Generating recommendations for adjustment actions using a trained model of an online system
A trained model is used to generate market adjustment recommendations for a retailer associated with an online system. Upon displaying an item to a user of the online system for replacing an originally requested item and collecting user's engagement data in relation to the replacement item, the online system accesses a market adjustment model that is trained to generate a score for the user indicative of an affinity of the user in relation to the replacement item and generate one or more market adjustment recommendations for the retailer. The online system applies the market adjustment model to generate, based on the engagement data, behavioral information of the user and/or contextual information associated with the user, the score for the user and the one or more market adjustment recommendations for the retailer. The online system provides the one or more market adjustment recommendations to a computing system associated with the retailer.
1 . A method, performed at a computer system comprising a processor and a computer-readable medium having instructions executed by the processor, comprising:
receiving, via a network and from a device associated with a user, a request for an item submitted by the user via a user interface of the device associated with the user;
responsive to receiving the request and initiated by the processor executing the instructions, searching through a database of the computer system by an order management module of the computer system to identify that the item is unavailable at an entity;
upon identifying that the item is unavailable, accessing an item replacement model, wherein the item replacement model is a machine-learning model trained to generate a replacement score for each candidate replacement item from a set of candidate replacement items;
applying, by an item replacement module of the computer system initiated by the processor executing the instructions, the item replacement model to user data associated with the user and one or more features for each candidate replacement item to generate the replacement score for each candidate replacement item;
selecting, using the replacement score for each candidate replacement item, a replacement item from the set of candidate replacement items;
causing, by a content presentation module of the computer system initiated by the processor executing the instructions, a user interface of the device associated with the user to display the replacement item for recommendation to the user and inclusion into a cart;
receiving, in real time via the network and from the device associated with the user, engagement data with information about an engagement of the user with the replacement item via the user interface;
collecting, by a data gathering module of the computer system initiated by the processor executing the instructions, chat data exchanged between the device associated with the user and a device associated with a picker who is servicing an order placed by the user via the user interface of the device associated with the user;
applying, by an adjustment module of the computer system initiated by the processor executing the instructions, a natural language processing to the chat data to extract activity information of the user;
accessing an adjustment model, wherein the adjustment model is a machine-learning model trained to:
generate a score for the user indicative of an affinity of the user in relation to the replacement item, and
generate a real time alert for the entity;
applying, by the adjustment module initiated by the processor executing the instructions, the adjustment model to the engagement data, the activity information of the user, and contextual information associated with the user to generate the score for the user and the real time alert for the entity; and
communicating, via the network and from the adjustment module, the real time alert to a computing system associated with the entity prompting the entity to perform one or more immediate actions.
2 . The method of claim 1 , further comprising:
clustering, using the score for the user, the user into a cluster of users, each user in the cluster of users associated with a likelihood of switching from a conversion of the item to a conversion of the replacement item greater than a threshold value; and
storing, in the database, information about the cluster of users.
3 . The method of claim 1 , further comprising:
generating the activity information of the user, wherein the activity information comprises at least one of: information about a brand of the replacement item converted by the user relative to a brand of the item, information about a type of the replacement item converted by the user relative to a type of the item, a discussion between the user and a picker during a fulfillment process associated with the replacement item, information about a future reordering of the replacement item by the user, or an appeasement request from the user associated with the item.
4 . The method of claim 1 , further comprising:
generating the contextual information associated with the user, wherein the contextual information comprises at least one of: demographic information about the user, a timestamp when the request for the item was made, content of the cart, or information about one or more other replacement items included in the cart.
5 . The method of claim 1 , wherein applying the adjustment model further comprises:
applying the adjustment model to the engagement data, the activity information of the user and the contextual information associated with the user to generate the real time alert including information about one or more reasons for the user declining to convert the replacement item.
6 . The method of claim 1 , wherein applying the adjustment model further comprises:
applying the adjustment model to the engagement data, the activity information of the user and the contextual information associated with the user to generate the real time alert including a recommendation for generating a coupon for incentivizing a conversion of the replacement item.
7 . The method of claim 1 , wherein applying the adjustment model further comprises:
applying the adjustment model to the engagement data, the activity information of the user and the contextual information associated with the user to generate the real time alert including a recommendation for at least one of branding or advertising a group of items including the replacement item which was converted by the user.
8 . The method of claim 1 , wherein applying the adjustment model further comprises:
applying the adjustment model to the engagement data, the activity information of the user and the contextual information associated with the user to generate the real time alert including a recommendation for the entity to combine the replacement item with one or more other items into a single item.
9 . The method of claim 1 , wherein applying the adjustment model further comprises:
applying the adjustment model to the engagement data, the activity information of the user and the contextual information associated with the user to generate the real time alert including an advertisement of the replacement item.
10 . The method of claim 1 , wherein applying the adjustment model further comprises:
applying the adjustment model to the engagement data, the activity information of the user and the contextual information associated with the user to generate the real time alert including a recommendation for adjusting stocking of at least one of the item or the replacement item.
11 . The method of claim 1 , further comprising:
collecting data with information about one or more actions conducted by the entity in response to the real time alert; and
re-training the adjustment model by updating, using the collected data, a set of parameters of the adjustment model.
12 . The method of claim 1 , wherein communicating the real time alert comprises:
communicating the real time alert to the computing system associated with the entity prompting the entity to perform the one or more immediate actions including at least one of adjusting a stock of one or more items, generating advertisements for the one or more items, providing discount coupons for conversion of the one or more items, or generating one or more new items.
13 . A computer program product comprising a non-transitory computer readable storage medium having instructions encoded thereon that, when executed by a processor of a computer system, cause the processor to perform steps comprising:
receiving, via a network and from a device associated with a user, a request for an item submitted by the user via a user interface of the device associated with the user;
responsive to receiving the request and initiated by the processor executing the instructions, searching through a database of the computer system by an order management module of the computer system to identify that the item is unavailable at an entity;
upon identifying that the item is unavailable, accessing an item replacement model, wherein the item replacement model is a machine-learning model trained to generate a replacement score for each candidate replacement item from a set of candidate replacement items;
applying, by an item replacement module of the computer system initiated by the processor executing the instructions, the item replacement model to user data associated with the user and one or more features for each candidate replacement item to generate the replacement score for each candidate replacement item;
selecting, using the replacement score for each candidate replacement item, a replacement item from the set of candidate replacement items;
causing, by a content presentation module of the computer system initiated by the processor executing the instructions, a user interface of the device associated with the user to display the replacement item for recommendation to the user and inclusion into a cart;
receiving, in real time via the network and from the device associated with the user, engagement data with information about an engagement of the user with the replacement item via the user interface;
collecting, by a data gathering module of the computer system initiated by the processor executing the instructions, chat data exchanged between the device associated with the user and a device associated with a picker who is servicing an order placed by the user via the user interface of the device associated with the user;
applying, by an adjustment module of the computer system initiated by the processor executing the instructions, a natural language processing to the chat data to extract activity information of the user;
accessing a an adjustment model, wherein the adjustment model is a machine-learning model trained to:
generate a score for the user indicative of an affinity of the user in relation to the replacement item, and
generate a real time alert for the entity;
applying, by the adjustment module initiated by the processor executing the instructions, the adjustment model to the engagement data, the activity information of the user, and contextual information associated with the user to generate the score for the user and the real time alert for the entity; and
communicating, via the network and from the adjustment module, the real time alert to a computing system associated with the entity prompting the entity to perform one or more immediate actions.
14 . The computer program product of claim 13 , wherein the instructions further cause the processor to perform steps comprising:
clustering, using the score for the user, the user into a cluster of users, each user in the cluster of users associated with a likelihood of switching from a conversion of the item to a conversion of the replacement item greater than a threshold value; and
storing, in the database, information about the cluster of users.
15 . The computer program product of claim 13 , wherein the instructions further cause the processor to perform steps comprising:
applying the adjustment model to the engagement data, the activity information of the user and the contextual information associated with the user to generate the real time alert including information about one or more reasons for the user declining to convert the replacement item.
16 . The computer program product of claim 13 , wherein the instructions further cause the processor to perform steps comprising:
applying the adjustment model to the engagement data, the activity information of the user and the contextual information associated with the user to generate the real time alert including a recommendation for the entity to combine the replacement item with one or more other items into a single item.
17 . The computer program product of claim 13 , wherein the instructions further cause the processor to perform steps comprising:
applying the adjustment model to the engagement data, the activity information of the user and the contextual information associated with the user to generate the real time alert including a recommendation for adjusting stocking of at least one of the item or the replacement item.
18 . The computer program product of claim 13 , wherein the instructions further cause the processor to perform steps comprising:
collecting data with information about one or more actions conducted by the entity in response to the real time alert; and
re-training the adjustment model by updating, using the collected data, a set of parameters of the adjustment model.
19 . A computer system comprising:
a processor; and
a non-transitory computer-readable storage medium having instructions that, when executed by the processor, cause the computer system to perform steps comprising:
receiving, via a network and from a device associated with a user, a request for an item submitted by the user via a user interface of the device associated with the user;
responsive to receiving the request and initiated by the processor executing the instructions, searching through a database of the computer system by an order management module of the computer system to identify that the item is unavailable at an entity;
upon identifying that the item is unavailable, accessing an item replacement model, wherein the item replacement model is a machine-learning model trained to generate a replacement score for each candidate replacement item from a set of candidate replacement items;
applying, by an item replacement module of the computer system initiated by the processor executing the instructions, the item replacement model to user data associated with the user and one or more features for each candidate replacement item to generate the replacement score for each candidate replacement item;
selecting, using the replacement score for each candidate replacement item, a replacement item from the set of candidate replacement items;
causing, by a content presentation module of the computer system initiated by the processor executing the instructions, a user interface of the device associated with the user to display the replacement item for recommendation to the user and inclusion into a cart;
receiving, in real time via the network and from the device associated with the user, engagement data with information about an engagement of the user with the replacement item via the user interface;
collecting, by a data gathering module of the computer system initiated by the processor executing the instructions, chat data exchanged between the device associated with the user and a device associated with a picker who is servicing an order placed by the user via the user interface of the device associated with the user;
applying, by an adjustment module of the computer system initiated by the processor executing the instructions, a natural language processing to the chat data to extract activity information of the user;
accessing an adjustment model, wherein the adjustment model is a machine-learning model trained to:
generate a score for the user indicative of an affinity of the user in relation to the replacement item, and
generate a real time alert for the entity;
applying, by the adjustment module initiated by the processor executing the instructions, the adjustment model to the engagement data, the activity information of the user, and contextual information associated with the user to generate the score for the user and the real time alert for the entity; and
communicating, via the network and from the adjustment module, the real time alert to a computing system associated with the entity prompting the entity to perform one or more immediate actions.