modified unless proper operation of the embodiment requires a specific order of steps or actions.

[0163] The use of the term "first" in a claim relating to a feature or element does not necessarily imply the existence of a second or additional such feature or element. Elements described in a means-plus-function format are intended to comply with 35 USC. This will be explained in paragraph 6 of section 112. It will be apparent to those skilled in the art that changes can be made to the details of the above embodiments without departing from the basic principles of this disclosure.

[0164] The foregoing discussion is intended to illustrate the principles and various implementations of this disclosure. Once the foregoing disclosure is fully understood, many variations and modifications will become apparent to those skilled in the art. The appended claims are intended to be construed as covering all such variations and modifications.

[0165] The word “example” as used herein means intended as an example, instance, or illustration. Any aspect or design described herein as an “example” is not necessarily to be construed as preferred or advantageous relative to other aspects or designs. Rather, the use of the word “example” is intended to present concepts in a concrete manner. As used herein, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or.” That is, unless otherwise specified or clearly apparent from the context, “X comprises A or B” is intended to mean any natural inclusion. That is, if X comprises A; X comprises B; or X comprises both A and B, then “X comprises A or B” is satisfied in any of the foregoing cases. Additionally, unless otherwise specified or clearly apparent from the context to refer to the singular form, the articles “a (a)” and “an (an)” as used herein and in the appended claims should generally be construed as meaning “one or more.” Moreover, unless described as such, the use of the terms “an implementation” or “an implementation” throughout is not intended to refer to the same embodiment or implementation.

[0166] The systems, algorithms, methods, instructions, etc., described herein can be implemented in hardware, software, or any combination thereof. Hardware may include, for example, a computer, intellectual property (IP) core, application-specific integrated circuit (ASIC), programmable logic array, optical processor, programmable logic controller, microcode, microcontroller, server, microprocessor, digital signal processor, or any other suitable circuit. In the claims, the term "processor" should be understood to cover any of the foregoing hardware, whether alone or in combination. The terms "signal" and "data" are used interchangeably.

[0167] As used herein, the term "module" can include a packaged functional hardware unit designed for use with other components, a set of instructions executable by a controller (e.g., a processor executing software or firmware), processing circuitry configured to perform a specific function, and self-contained hardware or software components that interface with a larger system. For example, a module can include application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), circuits, digital logic circuits, analog circuits, discrete circuits, gates, and other types of hardware or combinations thereof. In other embodiments, a module can include memory storing instructions executable by a controller to implement the features of the module.

[0168] Furthermore, in one aspect, for example, the systems described herein can be implemented using a general-purpose computer or general-purpose processor and a computer program that, when executed, performs any of the corresponding methods, algorithms, and / or instructions described herein. Alternatively or alternatively, for example, a special-purpose computer / processor can be utilized, which may include additional hardware for performing any of the methods, algorithms, or instructions described herein.

[0169] Furthermore, all or part of the implementation of this disclosure may take the form of a computer program product accessible from, for example, a computer-usable or computer-readable medium. A computer-usable or computer-readable medium can be any device capable of, for example, tangibly containing, storing, communicating, or transmitting a program for use by or in connection with any processor. Such a medium can be, for example, an electronic, magnetic, optical, electromagnetic, or semiconductor device. Other suitable media are also available.

[0170] The above-described embodiments, implementations, and aspects are for the purpose of facilitating understanding of this disclosure and do not limit the scope of this disclosure. Rather, this disclosure is intended to cover various modifications and equivalent arrangements included within the scope of the appended claims, which should be interpreted in the broadest possible sense to encompass all such modifications and equivalent structures permitted by law.

Claims

1. A real-time bidding system, comprising: processor; as well as The memory includes instructions that, when executed by the processor, cause the processor to: Receive user input data; Generate a first machine learning model, which generates the expected performance of predictions based on the user input data; and Based on the expected performance of the forecast, adjust at least one bid for at least one of at least one product and at least one keyword associated with at least one market.

2. The system according to claim 1, wherein, The user input data includes the expected total bid value and the bid period.

3. The system according to claim 1, wherein, The user input data includes information associated with the at least one product.

4. The system according to claim 1, wherein, The user input includes information associated with the at least one market.

5. The system according to claim 1, wherein, The first machine learning model includes a tree-based machine learning model.

6. The system according to claim 1, wherein, The first machine learning model includes a Bayesian machine learning model.

7. The system according to claim 1, wherein, The instructions also cause the processor to generate a second machine learning model, which generates a predicted cost based on the user input.

8. The system according to claim 7, wherein, The second machine learning model includes a tree-based machine learning model.

9. The system according to claim 7, wherein, The second machine learning model includes a Bayesian machine learning model.

10. The system according to claim 7, wherein, The instructions also cause the processor to further adjust the bid of at least one of the at least one product and at least one keyword associated with the at least one market, based on the predicted expected cost.

11. A real-time bidding method, comprising: Receive user input data; Generate a first machine learning model, which generates the expected performance of the prediction based on the user input data; as well as Based on the expected performance of the forecast, adjust at least one bid for at least one of at least one product and at least one keyword associated with at least one market.

12. The method according to claim 11, wherein, The user input data includes the expected total bid value and the bid period.

13. The method according to claim 11, wherein, The user input data includes information associated with the at least one product.

14. The method according to claim 11, wherein, The user input includes information associated with the at least one market.

15. The method according to claim 11, wherein, The first machine learning model includes a tree-based machine learning model.

16. The method according to claim 11, wherein, The first machine learning model includes a Bayesian machine learning model.

17. The method of claim 11, further comprising generating a second machine learning model, the second machine learning model generating the expected cost of the prediction based on the user input.

18. The method according to claim 17, wherein, The second machine learning model includes a tree-based machine learning model.

19. The method of claim 17, wherein, The second machine learning model includes a Bayesian machine learning model.

20. The method of claim 17, further comprising adjusting the at least one bid of the at least one product and at least one keyword associated with the at least one market based on the predicted expected cost.

Images