Systems and methods for targeted item matching

Abstract

A portion of a plurality of items stored in an item data store are determined for targeted item matching. For a targeted item in the portion, a plurality of matching items to the targeted item are identified for inclusion in an item match set for the targeted item, and a plurality of scores are determined for the item match set. The plurality of scores include a score for each of the targeted item and the plurality of matching items based on a plurality of features for the respective item. A data structure associated with the item match set, and including the plurality of scores, is generated and stored for use in one or more subsequent processes, including item search and event-based item replacement processes.

Description

Attorney Docket No: 00273-0015-00304SYSTEMS AND METHODS FOR TARGETED ITEM MATCHINGCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of priority from U.S. Provisional Application No. 63 / 763,665, filed February 26, 2025, which is incorporated by reference herein in its entirety.TECHNICAL FIELD

[0002] The present disclosure relates generally to systems and methods for targeted item matching, and more specifically systems and methods for targeted item matching implemented to facilitate item discovery in an item search context for content generation and event-based item replacement within generated content.BACKGROUND

[0003] A content creation-based ecommerce platform may serve as a marketplace for merchants to engage content creators to help promote their items in exchange for a commission via content generated and published by the content creators. Within the marketplace, there may be instances of exact item matches (e.g., the same item and / or different variants of the same item sold by different merchants). In this context, an item search system and / or sub-platform of the ecommerce platform may be implemented to enable content creators to discover an item to include in promotional content. For example, the item search system and / or sub-platform may identify and rank items that are responsive to users’ queries for presentation as results. However, conventional ranking algorithms primarily rely on ranking factors or features such as keyword matching, sales volumes, or popularity. Utilization of the factors or features of conventional ranking algorithms favors identification of items of larger and / or more well-known merchants (e.g., despite other available matching items), resulting in a decreased diversity of items presented without consideration of earning opportunities for the content creators.

[0004] Additionally, item-based content generated and published by a content creator (e.g., based on discovery of the item via item search) may persist for consumer consumption throughout time. However, factors or features affecting the earning opportunity associated with the item may change overtime, which may impact a viability of the item-based content (e.g., render the item-based content lessAttorney Docket No: 00273-0015-00304valuable and / or invaluable to the content creator), if such earning opportunity decreases. For example, if several weeks after publication, the item is no longer available for purchase (e.g., is out of stock or sold out), the content can no longer be monetized.

[0005] The background description provided herein is for the purpose of generally presenting the context of the disclosure. Unless otherwise indicated herein, the materials described in this section are not prior art to the claims in this application and are not admitted to be prior art, or suggestions of the prior art, by inclusion in this section.SUMMARY

[0006] The techniques of this disclosure for targeted item matching improve the state of item discovery and item-based content viability (e.g., via event-based item replacement) in the context of a content creation-based ecommerce platform.

[0007] The techniques described herein relate to methods, systems, and / or non-transitory computer readable mediums for targeted item matching. For example, a portion of a plurality of items stored in an item data store may be determined for targeted item matching. For a targeted item in the portion, a plurality of matching items to the targeted item may be identified for inclusion in an item match set for the targeted item. A plurality of scores may be determined for the item match set, including a score for each of the targeted item and the plurality of matching items based on a plurality of features for the respective item. A data structure associated with the item match set, and including the plurality of scores, may be generated and stored. An event associated with the targeted item may be detected. Based on the detected event, content including the targeted item may be identified, a recommended matching item from the item match set for the targeted item may be determined based on the plurality of scores using the data structure, and a resource for the targeted item associated with the content may be replaced with a new resource for the recommended matching item.BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate various example embodiments and together with the description, serve to explain the principles of the disclosed embodiments.Attorney Docket No: 00273-0015-00304

[0009] FIG. 1 is a diagram showing an example of a marketplace environment, according to some aspects of the disclosure.

[0010] FIG. 2 is a flow chart showing an example process for determining conversion-based features, according to some aspects of the disclosure.

[0011] FIG. 3 is a flow chart showing an example process for targeted item matching, according to some aspects of the disclosure.

[0012] FIG. 4 is a flow chart showing an example process for determining, as part of the targeted item matching process, a feature-based score for a respective item of an item match set, according to some aspects of the disclosure.

[0013] FIG. 5 is a flow chart showing an example process for determining, as part of the feature-based score determination process, a conversion-based feature value for the respective item of the item match set, according to some aspects of the disclosure.

[0014] FIG. 6 is a flow chart showing an example method for processing a query for an item search that implements the targeted item matching, according to some embodiments of the disclosure.

[0015] FIG. 7 is a flow chart showing an example process for event-based item replacement that implements the targeted item matching, according to some embodiments of the disclosure.

[0016] FIG. 8 shows an implementation of a computer system that executes techniques presented herein, according to some embodiments of the disclosure.DETAILED DESCRIPTION

[0017] As briefly mentioned above, a content creation-based ecommerce platform may serve as a marketplace for merchants to engage content creators to help promote their items in exchange for a commission via content generated and published by the content creators. Within the marketplace, there may be many instances of exact item matches (e.g., the same item and / or different variants of the same item sold by different merchants). In this context, an item search system and / or sub-platform of the ecommerce platform may be implemented to enable content creators to discover an item to include in promotional content. For example, the item search system and / or sub-platform may identify and rank items that are responsive to users’ queries for presentation as results. However, conventional ranking algorithms primarily rely on ranking factors or features such as keywordAttorney Docket No: 00273-0015-00304matching, sales volumes, or popularity. Utilization of the factors or features of conventional ranking algorithms favors identification of items of larger and / or more well-known merchants (e.g., despite other available matching items), resulting in a decreased diversity of items presented without consideration of earning opportunities for the content creators.

[0018] For example, because larger and / or well-known merchants typically offer a larger number and wider range of items for sale, there is an increased likelihood of items from these larger and / or well-known merchants matching the content creator’s search query. This match leads to a higher exposure of items from larger and / or well-known merchants within search results, and a corresponding increased number of item sales from larger and / or well-known merchants due to the exposure and / or popularity that may elevate a rank of these items within the search results.Additionally, conventional ranking algorithms often fail to include factors or features indicative of a quality or performance of the items (e.g., fail to include conversionbased factors or features that indicate how often views of an item are converted into purchases and / or account for returns). Resultantly, despite offering the same exact items to those of the larger and / or well-known merchants for sale through the ecommerce platform, smaller and / or mid-sized merchants that are less well-known may struggle to achieve the visibility via item search that is necessary to engage content creators to promote their items. If not kept in balance or check, the larger and / or well-known merchants may dominate the marketplace, potentially leading to manipulative, monopolistic practices that stifle competition and leave consumers with only lower quality item options. Also, the larger and / or well-known merchants dominating within the marketplace may lead to overall lower commission rates paid to the platform, advertisers, and / or the content creators, which reduces the incentive for item promotion.

[0019] Further, item-based content generated and published by a content creator (e.g., based on discovery of the item via item search) may persist for consumer consumption throughout time. However, various events and / or factors or features affecting the earning opportunity associated with the item may change over time, which may impact a viability of the item-based content (e.g., render the item-based content less valuable and / or invaluable to the content creator), if such earning opportunity decreases. For example, if several weeks after publication, the item is noAttorney Docket No: 00273-0015-00304longer available for purchase (e.g., is out of stock or sold out), the content can no longer be monetized.

[0020] The present disclosure solves these problems and / or other problems described above or elsewhere in the present disclosure, namely by implementing a targeted item matching platform configured to generate item match sets for targeted items identified within an item data store for subsequent use in item discovery and / or event-based item replacement processes.

[0021] The item data store (e.g., an item catalog) for an ecommerce platform may store an exuberant number of items with new items being ingested and stored continuously. As an illustrative example, a weekly ingestion volume may range from thousands to millions of new items. The ecommerce platform does not have sufficient processing and / or storage resources required (or it would otherwise be highly inefficient to dedicate such required resources) to perform item matching for each and every item in the item data store. Therefore, the targeted item matching platform may first identify a portion of the items stored in the item data store to target for item matching (e.g., identify the targeted items). The targeted items may include items for which there is high user interest. The high user interest may be indicative of a likelihood of the item being included in content generated and published by content creators via the content creation platform 114, and thus a likelihood of generating business (e.g., item purchases) on the marketplace.

[0022] Once the targeted items are identified, matching items to the targeted items may be identified for inclusion in item match sets for the targeted items. A matching item may include a same or exact (e.g., an identical) item to the targeted item that is sold by a different merchant than a merchant of the targeted item.Additionally or alternatively, a matching item may include a highly similar item to the targeted item (e.g., a variant of the item, if a multi-variant item) that is sold by a different merchant than the merchant of the targeted item.

[0023] Feature-based scores for each respective item within the item match sets (e.g., the targeted item and the matching items) may be determined for use as an evaluation or ranking metric among the items within a given item match set. In some examples, the score may be based on a conversion-based feature that is indicative of an earning opportunity associated with the respective item, such as an earnings per click (EPC) value. The EPC value may be determined, in part, based on monitoring of resources (e.g., merchant websites) for the plurality of items includedAttorney Docket No: 00273-0015-00304in the item data store to identify various actions utilized in the EPC value determination. For example, the EPC value for an item may be determined based on a merchant conversion rate associated with the resource for the item, a number of actual item orders to account for returns, an average item order value, and an average item commission rate. The score may be further based on a current commission amount to remove any potential bias towards lower priced items created by the EPC value.

[0024] Further, multiple conversion-based features (e.g., multiple EPC values) may be determined within specific categories corresponding to category-based hierarchical levels of a hierarchical structure of the item data store that is storing the items. For example, the hierarchical levels of the hierarchical structure of the item data store may include, from highest to lowest level, parent categories, child categories, and items. EPC values may be aggregated up / across the levels of the hierarchical structure. For example, a child category EPC value may be determined based on item EPC values across the child category, and a parent category EPC value may be determined based on child category EPC values for child categories across the parent category. This provides for granularity when the EPC value is included as a scoring feature for the item match set, and particularly when an item being scored does not have an item EPC value, to help ensure that an item is evaluated within its relevant context (e.g., within the child category and / or parent category to which the item belongs), leading to more accurate and fair scoring. In other words, in recognition that item performance can vary significantly depending on the child and / or parent category the item belongs to, the granularity allows for nuanced decisions or logic associated with choice of EPC value for use in the scoring.

[0025] EPC value-based scoring promotes items from merchants having higher conversion metrics (e.g., having a higher likelihood of converting user interest into purchases) and / or merchants providing more consistent / stable and reasonable commission rates. Therefore, when the scores within an item match set are implemented in the subsequent item discovery and / or event-based item replacement processes, such promotion may (1 ) help to increase a diversity of merchants represented, including smaller to mid-sized merchants, (2) help to maximize earning opportunity for content creators and thus continue to incentivize content promotion, and (3) help to promote provision of higher quality items to consumers. For example,Attorney Docket No: 00273-0015-00304for item discovery, in addition to a targeted item determined to be responsive to a content creator’s query being displayed, the matching items to the targeted item in the item match set may also be displayed. The matching items may be displayed in association with the targeted item based on the scores to indicate earning opportunities of the respective matching items relative to the targeted item.

[0026] Because item-related information is dynamic and constantly being changed by merchants overtime (e.g., changes in commission rate, changes in item availability, etc.), the ability to implement the targeted item matching beyond item discovery, and specifically at a later time when an event triggering item replacement is detected, an earning opportunity for the content creator may continuously be maximized in real time. For example, when an event is detected that may potentially affect or alter the earning opportunity associated with a targeted item, a matching item from the item match set for the targeted item having a better earning opportunity than the targeted item (e.g., based on the scores) may be recommended. Any content including the targeted item may be identified, and the resource for the targeted item associated with that content may be replaced with a new resource for the recommend matching item. Resultantly, when a consumer interacts with the content, the consumers’ device may be automatically routed to the new resource.

[0027] While specific examples included throughout the present disclosure involve processes occurring within a content creation-based ecommerce platform context, it should be understood that techniques according to this disclosure may be adapted to any ecommerce platform. It should also be understood that the examples above are illustrative only. The techniques and technologies of this disclosure may be adapted to any suitable activity.

[0028] FIG. 1 is a diagram showing an example of a marketplace environment 100, according to some aspects of the disclosure. The marketplace environment 100 may include a plurality of computing devices 102 communicating with one or more server-side systems 106 over one or more networks, such as network 104.

[0029] The server-side systems 106 may include an ecommerce platform 108 and one or more data storage system(s) 110, among other systems. In some examples, the ecommerce platform 108 may be a content creation-based platform associated with a first entity. As one non-limiting example, the content creationbased platform may serve as a marketplace for merchants to engage content creators to help promote their items in exchange for a commission via contentAttorney Docket No: 00273-0015-00304generated and published by the content creators for consumption by consumers. The content creation-based platform may enable content creators to, among other things, search for items (e.g., products) to include in monetized content generated and published via the content creation-based platform. Additionally, the content creation-based platform may enable consumers to, among other things, interact with the content that is published. The content creation-based platform monitors the interactions to help facilitate the monetization of the content, as well as update conversion-based features determined for use in feature-based scoring of matching items, among other ranking mechanisms, used in item search and / or event-based item replacement processes, as described below.

[0030] In some embodiments, the ecommerce platform 108 and the data storage system(s) 110 may be associated with a common entity (e.g., the first entity). In such embodiments, the ecommerce platform 108 and the data storage system(s) 110 may be part of a cloud service computer system (e.g., in a data center). In other embodiments, the ecommerce platform 108 and the data storage system(s) 110 may be associated with a different entity than one another. For example, the data storage system(s) 110 may be associated with a third party that provides data storage services to the first entity. The systems and devices of the marketplace environment 100 may communicate in any arrangement. As will be discussed herein, systems or devices of the marketplace environment 100 may communicate in order to facilitate item searches and content interactions, among other activities.

[0031] The network 104 over which the one or more components of the marketplace environment 100 communicate may include one or more wired or wireless networks, such as a wide area network (“WAN”), a local area network (“LAN”), personal area network (“PAN”), a cellular network (e.g., a 3G network, a 4G network, a 5G network, etc.) or the like. In some embodiments, the network 104 includes the Internet, and information and data provided between various systems occurs online. “Online” may mean connecting to or accessing source data or information from a location remote from other devices or networks coupled to the Internet. Alternatively, “online” may refer to connecting or accessing an electronic network (wired or wireless) via a mobile communications network or device. The Internet is a worldwide system of computer networks — a network of networks in which a party at one computer or other device connected to the network can obtain information from any other computer and communicate with parties of otherAttorney Docket No: 00273-0015-00304computers or devices. The most widely used part of the Internet is the World Wide Web (often-abbreviated “WWW’ or called “the Web”). A “website page” or “web page” or “page” generally encompasses a location, data store, or the like that is, for example, hosted or operated by a computer system so as to be accessible online, and that may include data configured to cause a program such as a web browser to perform operations such as send, receive, or process data, generate a visual display or an interactive interface, or the like. The computing devices 102 and one or more of the server-side systems 106 may be connected via the network 104, using one or more standard communication protocols. The computing devices 102 and one or more of the server-side systems 106 may transmit and receive communications from each other across the network 104, as discussed in more detail below.

[0032] The computing devices 102 may be configured to enable access to or interaction with other systems in the marketplace environment 100. The computing devices 102 may be any type of computer system such as, for example, a desktop computer, a laptop computer, a tablet, a smart cellular phone (e.g., a mobile phone), a smart watch or other electronic wearable, etc.

[0033] In some embodiments, the computing devices 102 may include one or more electronic application(s), e.g., a program, plugin, browser extension, etc., installed on a memory of the respective devices. The electronic application(s) may include one or more of system control software, system monitoring software, software development tools, etc. In some embodiments, the electronic application(s) may be associated with one or more of the other components in the marketplace environment 100. For example, one or more of the electronic application(s) may include client applications associated with one or more of the server-side systems 106 (e.g., a client application of the ecommerce platform 108). In some examples, one or more of the client applications may include thick client applications that are locally installed on the computing devices 102 (e.g., desktop applications or mobile applications). In other examples, one or more of the client applications may include thin client applications (e.g., web applications) that are rendered via a web browser launched on the computing devices 102.

[0034] Additionally, the computing devices 102 may generate, or may cause to be generated, one or more graphical user interfaces (GUIs) based on instructions or information stored in the memory, instructions or information received from the other systems in the marketplace environment 100, or the like and may cause the GUIs toAttorney Docket No: 00273-0015-00304be displayed via a display of the computing devices 102. The GUIs may be, e.g., mobile application interfaces or browser user interfaces and may include text, input text boxes, selection controls, or the like. The display may include a touch screen or a display with other input systems (e.g., a mouse, keyboard, etc.) for the users of the respective devices to control the functions thereof.

[0035] As shown in FIG. 1, the computing devices 102 may include a first computing device 102A and a second computing device 102B. The first computing device 102A may be associated with a content creator subscribed to the ecommerce platform 108 (e.g., is a content creator computing device), whereas the second computing device 102B may be associated with a consumer or a follower of a content creator (e.g., is a consumer computing device).

[0036] In some embodiments, the content creator may interact with the ecommerce platform 108 through a client application running on the first computing device 102A to generate content to recommend or otherwise promote an item of interest for which the content creator is seeking monetization for (e.g., seeking commissions based on purchases of the item resulting from consumer interactions with the content). As part of the content generation process, the content creator may select an item search feature of the client application, and utilize this feature to generate a query to facilitate identification of the item for inclusion in the content. Upon generation and submission of the query, the client application may transmit the query over the network 104 to the ecommerce platform 108. In return, and as described in more detail below, the ecommerce platform 108 may provide computerexecutable instructions, via the client application over the network 104, to enable the first computing device 102A to construct and display a user interface to present a subset of items determined to be responsive to the query. The item may then be selected from the subset for inclusion in the content, and a resource for the item may be associated with the content.

[0037] In other embodiments, the consumer or the follower may use the second computing device 102B to interact with item-based content (e.g., search for, view, and / or make a selection to be routed to an item resource associated with the content) generated and published by content creators via the content creation platform. For example, the consumer or the follower may interact with the ecommerce platform 108 through a client application running on the second computing device 102B to utilize an item search feature of the client application to tryAttorney Docket No: 00273-0015-00304to locate content associated with one or more items of interest that may or may not be associated with the content creator. In some examples, the consumer or the follower may interact with the content through another platform, such as a social media platform, over which content generated via the content creation platform was re-published or shared, for example.

[0038] The ecommerce platform 108 may include one or more server devices (or other similar computing devices) for executing a variety of marketplace related services. In some examples, and as illustrated in FIG. 1, the ecommerce platform 108 may include a plurality of sub-platforms or systems, such as a search platform 112, a content creation platform 114, a behavior monitoring system 116, a conversion-based feature determination system 118, and a targeted item matching platform 120, each configured to perform a different subset of tasks to execute the marketplace related services, and each executable independently or in conjunction with one another. In other examples, one or more of the plurality of sub-platforms or systems may be external systems providing the services as third-party services to the ecommerce platform 108. As one non-limiting example, the behavior monitoring system 116 may be an external system providing behavior monitoring services to the ecommerce platform 108.

[0039] Example marketplace related services performed by the various subplatforms or systems of the ecommerce platform 108 may include, but are not limited to, tasks associated with: executing item and / or content searches for content creators and / or consumers; enabling content creation by content creators; behavioral monitoring associated with items within the ecommerce platform 108 and / or as resources for the items are accessed over the network 104 by the computing devices 102 for use in various determinations across one or more of the sub-platforms or systems; determining conversion-based features for use in various scoring and / or ranking determinations across one or more of the sub-platforms or systems; and targeted item matching to facilitate item discovery in the item search context and / or event-based item replacement.

[0040] The targeted item matching platform 120 configured to perform the targeted item matching may include a plurality of sub-systems, including an item prioritization system 122, a matching system 123, and / or a scoring system 124. As described in greater detail below, the item prioritization system 122 may be configured to identify a portion of the items stored in the item data store 126 to targetAttorney Docket No: 00273-0015-00304for item matching. The prioritization may be based on collected user intent signals indicative of higher user interest in the items included within the identified portion (e.g., a higher likelihood that these targeted items will be included in content generated and published by content creators via the content creation platform 114). For each targeted item in the portion, the matching system 123 may be configured to identify matching items (e.g., exact item matches) for inclusion in an item match set for the targeted item, and the scoring system 124 may be configured to determine scores for each of the targeted item and the matching items. The targeted item matching platform 120 may generate and store a data structure associated with the item match set for each targeted item that includes the scores for use in one or more subsequent processes (e.g., item discovery in the item search context and / or eventbased item replacement).

[0041] The data storage system(s) 110 may include a server system, computer-readable memory such as a hard drive, flash drive, disk, etc. In some embodiments, the data storage system(s) 110 include or interact with an application programming interface for exchanging data to other systems, e.g., one or more of the other components of the marketplace environment 100, such as at least the ecommerce platform 108. The data storage system(s) 110 may include a plurality of data stores, such as an item data store 126, a content data store 128, a behavior data store 130, a feature data store 132, a user intent data store 134, and an item match data store 136. The data stores may include or act as a repository or source for various types of data.

[0042] The item data store 126 may be configured to store a plurality of items queryable via the search platform 112 of the ecommerce platform 108. For example, the item data store 126 may be an item or product catalog. The items may include products and / or services that each have a plurality of attributes unique to a type of the respective item. To provide an illustrative example, the items may include goods or merchandise, such as products available for purchase or sale by one or more merchants. Example item attributes stored in association with a given item in the item data store 126 may include a title, a description, an image, a brand, a collection, a size, a color, a style, and / or a merchant, among other similar attribute types, associated with the given item. Other example item information stored may include a price of the item, a commission rate for the item, a current commission amount, or other similar attribute that is defined by the merchant and dynamic in nature. TheAttorney Docket No: 00273-0015-00304items stored in the item data store 126 may include items scraped from external resources (e.g., scraped from the Internet), items previously interacted with by users (e.g., in the content creation platform scenario, items previously included in generated content and / or saved, favorited, or otherwise interacted with by a content creator), or the like. The item data store 126 may be continuously updated to include new items as they are identified via scraping and / or via user interactions.

[0043] In some embodiments, the item data store 126 may be a vector-based database configured to store the items as vector representations of the items (e.g., as item vectors) in a vector space. A vector for an item may be a mathematical representation generated based on one or more of the attributes comprising the data object.

[0044] Additionally, the item data store 126 may include (e.g., may be organized according to) a hierarchical structure having a plurality of hierarchical levels. The hierarchical structure may be category based. For example, the hierarchical structure may include, from a highest hierarchical level to a lowest hierarchical level, a plurality of parent categories, one or more child categories for one or more of the plurality of parent categories, and respective items that are each associated with at least one of the parent categories and / or one of the one or more child categories for a respective parent category. In some examples, the child categories may further include one or more additional levels of child sub-categories. To provide an illustrative example, a parent category may include clothing, a child category of the clothing parent category may include bottoms, a child sub-category of the bottoms child category may include jeans, and an item of the child sub-category of jeans may be a particular brand and style of jeans. As described in detail below, the hierarchical structure of the item data store 126 may be leveraged by the conversion-based feature determination system 118 to determine one or more conversion-based features associated with an item at one or more hierarchical levels corresponding to the hierarchical structure for subsequent use in scoring and / or ranking algorithms, such as scoring performed by scoring system 124 of the targeted item matching platform 120.

[0045] The content data store 128 may be configured to store any content including an item from the plurality of items that are stored in the item data store 126, such as content generated by a content creator via the content creation platform 114. For example, the item may be an item that was selected from item search results for inclusion in content generated and published by the content creator forAttorney Docket No: 00273-0015-00304monetization purposes. A resource for the item may be stored in association with the content in the content data store 128 such that when a consumer or follower interacts with the content via the second computing device 102B, the second computing device 102B can be caused to access (e.g., can be routed to) the resource for the item. In some examples, the resource for the item may be replaced with an alternative, new resource for a different item (e.g., a matching item) at a later point in time when an event is detected, as described in detail below.

[0046] The behavior data store 130 may be configured to store actions performed in association with items stored in the item data store 126. The actions may be identified based on behavior monitoring, performed by the behavior monitoring system 116, of resources for the items (e.g., merchant websites or web pages for the items) as the resources are accessed over one or more networks, such as the network 104. A count or number associated with one or more different types of actions may be monitored for and stored. Example types of actions monitored for may include a click that causes access to the resource, and / or a purchase of the target item via the resource as a result of the click. In some examples, the count or number stored may be updated in real-time or on a periodic basis.

[0047] The feature data store 132 may be configured to store one or more values determined for one or more features of at least a subset of items stored in the item data store 126. For example, the values may include at least conversion-based feature values for the subset of items. The conversion-based feature determination system 118 may determine the conversion-based feature values based, in part, on the information stored in the behavior data store 130 for the subset of items and / or information stored in association with the subset of items in the item data store 126. In some examples, the feature values stored may be updated in real-time as the information stored in the behavior data store 130 and / or item data store 126 is updated or on a periodic basis.

[0048] The user intent data store 134 may be configured to store a plurality of signals indicative of user intent that are tracked as one or more of the items stored in the item data store 126 are interacted with via the computing devices 102 over the network 104 (e.g., interacted with internally within the ecommerce platform 108). Example types of signals tracked across users of the ecommerce platform 108 (e.g., content creators and / or consumers) may include a link generated for an item by a content creator, a favoring of an item by a content creator, a selection of an itemAttorney Docket No: 00273-0015-00304from a set of search results by a content creator, and / or brand popularity indicated by user interactions. As described in more detail below, the signals may be used by the targeted item matching platform 120, and particularly the item prioritization system 122, to identify a portion of the items stored in the item data store 126 to target for item matching. In some examples, the user intent data store 134 may be combined with the behavior data store 130 into a single data store including interaction / behavior information occurring both internally within the ecommerce platform 108 and externally as the resources for the items are accessed over the network 104 by the computing devices 102.

[0049] The item match data store 136 may be configured to store the data structures associated with the item match sets determined for the portion of the items to be targeted. An example data structure may include a graph that includes, for each of the targeted item and the matching items to the targeted item in the item match set: an item identifier, one or more category-based identifiers corresponding to the hierarchical structure of the item data store 126 (e.g., parent and / or child category identifiers), a resource (e.g., a link to a website, web page, or application of a merchant associated with the respective item), and / or the score determined for the respective item by the scoring system 124.

[0050] Although depicted as separate components in FIG. 1, it should be understood that a component or portion of a component in the system of the marketplace environment 100 may, in some embodiments, be integrated with or incorporated into one or more other components. For example, one or more of data storage system(s) 110 may be integrated with the ecommerce platform 108 or the like. As another example, the item prioritization system 122, the matching system 123, and / or the scoring system 124 may comprise a single system of the targeted item matching platform 120. As a further example, one or more of the search platform 112, the content creation platform 114, the behavior monitoring system 116, the conversion-based feature determination system 118, and / or the targeted item matching platform 120 may comprise a single system of the ecommerce platform 108. In some embodiments, operations or aspects of one or more of the components discussed above may be distributed amongst one or more other components. Any suitable arrangement or integration of the various systems and devices of the marketplace environment 100 may be used.Attorney Docket No: 00273-0015-00304

[0051] In the following disclosure, various acts may be described as performed or executed by a component from FIG. 1, such as the computing devices 102 or one or more of the server-side systems 106, or components thereof. However, it should be understood that in various embodiments, various components of the marketplace environment 100 discussed above may execute instructions or perform acts including the acts discussed below. An act performed by a device may be considered to be performed by a processor, actuator, or the like associated with that device. Further, it should be understood that in various embodiments, various steps may be added, omitted, or rearranged in any suitable manner.

[0052] FIG. 2 is a flow chart showing an example process 200 for determining conversion-based features, according to some aspects of the disclosure. The process 200 may be performed by one or more of the server-side systems 106, such as one or more components of the ecommerce platform 108, including the behavior monitoring system 116 and the conversion-based feature determination system 118.

[0053] At step 202, the process 200 may include, for an item stored in the item data store 126, monitoring behavior associated with the item as a resource for the item is accessed over a network, such as the network 104. The resource for the item may be a website, a web page, or other similar page of a merchant that sells the item, and includes information associated with the item (e.g., an image, description, price, availability, etc.). The resource may be accessed via the computing devices 102 over the network 104, and behavior of the users performed in association with the item as the resource is accessed via the computing devices 102 may be monitored.

[0054] At step 204, the process 200 may include identifying one or more actions performed in association with the item based on the monitoring. The actions may include a click causing the resource for the item to be accessed and displayed on the computing devices 102 and / or interaction(s) with the item via the resource, such as a view, a purchase, etc. A count or number associated with the different types of actions identified may be stored in the behavior data store 130. In some examples, the count or number stored may be updated in real-time or on a periodic basis.

[0055] At step 206, the process 200 may include, based on the one or more actions, determining one or more conversion-based features at one or more of a plurality of hierarchical levels corresponding to a hierarchical structure of the item data store 126. As described above with reference to FIG. 1 , the hierarchicalAttorney Docket No: 00273-0015-00304structure of the item data store 126 may be category-based and include, from a highest hierarchical level to a lowest hierarchical level, a plurality of parent categories, one or more child categories for one or more of the parent categories, and respective items that are each associated with at least one of the parent categories and / or one or more child categories for a respective parent category.

[0056] Based on the actions performed in association with the item, at least a first conversion-based feature may be determined. The first conversion-based feature may be determined at a first hierarchical level corresponding to the item (e.g., the lowest hierarchical level of the hierarchical structure of the item data store 126). In some examples, the first conversion-based feature is an earnings per click (EPC) value for the item. The EPC value may be based on a merchant conversion rate associated with the resource for the item, a number of actual item orders, an average item order value, and an average item commission rate. The first conversion-based feature may also be referred to herein as a first EPC value for the item or an item EPC value. An example EPC formula is included below:EPC = (Converted click / clickout) * ((Orders * (1 - Return Rate)) / Converted click) * Average commission rate * Average order value.

[0057] Within the formula, converted click is a number of clicks received that caused the resource for the item to be accessed and resulted in a purchase or sale of the item, whereas clickout is a total number of clicks that caused the resource for the item to be accessed. Therefore, the division of the converted click by the clickout provides the merchant conversion rate associated with the resource for the item, which indicates an efficiency of the traffic that lands on the resource. In determining this efficiency, other factors may be taken into account, such as an attribution window (e.g., transactions outside the attribution window won’t be classified as converted clicks), action caps (e.g., if a transaction occurred after a cap was met, the transaction won’t be associated with a converted click), trackability, inventory availability, and an overall merchant conversion rate (e.g., including conversion rates for other items sold by the merchant).

[0058] Based on the inclusion of the converted click in both the numerator of the first parentheses and in the denominator of the second parentheses (e.g., cancelling one another out), the EPC formula may be simplified as follows:EPC = ((Orders * (1 - Return Rate)) / clickout)* Average commission rate * Average order value.Attorney Docket No: 00273-0015-00304

[0059] Within the formula (full or simplified), orders refers to a total number of orders (e.g., purchases) of the item from the merchant, and return rate is a number of times an order, from the total number of orders, has been returned to the merchant. Thus, the EPC formula includes the actual number of orders, taking the merchant’s return rate into consideration.

[0060] A commission rate is a percentage of a total sale value that is set by the merchant and paid to a content creator, for example, when the item is sold as a result of a consumer interacting with the content creator’s content including the item. Therefore, within the formula, average commission rate is an average of commission rates set by the merchant over a period of time. Average order value is an average of total values of a sale (e.g., average sale prices) for the item over the same period of time that, when multiplied by the average commission rate within the formula, indicates an average amount of commission for the item.

[0061] In some examples, the number of orders, number of returns, return rate, commission rates, and / or prices may be received periodically from the merchant. Additionally and / or alternatively, at least a portion of these values may be tracked internally by the ecommerce platform 108 and / or components thereof (e.g., obtained as part of the data collected by the behavior monitoring system 116 when actions associated with the merchant resources are detected). In some examples, when there is discrepancy between data received from the merchant and the internally tracked data, the ecommerce platform 108 may initiate recovery actions with the merchant to identify and recover any lost data creating the discrepancy.

[0062] Thus, as a whole, the EPC yields the amount of commission for the item per clickout. In other words, the EPC provides a signal for potential earning opportunity associated with the item, where a higher EPC may indicate a higher (e.g., more optimal) earning opportunity due to a higher likelihood of user interest in the item being converted to a purchase.

[0063] In some examples, the first conversion-based feature determined for the item (e.g., the first EPC value) may then be aggregated up to yield additional EPC values at one or more higher hierarchical levels corresponding to a child category and / or a parent category under which the item falls (e.g., is associated with) in the hierarchical data structure of the item data store 126. For example, to determine a second conversion-based feature at a second hierarchical level for the child category, the first conversion-based feature determined at the first hierarchical levelAttorney Docket No: 00273-0015-00304for the item may be aggregated with one or more other first conversion-based features determined at the first hierarchical level for one or more other items stored in the item data store 126 that are associated with a same child category as the item. The second conversion-based feature may also be referred to herein as a second EPC value for the child category or a child category EPC value.

[0064] As another example, to determine a third conversion-based feature at a third hierarchical level for the parent category, the second conversion-based feature determined at the second hierarchical level for the child category may be aggregated with one or more other second conversion-based features determined at the second hierarchical level for one or more other child categories associated with a same parent category. The third conversion-based feature may also be referred to herein as a third EPC value for the parent category or a parent category EPC value.

[0065] For brevity and clarity, the item, child category, and parent category EPC values are discussed herein as examples of the conversion-based features at the hierarchical levels corresponding to the structure of the item data store 126.However, in other examples, additional EPC values may be determined at further hierarchical levels, such as child sub-category levels. An item’s performance can vary significantly depending on a parent category, child category, or child subcategory the item belongs to. By determining EPC values at the plurality of hierarchical levels to provide more granularity, as opposed to a more general EPC value, items may be evaluated within a relevant context for use in dynamic scoring and / or ranking, a described in detail below.

[0066] At step 208, the process 200 may include storing the one or more conversion-based features for subsequent use in one or more processes. For example, the conversion-based features may be stored in the feature data store 132. Within the feature data store 132, the first conversion-based feature (e.g., the item EPC value) may be stored in association with the item, the second conversion-based feature (e.g., the child category EPC value) may be stored in association with the respective child category, and the third conversion-based feature (e.g., the parent category EPC value) may be stored in association with the respective parent category.

[0067] One example process in which at least one of the conversion-based features may be subsequently used is targeted item matching to determine a plurality of scores for each of a targeted item and matching items to the targetedAttorney Docket No: 00273-0015-00304item, as described in detail with reference to FIGs. 3-5. The determined scores may then be used for item ranking and / or recommendation processes. For example, in an item discovery or query process in which the targeted item is determined to be responsive to the query, the matching items may be displayed in association with the targeted item within the item search results based on the scores, as described in detail with reference to FIG. 6. In some examples, when the targeted item is selected from the item search results for inclusion in content that is generated and published (e.g., via the content creation platform 114), a resource for the targeted item is associated with the content. As another example, in an event-based item replacement process, the scores may be used to identify a recommended matching item from the matching items when an event is detected that triggers a replacement of the resource for the targeted item associated with the content with a new resource for the recommended matching item, as described in detail with reference to FIG. 7.

[0068] For brevity and clarity, the process 200 describes the determination of one or more conversion-based features at one or more of the plurality of hierarchical levels for one item from the plurality of items stored in the item data store 126.However, the process 200 may be performed for each item within at least a subset of the plurality of items for which actions have been identified based on the monitoring.

[0069] Accordingly, certain embodiments may be performed for conversion-based feature determination. The process 200 described above is provided merely as an example, and may include additional, fewer, different, or differently arranged steps than depicted in FIG. 2 describing the process 200 and / or portions thereof.

[0070] FIG. 3 is a flow chart showing an example process 300 for targeted item matching, according to some aspects of the disclosure. The process 300 may be performed by one or more of the server-side systems 106, such as one or more components of the targeted item matching platform 120, including the item prioritization system 122, the matching system 123, and / or the scoring system 124.

[0071] At step 302, the process 300 may include determining a portion of the plurality of items stored in the item data store 126 to target for item matching (e.g., determining targeted items). The determination may include receiving and processing a plurality of prioritization signals, where at least a portion of the prioritization signals include user intent signals. For example, the item prioritization system 122 may receive from the user intent data store 134 a plurality of signals indicative of user intent (e.g., user intent signals). The user intent signals may beAttorney Docket No: 00273-0015-00304collected and stored by the behavior monitoring system 116 based on monitored or tracked user behavior within the ecommerce platform 108. Example types of user intent signals collected may include a link generated for an item by a content creator, a favoriting of an item by a content creator, a selection of an item from a set of search results by a content creator, and / or brand popularity indicated by user interactions (e.g., top A / brands based on how often items of the brand are interacted with by content creators and / or consumers). The user intent signals may be stored based on the signal type to enable quantitative metrics for each signal type to be generated for a given item.

[0072] The item prioritization system 122 may determine the portion of items to target by processing the plurality of signals. For example, items associated with user intent signals indicative of higher user interest in the items (e.g., having higher qualitative metrics above a predefined threshold) may be identified for targeting. Specifically, higher user interest for these items may translate to a higher likelihood that these items will be included in content generated and published by content creators via the content creation platform 114, and thus should be targeted for item matching.

[0073] In some examples, as part of the signal processing, the signals may be weighted based on the signal type. The weights applied to each signal type may be different to reflect a level of user interest or intent attributed to each signal type. To provide an illustrative example, a generated item link may receive a highest weight, a favorited item may receive an intermediate weight, and an item selection from a set of search results (e.g., an item click) may receive a lowest weight, among these signal types, to correspond to the user’s level of interest. That is, when a user clicks on an item to view the item, user interest is indicated, but at a lower level than the user actively favoriting that item (e.g., after viewing) and / or generating a link for that item. Additionally, a weight for a given signal type may be further broken down into a plurality of weights. For example, for link generation, a different weight may be applied based on a creator segment the content creator who generated the link belongs to (e.g., with higher weights applied to more active and / or higher earning content creators as identified by segment).

[0074] In other aspects, as part of the processing, additional signals may be considered and weighted in conjunction with the user intent signals. As one example, weight(s) may be applied to prioritize items associated with a brand that is sold byAttorney Docket No: 00273-0015-00304multiple merchants, and thus have an increased likelihood of having exact item matches stored in the item data store 126, as described in detail below. As another example, weights may be applied to prioritize items associated with one or more specific parent categories, child categories, and / or child sub-categories. As a further example, weights may be applied to prioritize items associated with one or more currencies (e.g., to promote items being sold by merchants in one or more countries associated with the one or more currencies).

[0075] The above-described weights may be variable or adjustable, and may be defined by administrators associated with the ecommerce platform 108. The combination of prioritization signals utilized for the determination may be similarly variable or adjustable. In some examples, the determination of targeted items for item matching performed at step 302 may be repeated on a periodic basis (e.g., daily, weekly, etc.) using a same and / or different combination of prioritization signals.

[0076] The item data store 126 may store an exuberant number of items with new items being ingested and stored continuously. As an illustrative example, a weekly ingestion volume may range from thousands to millions of the items. The ecommerce platform 108 does not have sufficient processing and / or storage resources required (or it would otherwise be highly inefficient to dedicate such required resources) to perform item matching, as described below with reference to steps 304 through 308, for each and every item in the item data store 126.Therefore, the determination of the portion of items from the item data store 126 to target for item matching at step 302, enables the ecommerce platform 108 to dedicate the processing and storage resources to a smaller, more manageable number of items that have a highest likelihood of being included in content generated and published by content creators via the content creation platform 114, and thus have a highest likelihood of generating business (e.g., item purchases) on the marketplace.

[0077] At step 304, the process 300 may include, for a targeted item in the portion, identifying a plurality of matching items to the targeted item for inclusion in an item match set for the targeted item. As described in detail with reference to FIG.1, the item data store 126 may store item attributes associated with each of the plurality of items stored therein. To identify the matching items, the matching system 123 may query the item data store 126 to obtain at least a portion of item attributes stored for the targeted item, and a corresponding portion of item attributes for eachAttorney Docket No: 00273-0015-00304of a plurality of candidate matching items from the plurality of items for comparison against the item attributes for the targeted item (e.g., using an item matching algorithm). Example item attributes considered for the comparison (e.g., provided as input to the item matching algorithm) may include a title, a description, a price, a brand, a merchant, and / or an image.

[0078] Based on the comparison, at least a subset of the candidate matching items may be determined as the matching items for inclusion in the item match set. For example, the corresponding item attributes for each of the subset of the candidate matching items may have a similarity to the item attributes for the targeted item that is above a predetermined similarity threshold (e.g., an output of the item matching algorithm is a similarity value above the predetermined similarity threshold). Resultantly, a matching item may include a same or exact (e.g., an identical) item to the targeted item that is sold by a different merchant than the merchant of the targeted item. Additionally or alternatively, a matching item may include a highly similar item to the targeted item that is sold by a different merchant than the merchant of the targeted item. The similar item may include at least a portion of the same attributes as the item (e.g., the similar item may be a same style but different color or size than the item). In other words, the similar item may be an item variant of the targeted item. In some examples, a predefined number of matching items may be identified. As one non-limiting example, five or six matching items may be identified.

[0079] In some aspects, more than one predetermined similarity threshold may be utilized by the item matching algorithm. For example, a first confidence list (e.g., a high confidence list) may be generated to include a first subset of candidate matching items having a similarity to the targeted item above a higher, first predetermined similarity threshold. The first subset of candidate matching items included in the first confidence list may be automatically identified as matching items for inclusion in the item match set for the targeted item. A second confidence list (e.g., a low confidence list) may be generated to include a second subset of candidate matching items having a similarity to the targeted item above a lower, second predetermined similarity threshold. The second confidence list may be included within a communication (e.g., a notification, alert, message, etc.) that is transmitted to an administrator of the ecommerce platform 108 to request feedback on whether any of the second subset of candidate matching items are matchingAttorney Docket No: 00273-0015-00304items to the targeted item. Based on the feedback received in response to the communication, any of the second subset of candidate matching items indicated by the administrator as matching items may be included in the item set match for the targeted item. Additionally or alternatively, the feedback may be utilized to update, modify, or otherwise adjust the item matching algorithm to improve the accuracy thereof.

[0080] At step 306, the process 300 may include determining a plurality of scores for the item match set, including a score for each of the targeted item and the plurality of matching items in the item match set based on a plurality of features for the respective item. As described in more detail with reference to FIG. 4, a feature value associated with each feature of the plurality of features may be determined, weighted, and normalized for use in determining the score for a respective item within the item match set. Any combination of features may be used for determining the score.

[0081] In some aspects, the features may include a conversion-based feature. An example of the conversion-based feature may be an EPC value indicative of an amount of commission per clickout (e.g., signaling earning opportunity) that is determined as described above with reference to FIG. 2. In some examples, one or more of the items within the item match set (e.g., the targeted item and / or one or more of the matching items to the targeted item) may be items that are included within the subset of items for which conversion-based features have been determined, as described with reference in FIG. 2. Additionally or alternatively, one or more of the items within the item match set may not be included within this subset. As described in more detail with reference to FIG. 5, whether an item is included in the subset may impact which conversion-based feature is available for use as the feature value when determining the score. Specifically, a conversion-based feature at a lowest hierarchical level of the plurality of hierarchical levels available from the feature data store 132 may be determined and retrieved for use. In some examples, when a conversion-based feature for the item itself is not available, a conversionbased feature for one or more semantically similar items to the item may be used to derive the item’s score.

[0082] The plurality of features may also include a current commission amount (e.g., current commission dollars). The current commission amount may be determined based on a current price of the item and a current commission rate forAttorney Docket No: 00273-0015-00304the item that can be obtained from the item data store 126. Use of the conversionbased feature (e.g. the EPC value) alone in the score determination may create a bias towards lower priced items. Specifically, lower priced items may often have higher commission rates and a higher number of actual item orders (which are each accounted for in the EPC determination). However, an item having higher commission rates and / or number of actual item orders doesn’t necessarily equate to that item providing a content creator a better or more optimal earning opportunity. To provide an illustrative example, a lower commission rate of 6% on a $100 item yields a more optimal earning opportunity (e.g., a current commission amount of $6 per item) than a higher commission rate of 8% on a $10 item (e.g., a current commission amount of only $0.80 per item). Therefore, including the current commission amount as an additional feature for use in determining the score in conjunction with the EPC value accounts for this bias to yield a more accurate evaluation of earning opportunity (e.g., as reflected by the score).

[0083] In some examples, in addition to, or alternatively from, the EPC value and the current commission amount, other features and / or values associated with the item may be considered when determining the score. As one illustrative example, a popularity-based feature, such as a number of times the item has been viewed (e.g., selected in response to an item search but not linked), may be determined and used in conjunction with, or alternatively from, the EPC value and the current commission amount to determine the score.

[0084] At step 308, the process 300 may include generating and storing a data structure associated with the item match set for the targeted item for subsequent use in one or more processes. The data structure may be stored in association with the targeted item at the item match data store 136. In some examples, the data structure may be a graph that is built based on the item match set. For example, for each item in the item match set (e.g., each of the targeted item and the matching items), the graph may include an item identifier, one or more category-based identifiers corresponding to the hierarchical structure of the item data store 126 (e.g., parent and / or child category identifiers), a resource (e.g, a link to a website, web page, or application of a merchant associated with the respective item), and the score determined for the respective item.

[0085] One example process that the data structure may be used or referenced for is an item discovery or query process in which the targeted item is determined toAttorney Docket No: 00273-0015-00304be responsive to the query. For example, the data structure may be referenced to identify and display the matching items in association with the targeted item within the item search results based on the scores, as described in detail with reference to FIG. 6. Another example process that the data structure may be used for is an eventbased item replacement process. For example, the data structure may be referenced to identify a recommended matching item from the matching items based on the scores when an event is detected to cause replacement of the resource for the targeted item with a new resource for the recommended matching item, as described in detail with reference to FIG. 7.

[0086] For brevity and clarity, steps 304 through 308 of the process 300 describe the item matching process for one of the targeted items identified within the portion at step 302. However, steps 304 through 308 of the process 300 may be performed for each targeted item within the portion identified at step 302.

[0087] As previously mentioned, new items (and potentially matching items) may be continuously added to the item data store 126. Additionally, the features used to determine the scores for the item match sets are dynamic in nature and may change drastically over a period a time. Therefore, one or more steps of the process 300 may be performed repeatedly on a periodic basis to generate new item match sets for newly identified targeted items, adjust item match sets for existing targeted items (e.g., adjust matching items and / or one or more of the scores), and / or remove item match sets (e.g., delete the data structure associated therewith) for items that are no longer being targeted.

[0088] Accordingly, certain embodiments may be performed for targeted item matching. The process 300 described above is provided merely as an example, and may include additional, fewer, different, or differently arranged steps than depicted in FIG. 3 describing the process 300 and / or portions thereof.

[0089] FIG. 4 is a flow chart showing an example process 400 for determining, as part of the targeted item matching process, a feature-based score for a respective item of an item match set, according to some aspects of the disclosure. The process 400 may be performed by one or more of the server-side systems 106, such as one or more components of the targeted item matching platform 120, including the scoring system 124. One or more steps of the example process 400 may be used to perform step 306 of the process 300 for determining the plurality of scores for the item match set, as described with reference to FIG. 3.Attorney Docket No: 00273-0015-00304

[0090] At step 402, the process 400 may include, for a feature of the plurality of features (forming the basis of the score determination), determining a feature value for a respective item of the item match set. The respective item may include the targeted item or one of the matching items to the targeted item included in the item match set. How the feature value is determined is based on (e.g., is dependent on) the feature type.

[0091] For example, for a conversion-based feature of the respective item, the feature value may be an EPC value obtained from the feature data store 132 that is indicative of an amount of commission per clickout (e.g., signaling earning opportunity). Specifically, and as described in more detail with reference to FIG. 5, the EPC value at a lowest hierarchical level of the plurality of hierarchical levels available from the feature data store 132 may be used as the feature value.

[0092] Additionally, dependent on a hierarchical level of the EPC value available for use as the feature value, a factor (e.g., a percentage value) may be applied to adjust the EPC value obtained from the feature data store 132. For example, a value of the factor applied causes the EPC value to decrease when higher hierarchical levels of the EPC value are used as the feature value. The EPC value is decreased to account for the greater generalization (and thus lower accuracy) of the EPC value at the higher child category and / or parent category hierarchical levels as opposed to the item hierarchical level. This helps to, for example, prevent a parent or child category’s EPC value from unfairly boosting items from merchants due to high EPC values associated with items in other child categories or sub-categories that are not relevant to the targeted item and / or matching items at hand. Specifically, a value of the factor applied to a parent category EPC value may be a lesser value than a value of the factor applied to a child category EPC value to more significantly decrease the parent category EPC due to the higher hierarchical level associated with the parent category than the child category. Further, a value of the factor applied to the EPC value may be variable dependent on the particular parent category or child category the EPC value is associated with. For example, a value of the factor applied to a first parent category EPC value (e.g., for a parent category of “shoes”) may be different from a value of the factor applied to a second parent category EPC value (e.g., for a parent category of “clothing”).

[0093] In addition to the conversion-based feature, another example feature of the plurality of features used as a basis for the score determination is a currentAttorney Docket No: 00273-0015-00304commission amount. The current commission feature value for the respective item may be determined as a function of (e.g., a multiplication of) a current price of the item and a current commission rate for the item that can each be obtained from the item data store 126.

[0094] At step 404, the process 400 may include applying a weight associated with the feature to the feature value determined at step 402 to obtain a weighted feature value for the respective item. A value of the weight (e.g., a weight value) applied to the feature value may indicate a relative importance or contribution of the feature, as compared to other features, to the overall score. As one non-limiting example, the weight applied to the EPC value may be of a higher weight value than the weight applied to the current commission feature value, indicative of a higher importance or contribution of EPC value, as opposed to current commission amount, in yielding an accurate depiction of earning opportunity associated with the respective item.

[0095] At step 406, the process 400 may include normalizing the weighted feature value to obtain a normalized, weighted feature value for the respective item.Example normalization techniques applied may include min-max scaling, z-score scaling, or decimal scaling, among other similar normalization processes.

[0096] Steps 402, 404, and 406 may be repeated for each feature of the plurality of features to obtain a plurality of normalized, weighted feature values for the respective item associated with the plurality of features.

[0097] At step 408, the process 400 may include determining the score for the respective item based on the plurality of normalized, weighted feature values for the respective item associated with the plurality of features. The normalized, weighted feature values may be combined together using an operation (e.g., may be added, subtracted, multiplied, or divided) to determine the score for the respective item. An example formula for the score is included below:Score = Normalize(w1 * current commission amount) + Normalize(w2 * EPC value).

[0098] Steps 402 through 408 of the process 400 may then be repeated for each remaining item included in the item match set. Resultantly, a plurality of scores, including a score for each of the targeted item and the matching items may be determined and stored within the data structure associated with the item match set.Attorney Docket No: 00273-0015-00304

[0099] Accordingly, certain embodiments may be performed for a feature-based score determination as part of the targeted item matching process. The process 400 described above is provided merely as an example, and may include additional, fewer, different, or differently arranged steps than depicted in FIG. 4 describing the process 400 and / or portions thereof.

[0100] FIG. 5 is a flow chart showing an example process 500 for determining, as part of the feature-based score determination process, a conversion-based feature value for the respective item of the item match set, according to some aspects of the disclosure. The process 500 may be performed by one or more of the server-side systems 106, such as one or more components of the targeted item matching platform 120, including the scoring system 124. One or more steps of the example process 500 may be used to perform step 402 of the process 400 for determining a conversion-based feature value (e.g., an EPC value) for the respective item of the item match set, as described with reference to FIG. 4. The conversion-based feature may be one of the features used as a basis for determining the score, as described with reference to step 306 of the process 300 of FIG. 3.

[0101] At step 502, the process 500 may include determining an availability of one or more conversion-based features (e.g., one or more EPC values) at one or more of the plurality of hierarchical levels. Based on the determination, the conversion-based feature at the lowest hierarchical level available may be used as the conversion-based feature value. To determine the availability, the scoring system 124 may query the feature data store 132 using the respective item and / or one or both of a child category and a parent category associated with the respective item. The respective item may be the targeted item or one of the matching items to the target item included in the item match set.

[0102] The availability may be based, in part, on whether the respective item is one of the subset of items for which at least a first conversion-based feature has been determined at a first hierarchical level corresponding to the item (and optionally aggregated up to one or more additional conversion-based features at one or more higher hierarchical levels corresponding to the child category or the parent category). In other words, the availability may be based on whether or not the respective item is an item for which actions have been identified via the behavior monitoring to allow for conversion-based feature determination.Attorney Docket No: 00273-0015-00304

[0103] To determine the conversion-based feature at the lowest hierarchical level available, at decision 504, a first determination is made as to whether a first conversion-based feature for the respective item is available (e.g., whether the item EPC value is available). As described in detail with reference to FIG. 2, the first conversion-based feature is determined at a first hierarchical level corresponding to the item, which is the lowest hierarchical level available. Therefore, if at decision 504, the first conversion-based feature for the item is determined to be available (e.g., because the respective item is one of the subset of items for which conversionbased feature determination has been performed), at step 506, the process 500 may include using a value for the first conversion-based feature (e.g., the item EPC value) as the feature value. Otherwise, if at decision 504, the first conversion-based feature for the respective item is determined to be unavailable, the process 500 may proceed via path A associated with a semantic similarity approach or path B associated with a category approach.

[0104] If the process 500 proceeds with path A, at step 508, the process 500 may include determining a conversion-based feature for one or more semantically similar items for use as the feature value for the respective item. For example, one or more semantically similar items to the respective item may be identified. The semantically similar items may be determined based on a similarity (e.g., above a predetermined threshold) between item vectors of the respective item and each semantically similar item. To provide an illustrative example, the respective item is "guest wedding dress", but no conversion-based feature is stored for “guest wedding dress” in the feature data store 132 based on a lack of action data obtained for this item via the behavior monitoring (e.g., the first conversion-based feature for the respective item is determined to be unavailable). However, items having a title of "wedding guest dress" or "wedding dress for guest” that are semantically similar to “guest wedding dress” may be identified as semantically similar items. Additionally, the semantically similar items identified may be items for which a first conversion-based feature has been determined at a first hierarchical level corresponding to the semantically similar items (e.g, semantically similar items to the respective item that are included in the subset of items). The conversion-based feature for use as the feature value for the respective item may then be derived based on the first conversion-based feature for each of the one or more semantically similar items (e.g., obtained from the feature data store 132). For example, the first conversion-based feature for each of the oneAttorney Docket No: 00273-0015-00304or more semantically similar items may be weighted and / or combined to determine the conversion-based feature for the respective item.

[0105] If the process 500 proceeds with path B, at decision 510, the process 500 may include a second determination as to whether a second conversion-based feature fora child category associated with the respective item (e.g. whether the child category EPC value is available). As described in detail with reference to FIG.2, the second conversion-based feature is determined at a second hierarchical level corresponding to the child category based on an aggregation of first conversionbased features determined at the first hierarchical level across items within the child category. The second hierarchical level corresponding to the child category is a higher hierarchical level than the first hierarchical level corresponding to the item, but a lower hierarchical level than the third hierarchical level corresponding to the parent category. Therefore, if at decision 510, the second conversion-based feature for the child category is determined to be available, at step 512, the process 500 may include using a value for the second conversion-based feature (e.g., the child category EPC value) as the feature value. In some examples, the value for the second conversion-based feature may be adjusted by a first factor (e.g., decreased by a first percentage) to account for a lower accuracy of the value given the higher hierarchical level. Otherwise, if at decision 510, the second conversion-based feature for the child category is determined to be unavailable, the process 500 may proceed to decision 514. Alternatively, in examples where no child category is associated with the respective item, then the process skips decision 510 and proceeds to decision 514.

[0106] At decision 514, the process 500 may include a third determination as to whether a third conversion-based feature for a parent category associated with the respective item is available (e.g. whether the parent category EPC value is available). As described in detail with reference to FIG. 2, the third conversion-based feature is determined at a third hierarchical level corresponding to the parent category based on an aggregation of second conversion-based features determined at the second hierarchical level across child categories of the parent category. The third hierarchical level corresponding to the parent category is a higher hierarchical level than the first hierarchical level corresponding to the item and the second hierarchical level corresponding to the child category.Attorney Docket No: 00273-0015-00304

[0107] If at decision 514, the third conversion-based feature for the parent category is determined to be available, at step 516, the process 500 may include using a value for the third conversion-based feature (e.g., the parent category EPC value) as the feature value. In some examples, the value for the third conversionbased feature may be adjusted by a second factor (e.g., decreased by a second percentage) to account for a lower accuracy of the value given the higher hierarchical level. Additionally, the second factor may be a different (e.g., lesser) value than the first factor applied to adjust the second conversion-based feature to cause an even greater decrease in the value of the third conversion-based feature. Otherwise, if at decision 514, the third conversion-based feature for the parent category is determined to be unavailable, the process 500 may end or alternatively proceed to path A, as described above. Alternatively, in examples where no parent category is associated with the respective item, then the process skips decision 514, the process 500 ends or proceeds to path A.

[0108] Accordingly, certain embodiments may be performed for conversion-based feature determination. The process 500 described above is provided merely as an example, and may include additional, fewer, different, or differently arranged steps than depicted in FIG. 5 describing the process 500 and / or portions thereof.

[0109] FIG. 6 is a flow chart showing an example method 600 for processing a query for an item search that implements the targeted item matching, according to some embodiments of the disclosure. The method 600 may be performed by one or more of the server-side systems 106, such as one or more components of the search platform 112.

[0110] At step 602, the method 600 may include receiving, from the first computing device 102A, a query for an item search. To provide an illustrative example, the client application associated with the ecommerce platform 108 described above with reference to FIG. 1 may be running on the first computing device 102A associated with the content creator, and the content creator may select to navigate to an item search feature of the client application. In response to the selection, the client application may display a search user interface (e.g., an item search page) having one or more control elements, such as a search field and / or one or more search filter control elements, for entering the query. As one example, the content creator may intend to generate content that promotes or otherwise recommends (e.g., for monetization purposes) running shoes that are lightweightAttorney Docket No: 00273-0015-00304and casual. To help the content creator identify a pair of running shoes for inclusion in the content to establish a monetization pathway, the content creator may enter “lightweight casual running shoes” as the query via the search field of the search user interface. The first computing device 102A, via the client application, may then transmit the query over the network 104 to the search platform 112.

[0111] At step 604, the method 600 may include determining and ranking a subset of items from the plurality of items stored in the item data store 126 that are responsive to the query. In some examples, the query may be interpreted to, among other things, identify or extract an item category associated with the query. An example item category may include a child category and / or a parent category of the hierarchical structure of the item data store 126 associated with the query. For example, one or more natural language processing techniques may be applied to identify or extract an item category associated with the query. In other examples, rule-based logic leveraging the hierarchical structure of the item data store 126 may be applied to identify or extract the item category associated with the query.Continuing with the above example where the query of “lightweight casual running shoes” is received, the item category identified may include a “running shoes” child category and / or a “shoes” parent category.

[0112] In further examples, the query may be tokenized and transformed into a query vector. For example, the query may be comprised of text that is converted (e.g., by dividing the text) into a plurality of tokens using sentence, word, sub-word, and / or character-based tokenization, among other tokenization techniques known in the art. The query vector may be generated based on the tokens, where the query vector mathematically or numerically represents the meaning of and / or relationship among the tokens. For example, the tokens may be parsed and interpreted to identify attributes of the item being searched for in the query.

[0113] To determine the subset of items responsive to the query, the search platform 112 may query the item data store 126 using the query vector (and optionally the item category for a more contextualized search). As discussed in detail with reference to FIG. 1 , the item data store 126 may be a vector-based database configured to store the items as vector representations of the items (e.g., as item vectors) in a vector space. A search of the vector space (e.g., a multi-dimensional graph search) of the item data store 126 may be performed using the query vector to identify a plurality of item vectors that are within a predefined distance of the queryAttorney Docket No: 00273-0015-00304vector in the vector space. The item vectors being within the predefined distance of the query vector indicate a similarity in one or more attributes of items that the item vectors are representing with the attributes of the item being queried for, and thus a relevancy or responsiveness of those items to the query. Resultantly, the item vectors identified may represent the items determined to be responsive to the query.

[0114] The subset of items determined to be responsive to the query may then be ranked. In some examples, a ranking algorithm may process a plurality of item features to determine a rank score for each item of the subset. Example features may include the conversion-based feature (e.g., determined in a same or similar manner as the conversion-based features is determined for the score determined for targeted item matching) and a similarity feature. The similarity feature may represent a word similarity between the text of the query and the text and / or image-based item information that is stored in association with the item in the item data store 126. A value for the similarity feature may be determined using cosine similarity. For example, for each of the items in the subset determined to be responsive to the query, a cosine of an angle between the item vector representing the item (e.g., the item vector including the text and / or image-based item information) and the query vector (e.g., representing the query text) in the vector space may be determined to yield a cosine similarity value for the respective item. A smaller cosine similarity value indicates increased similarity of the item vector to the query vector, and thus a likely higher relevance or responsiveness of the respective item to the query.

[0115] In addition to the conversion-based feature and similarity feature, other example features that are obtained and used for determining the rank score may include a commission-based feature (e.g., a commission rate, a commission amount, etc.), a text property-based relevancy feature (e.g., a best matching score estimating a relevance of the item to the query based on text properties of the item), and / or an interaction-based relevancy feature (e.g., an estimated relevance of the item based on content creator and / or consumer search interactions with the item over time), among other similar features. Each of the features may be weighted and combined with one another using a function to determine the rank score.

[0116] For each item in the subset of items determined to be responsive to the query, the EPC value, the similarity value, and any other feature values obtained for the item may be separately normalized and weighted to obtain weighted featureAttorney Docket No: 00273-0015-00304scores for the features that are combined with one another to generate the rank score. An example formula for rank score is included below:Rank Score = Sum(Feature scorej * weightj).

[0117] Continuing with the illustrative example, a first item “lightweight casual running shoe” sold by a first merchant and a second item “lightweight running shoe” sold by a second merchant may be two example items determined to be responsive to the query of “lightweight casual running shoes”. The first merchant is a larger, well-known merchant in comparison to the second merchant. Based on monitoring, 100 purchases of the first item resulted from 1 ,000 clicks to access the resource for the first item, whereas 20 purchases of the second item resulted from 100 clicks to access the resource for the second item. Therefore, although the second item has lower sales volume, the second item has a higher conversion-rate (e.g., higher likelihood of user interest converting into a purchase), which will be reflected in the EPC value. For example, the EPC value determined for the first item and the second item may be 0.8 and 1.6, respectively, and the normalized and weighted EPC values (e.g., EPC feature scores) for the first item and the second item may be 1.6 and 3.2, respectively, which is indicative of a greater amount of commission per clickout associated with the second item (e.g., a higher earning opportunity).

[0118] The normalized and weighted similarity values (e.g., similarity feature scores) for the first item and the second item may be 6 and 4.8 respectively, which is indicative of the closer similarity of item information of the first item to the text of the query (e.g., closer word similarity). When the EPC feature scores and similarity feature scores are combined (e.g., summed), the rank scores for the first item and the second item may be 7.6 and 8, respectively. Therefore, the rank score for the second item is higher than the first item, even though it is less similar to the query than the first item. This demonstrates that smaller or mid-sized merchants having lower sales volume, may nonetheless achieve visibility in search results when items sold by these merchants have high EPC values indicative of their items meeting user preferences (e.g., as quantified by higher conversion rates / traffic efficiency on the merchants’ resources for the items) and commission rates meeting expectations of content creators. In this illustrative example, only the conversion-based feature and similarity feature are discussed as example features used as part of the rank score. However, in other examples, values for multiple other features may also beAttorney Docket No: 00273-0015-00304determined, normalized, and weighted for use in combination with the EPC and similarity feature sores in the rank score.

[0119] In some examples, the search platform 112 may then sort (e.g., order) the items within the subset based on the rank score determined for each item. For example, the items may be ordered from a highest rank score to a lowest rank score. Continuing the above example, the second item may be ordered higher than the first item.

[0120] In some examples, the subset of items determined to be responsive to the query at step 606 may include a targeted item. Continuing with the illustrative example, the second item “lightweight running shoe” sold by the second merchant may be a targeted item. In such examples, at step 606, the method 600 may include identifying, using the data structure, the item match set for the targeted item. For example, the search platform 112 may query the item match data store 136 using the targeted item to identify the data structure associated with the item match set for the targeted item. The data structure comprises various information, including the determined plurality of scores, for each of the targeted item and the matching items included in the item match set. Continuing with the illustrative example, the matching items of the second item “lightweight running shoe” sold by the second merchant may include the exact same “lightweight running shoe” that is sold by different merchants (e.g., same brand and style of shoe but sold by a third merchant, fourth merchant, etc.). The respective scores of the targeted item and matching items reflect or signal earning opportunity as a function of the EPC value and the current commission amount for the items (e.g., features dynamically driven by merchant-defined prices and commission rates, as well as merchant conversion rates, etc.).

[0121] Step 606 may be repeated for each targeted item included in the response. In some examples, step 606 may alternatively be performed at a later step in the method 600 (e.g., after step 610), as described in more detail below.

[0122] At step 608, the method 600 may include generating and transmitting, to the first computing device 102A, computer-executable instructions configured to cause the first computing device 102A to construct and display a user interface that displays, as results of the item search, at least a portion of the ranked subset of items. The user interface may be a search results user interface of the client application of the ecommerce platform 108 running on the first computing device 102A.Attorney Docket No: 00273-0015-00304

[0123] The instructions may include information associated with how to visually display the ranked subset of items within the user interface. For example, the instructions may include a position or arrangement of the items relative to one another within the user interface based on the ranking or ordering. To provide an illustrative example, the arrangement may indicate that the items are to be displayed row by row, with / V items per row listed in the order provided from left to right.Additionally, the instructions may include a style or format for presentation of the items, along with any images and / or text associated with the items (e.g., obtained from the item data store 126) to display in accordance with the style or format.

[0124] The portion of the ranked subset of items displayed may include the targeted item (or multiple targeted items). In such examples, the matching items in the item match set for the targeted item may be displayed in association with the targeted item based on the plurality of scores. Resultantly, the content creator is provided not only the most relevant items to the query (e.g., the first item, the second item, etc.), but also exact matching items to any targeted items included in the results as alternative options that may potentially provide an improved or more optimal earning opportunity. For example, the matching items may be ordered or arranged for display relative to one another and / or the targeted item based on the scores (e.g., from highest score to lowest score indicative of highest to lowest earning opportunity). In some examples, the score corresponding to each of the targeted and matching items itself may be displayed.

[0125] In other examples, the matching items may be ordered or arranged for display relative to one another and / or the targeted item further based on preferences of the content creator. For example, the initial order or arrangement based on the scores may be adjusted or updated based on the content creator preferences.Example preferences may include preferred retailers (e.g., merchants), preferred brands, or preferred designers, as well as retailers to be avoided, brands to be avoided, or designers to be avoided, among other similar preference types. As one non-limiting example, a highest score may be for a matching item associated with a large retailer. However, based on a preference of the content creator to not promote items from large retailers, the matching item with the highest score may be moved to a lower priority position in the display.

[0126] In some aspects, the matching items may be displayed in a manner to signal to a user that the matching items are optional alternatives to the targeted itemAttorney Docket No: 00273-0015-00304(e.g., may be displayed in a separate or pop-up window, beneath or nested within the targeted item, etc.). In some examples, when step 606 may alternatively be performed after step 610, the matching items may not yet have been identified and thus will not be displayed as part of the initial search results.

[0127] At step 610, the method 600 may include receiving, from the first computing device 102A, an indication of a selection of the targeted item from at least the portion of the ranked subset of items displayed via the user interface.

[0128] In some examples, step 606 may be performed only after the targeted item has been selected. In such examples, a separate user interface (e.g., a pop-up window) may be presented to the user to display the matching items in association with the targeted item based on the plurality of scores. In some examples, the matching items may be indicated as options and / or, dependent on the relative scores, recommended matching items, for alternative selection. Waiting to identify the item match set for use in displaying the matching items until the targeted item is selected (e.g., until user interest in the targeted item is indicated) may help to conserve processing resources, particularly when multiple targeted items are included in the subset of responsive items to the query.

[0129] The targeted item selected may be included in content generated and published by the content creator via the content creation platform 114 to promote or otherwise recommend (e.g., for monetization purposes) the target item. For example, the content creator may select the higher ranked second item as the pair of lightweight casual running shoes for inclusion in the content.

[0130] At step 612, the method 600 may include associating a resource for the targeted item with the content. The resource may be a website, web page, or other similar page of the merchant selling the targeted item. The resource may be saved in association with the content within the content data store 128. The resource may be subsequently referenced to provide access to the resource and / or replaced with an alternative resource in response to detecting a triggering event, as described in more detail with reference to the event-based item replacement process of FIG. 7.

[0131] Accordingly, certain embodiments may be performed for query processing. The method 600 described above is provided merely as an example, and may include additional, fewer, different, or differently arranged steps than depicted in FIG.6 describing the method 600 and / or portions thereof.Attorney Docket No: 00273-0015-00304

[0132] FIG. 7 is a flow chart showing an example process 700 for event-based item replacement that implements the targeted item matching, according to some embodiments of the disclosure. The process 700 may be performed by one or more of the server-side systems 106, such as one or more components of the ecommerce platform 108, including the targeted item matching platform 120.

[0133] At step 702, the process 700 may include detecting an event associated with a targeted item. For illustrative purposes, the targeted item for which the event is detected may be the targeted item that was selected for inclusion in the content, and the resource for the targeted item was associated with the content at steps 610 and 612 of the method 600 described with reference to FIG. 6. However, events may be detected for any of the targeted items.

[0134] In some examples, the event may be a detected change associated with the targeted item that affects and / or alters an earning opportunity associated with the targeted item. Example changes may be associated with an availability of the targeted item for sale by the merchant. For example, a limited remaining availability of the targeted item for purchase (e.g., low stock), a temporary unavailability of the targeted item with continued purchase option (e.g., backorder), a temporary unavailability of the targeted item for purchase (e.g., out of stock), and / or a permanent unavailability of the targeted item (e.g., sell out). Other example changes associated with the targeted item may include a price drop and / or or a commission rate drop above one or more predetermined thresholds. In some aspects, one or more of these type of changes may be detected based on updated information provided by the merchant that are stored in association with the targeted item in the item data store 126. Additionally or alternatively, one or more of these type of changes may be detected based on a periodic scraping of the resource for the targeted item for this information.

[0135] In some examples, a merchant resource for the targeted item may be evaluated based on the above factors (e.g., item availability, price, commission rate, etc.) and may be classified as a healthy v. unhealthy resource or link. Classification of the merchant resource for the targeted item as an unhealthy resource or link may be an example event to trigger replacement.

[0136] In other examples, the event may be a detected inactivity associated with the targeted item in the item data store 126 exceeding a predefined time period. For example, the event may be triggered based on an elapse of the predefined timeAttorney Docket No: 00273-0015-00304period following a previous update associated with the targeted item in the item data store 126. Such inactivity may be indicative of a staleness (e.g., a declining user interest of and thus a declining earning opportunity of) the targeted item. To provide a non-limiting example, the predefined time period may be a 30 day period. If no changes or updates have been made in association with the targeted item in the item data store 126 (e.g., no changes to price, commission rate, availability and / or other attributes) in the past 30 days, the event may be triggered. In response to the detected inactivity, the targeted item matching process described above with reference to FIG. 3 may be repeated to reevaluate the targeted item and potentially update the item match set, including corresponding scores for the targeted item and the same or different matching items identified (if the targeted item is continued to be included in the portion of items determined for targeting).

[0137] At step 704, the process 700 may include determining to replace the resource for the targeted item associated with content including the targeted item based on the detected event. As described in detail above, the detected event may be indicative of an affected or altered earning opportunity associated with the targeted item. Therefore, it may be advantageous to replace the targeted item with one of the matching items to the targeted item within the item match set that provides a better or more optimal earning opportunity at that point in time.Specifically, rather than continuing to have consumers devices’ routed to the resource for the targeted item when they interact with the content, the resource of the targeted item associated with the content may be replaced with a new resource of the one of the matching items to instead enable an automatic routing of the consumers devices’ to the new resource, as described in detail in the following steps.

[0138] At step 706, the process 700 may include, using the data structure, identifying, from the item match set for the targeted item, a recommended matching item based on the plurality of scores. For example, the item match data store 136 may be queried using the targeted item to identify the data structure associated with the item match set for the targeted item. The data structure comprises various information, including the determined plurality of scores, for each of the targeted item and the matching items included in the item match set. The scores may be compared among one another, and a respective item from the item match set having the highest score (e.g., indicative of the highest earning opportunity) may beAttorney Docket No: 00273-0015-00304identified as a recommended item. As described in the process 700, one of the matching items may be identified as the recommended matching item. However, in other examples, the targeted item may have the highest score, and dependent on the type of event detected, the resource for the targeted item may remain associated with the content. Additionally, while a highest score is the criteria described herein for selecting the recommended item, additional or alternative criteria may be used.

[0139] In addition to, or alternatively from, the plurality of scores, the recommended matching item may be identified based on the preferences of the content creator. As one non-limiting example, a highest score may be for an item associated with a specific retailer. However, based on a preference of the content creator to not promote items from that specific retailer (e.g., the retailer is indicated as a retailer to avoid by the content creator), another item from the item match set (e.g., an item with a second highest score) may be selected as the recommended item despite having a lower score. As another example, a mid-score item within the item match set may be associated with a specific retailer indicated as a preferred retailer of the content creator, and thus may be selected as the recommended matching item despite not having the highest score.

[0140] At step 708, the process 700 may include replacing the resource for the targeted item associated with the content with a new resource for the recommended matching item. The recommended matching item may be sold by a different merchant than the merchant of the targeted item. The new resource of the different merchant may now be stored (instead of the resource of the merchant) in association with the content in the content data store 128.

[0141] In some examples, the resource replacement may occur automatically upon determining the recommended matching item. In other examples, a notification may be sent to the first computing device 102A of content creator to provide the content creator an option to replace, and the replacement may be performed upon a response to the notification selecting the replacement option. In some examples, the notification may also provide the content creator options to update the content based on the resource replacement. For example, the content creator may be enabled to modify a representative image of the item (e.g., a cover photo) or an item description, among other similar details, to reflect the recommended matching item within the content. The modified content may be published and stored in the content data store 128 (e.g., may be stored as a current version among one or moreAttorney Docket No: 00273-0015-00304previous versions and / or replace a previous version of the content if only one version is stored). The content creator may be enabled to perform such modifications to the content when the resource replacement occurs automatically as well.

[0142] As a result of the replacement, when an interaction with the content is detected (e.g., via the second computing device 102B), the second computing device 102B is caused to access the new resource for the recommended matching item now associated with the content. For example, the second computing device 102B may be routed (e.g., via a web browser or application) to the new resource to enable access thereto.

[0143] As an illustrative example, a consumer or a follower of the content creator that generated the content may utilize the second computing device 102B to access the content. In one example, the consumer or follower may utilize a search feature of the client application associated with the ecommerce platform 108 that is running on the second computing device 102B to identify and access the content. In other examples, the consumer or the follower may access the content through another platform, such as a social media platform, over which the content generated via the content creation platform was re-published or shared, for example.

[0144] The interaction detected may be a selection associated with the targeted item that is included in the content (e.g., a selection of an image of the targeted item, a description of the targeted item, and / or a link for the targeted item). The interaction may be indicative of user interest (e.g., an intent of the consumer or follower to view further information about the item and / or purchase the item).

[0145] Accordingly, certain embodiments may be performed for event-based item replacement. The process 700 described above is provided merely as an example, and may include additional, fewer, different, or differently arranged steps than depicted in FIG. 6 describing the process 700 and / or portions thereof.

[0146] FIG. 8 shows an implementation of a computer system 800 that executes techniques presented herein, including processes or operations depicted in, or described with respect to, FIGs. 2-7, according to some embodiments of the disclosure. For example, the computer system 800 may be configured as one of the computing devices 102, one of the server-side systems 106, or another device according to exemplary embodiments of this disclosure. In various embodiments, any of the systems herein may be a computer system 800 including, e.g., a data communication interface 820 for packet data communication. The computer systemAttorney Docket No: 00273-0015-00304800 may communicate with one or more other computer systems 800 using the electronic network 825. The electronic network 825 may include a wired or wireless network similar to the network 104 depicted in FIG 1.

[0147] The computer system 800 also may include a central processing unit (“CPU”), in the form of one or more processors 802, for executing program instructions 824. The program instructions 824 may include instructions for running one or more operations of the respective device or system. The computer system 800 may include an internal communication bus 808, and a drive unit 806 (such as read-only memory (ROM), hard disk drive (HDD), solid-state disk drive (SDD), etc.) that may store data on a computer readable medium 822, although the computer system 800 may receive programming and data via network communications. The computer system 800 may also have a memory 804 (such as random access memory (RAM)) storing instructions 824 for executing techniques presented herein, although the instructions 824 may be stored temporarily or permanently within other modules of computer system 800 (e.g., processor 802 or computer readable medium 822). The computer system 800 also may include user input and output ports 812 or a display 810 to connect with input and output devices such as keyboards, mice, touchscreens, monitors, displays, etc. The various system functions may be implemented in a distributed fashion on a number of similar platforms, to distribute the processing load. Alternatively, the systems may be implemented by appropriate programming of one computer hardware platform.

[0148] Program aspects of the technology may be thought of as “products” or “articles of manufacture” typically in the form of executable code or associated data that is carried on or embodied in a type of machine-readable medium. “Storage” type media include any or all of the tangible memory of the computers, processors or the like, or associated modules thereof, such as various semiconductor memories, tape drives, disk drives and the like, which may provide non-transitory storage at any time for the software programming. All or portions of the software may at times be communicated through the Internet or various other telecommunication networks. Such communications, e.g., may enable loading of the software from one computer or processor into another. Thus, another type of media that may bear the software elements includes optical, electrical and electromagnetic waves, such as used across physical interfaces between local devices, through wired and optical landline networks and over various air-links. The physical elements that carry such waves,Attorney Docket No: 00273-0015-00304such as wired or wireless links, optical links, or the like, also may be considered as media bearing the software. As used herein, unless restricted to non-transitory, tangible “storage” media, terms such as computer or machine “readable medium” refer to any medium that participates in providing instructions to a processor for execution.

[0149] While the disclosed methods, devices, and systems are described with exemplary reference to transmitting data, it should be appreciated that the disclosed embodiments may be applicable to any environment, such as a desktop or laptop computer, an automobile entertainment system, a home entertainment system, etc. Also, the disclosed embodiments may be applicable to any type of Internet protocol.

[0150] It should be understood that embodiments in this disclosure are exemplary only, and that other embodiments may include various combinations of features from other embodiments, as well as additional or fewer features.

[0151] It should be appreciated that in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. This method of disclosure, however, is not to be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the Detailed Description are hereby expressly incorporated into this Detailed Description, with each claim standing on its own as a separate embodiment of this invention.

[0152] Furthermore, while some embodiments described herein include some but not other features included in other embodiments, combinations of features of different embodiments are meant to be within the scope of the invention, and form different embodiments, as would be understood by those skilled in the art. For example, in the following claims, any of the claimed embodiments can be used in any combination.

[0153] Thus, while certain embodiments have been described, those skilled in the art will recognize that other and further modifications may be made thereto without departing from the spirit of the invention, and it is intended to claim all such changes and modifications as falling within the scope of the invention. For example,Attorney Docket No: 00273-0015-00304functionality may be added or deleted from the block diagrams and operations may be interchanged among functional blocks. Steps may be added or deleted to methods described within the scope of the present invention.

[0154] The above disclosed subject matter is to be considered illustrative, and not restrictive, and the appended claims are intended to cover all such modifications, enhancements, and other implementations, which fall within the true spirit and scope of the present disclosure. Thus, to the maximum extent allowed by law, the scope of the present disclosure is to be determined by the broadest permissible interpretation of the following claims and their equivalents, and shall not be restricted or limited by the foregoing detailed description. While various implementations of the disclosure have been described, it will be apparent to those of ordinary skill in the art that many more implementations are possible within the scope of the disclosure. Accordingly, the disclosure is not to be restricted except in light of the attached claims and their equivalents.

Claims

Attorney Docket No: 00273-0015-00304CLAIMSWhat is claimed is:

1. A computer-implemented method for targeted item matching, the method comprising:determining, from a plurality of items stored in an item data store, a portion of the plurality of items to target for item matching;for a targeted item in the portion, identifying a plurality of matching items to the targeted item for inclusion in an item match set for the targeted item;determining a plurality of scores for the item match set, including a score for each of the targeted item and the plurality of matching items in the item match set based on a plurality of features for the respective item;generating and storing a data structure associated with the item match set for the targeted item, the data structure including the plurality of scores;detecting an event associated with the targeted item; andbased on the detected event:identifying content including the targeted item, wherein a resource for the targeted item is associated with the content;using the data structure, identifying, from the item match set for the targeted item, a recommended matching item based on the plurality of scores; andreplacing the resource for the targeted item associated with the content with a new resource for the recommended matching item.

2. The computer-implemented method of claim 1 , wherein determining the portion of the plurality of items to target for item matching comprises:tracking a plurality of signals indicative of user intent as one or more of the plurality of items are interacted with via one or more user computing devices over a network; anddetermining the portion of the plurality of items by processing the plurality of signals, wherein each signal of the plurality of signals is weighted based on a signal type.Attorney Docket No: 00273-0015-003043. The computer-implemented method of claim 2, wherein the signal type includes one or more of: a generated item link, a favorited item, an item selection from a set of search results, or brand popularity.

4. The computer-implemented method of claim 1 , wherein the item data store is configured to store item attributes for each of the plurality of items, and identifying the plurality of matching items to the targeted item comprises:obtaining, from the item data store, the item attributes for the targeted item and the item attributes for a candidate matching item from the plurality of items; comparing the item attributes for the targeted item to the item attributes for the candidate matching item; anddetermining the candidate matching item as one of the plurality of matching items based on the comparing indicating a similarity above a similarity threshold.

5. The computer-implemented method of claim 4, wherein the item attributes include one or more of: a title, a description, a price, a brand, a merchant, or an image.

6. The computer-implemented method of claim 1 , wherein the plurality of features include a conversion-based feature, and the conversion-based feature for the respective item is an earnings per click (EPC) feature based on a merchant conversion rate associated with a resource for the respective item, a number of actual item orders, an average item order value, and an average item commission rate.

7. The computer-implemented method of claim 1 , wherein the plurality of features include a current commission amount.

8. The computer-implemented method of claim 1 , wherein the event is one or more of: a low stock associated with the targeted item, a backorder associated with the targeted item, an out of stock associated with the targeted item, a sell out associated with the targeted item, a price drop associated with the targeted item, a commission rate drop associated with the targeted item, or an inactivity associated with the targeted item in the item data store exceeding a predetermined time period.Attorney Docket No: 00273-0015-003049. The computer-implemented method of claim 1 , wherein the plurality of features include a conversion-based feature, and the method further comprises: determining a first subset of items, from the plurality of items stored in the item data store, for which one or more actions performed in association with each item in the first subset have been identified based on a monitoring of behavior associated with the plurality of items as a plurality of resources for the plurality of items are accessed over a network; andfor each item in the first subset, determining, based on the one or more actions, and storing, in a feature data store, one or more conversion-based features at one or more of a plurality of hierarchical levels corresponding to a hierarchical structure of the item data store, wherein the hierarchical structure includes the plurality of items and one or more of a plurality of child categories or a plurality of parent categories with which the plurality of items are associated.

10. The computer-implemented method of claim 9, wherein determining the plurality of scores for the item match set comprises:for each of the targeted item and the plurality of matching items in the item match set, determining a conversion-based feature at a lowest hierarchical level of the plurality of hierarchical levels available from the feature data store; and determining the score for each of the targeted item and the plurality of matching items in the item match set based on the plurality of features for the respective item, including the conversion-based feature.

11. The computer-implemented method of claim 9, further comprising:receiving, from a computing device, a query for an item search; determining and ranking, from the plurality of items stored in the item data store, a second subset of items responsive to the query, wherein the second subset of items include the targeted item;identifying, using the data structure, the item match set for the targeted item; generating and transmitting, to the computing device, computer-executable instructions configured to cause the computing device to construct and display a user interface that displays, as results of the item search, at least a portion of the ranked second subset of items including the targeted item, wherein the plurality ofAttorney Docket No: 00273-0015-00304matching items in the item match set are displayed in association with the targeted item based on the plurality of scores; andin response to receiving, from the computing device, an indication of a selection of the targeted item from at least the portion of the ranked second subset of items displayed via the user interface for inclusion in the content, associating the resource for the targeted item with the content.

12. A system for targeted item matching, the system comprising:one or more processors; andat least one memory storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations including:determining, from a plurality of items stored in an item data store, a portion of the plurality of items to target for item matching;for a targeted item in the portion, identifying a plurality of matching items to the targeted item for inclusion in an item match set for the targeted item;determining a plurality of scores for the item match set, including a score for each of the targeted item and the plurality of matching items in the item match set based on a plurality of features for the respective item;generating and storing a data structure associated with the item match set for the targeted item, the data structure including the plurality of scores; detecting an event associated with the targeted item; and based on the detected event:identifying content including the targeted item, wherein a resource for the targeted item is associated with the content;using the data structure, identifying, from the item match set for the targeted item, a recommended matching item based on the plurality of scores; andreplacing the resource for the targeted item associated with the content with a new resource for the recommended matching item.

13. The system of claim 12, wherein determining the portion of the plurality of items to target for item matching comprises:Attorney Docket No: 00273-0015-00304tracking a plurality of signals indicative of user intent as one or more of the plurality of items are interacted with via one or more user computing devices over a network; anddetermining the portion of the plurality of items by processing the plurality of signals, wherein each signal of the plurality of signals is weighted based on a signal type.

14. The system of claim 12, wherein the item data store is configured to store item attributes for each of the plurality of items, and identifying the plurality of matching items to the targeted item comprises:obtaining, from the item data store, the item attributes for the targeted item and the item attributes for a candidate matching item from the plurality of items; comparing the item attributes for the targeted item to the item attributes for the candidate matching item; anddetermining the candidate matching item as one of the plurality of matching items based on the comparing indicating a similarity above a similarity threshold.

15. The system of claim 12, wherein the plurality of features include a conversionbased feature, and the conversion-based feature for the respective item is an earnings per click (EPC) feature based on a merchant conversion rate associated with a resource for the respective item, a number of actual item orders, an average item order value, and an average item commission rate.

16. The system of claim 12, wherein the plurality of features include a current commission amount.

17. The system of claim 12, wherein the event is one or more of: a low stock associated with the targeted item, a backorder associated with the targeted item, an out of stock associated with the targeted item, a sell out associated with the targeted item, a price drop associated with the targeted item, a commission rate drop associated with the targeted item, or an inactivity associated with the targeted item in the item data store exceeding a predetermined time period.Attorney Docket No: 00273-0015-0030418. The system of claim 12, wherein the plurality of features include a conversionbased feature, and the operations further include:determining a first subset of items, from the plurality of items stored in the item data store, for which one or more actions performed in association with each item in the first subset have been identified based on a monitoring of behavior associated with the plurality of items as a plurality of resources for the plurality of items are accessed over a network; andfor each item in the first subset, determining, based on the one or more actions, and storing, in a feature data store, one or more conversion-based features at one or more of a plurality of hierarchical levels corresponding to a hierarchical structure of the item data store, wherein the hierarchical structure includes the plurality of items and one or more of a plurality of child categories or a plurality of parent categories with which the plurality of items are associated.

19. The system of claim 18, wherein determining the plurality of scores for the item match set comprises:for each of the targeted item and the plurality of matching items in the item match set, determining a conversion-based feature at a lowest hierarchical level of the plurality of hierarchical levels available from the feature data store; and determining the score for each of the targeted item and the plurality of matching items in the item match set based on the plurality of features for the respective item, including the conversion-based feature.

20. A non-transitory computer readable medium for targeted item matching, the non-transitory computer readable medium storing instructions which, when executed by one or more processors, cause the one or more processors to perform operations comprising:determining, from a plurality of items stored in an item data store, a portion of the plurality of items to target for item matching;for a targeted item in the portion, identifying a plurality of matching items to the targeted item for inclusion in an item match set for the targeted item;determining a plurality of scores for the item match set, including a score for each of the targeted item and the plurality of matching items in the item match set based on a plurality of features for the respective item;Attorney Docket No: 00273-0015-00304generating and storing a data structure associated with the item match set for the targeted item, the data structure including the plurality of scores;detecting an event associated with the targeted item; andbased on the detected event:identifying content including the targeted item, wherein a resource for the targeted item is associated with the content;using the data structure, identifying, from the item match set for the targeted item, a recommended matching item based on the plurality of scores; andreplacing the resource for the targeted item associated with the content with a new resource for the recommended matching item.

Images