Predicting performance for providing an item

• US 10,460,332 B1
• Filed: 02/20/2014
• Issued: 10/29/2019
• Est. Priority Date: 02/20/2014
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Kujat, Aaron Drew Alexander, Saxena, Avinash Chandra, Mosby, Joel Christopher
• Primary Examiner(s): Feacher, Renae
• Application Number:
  US14/185,798
• Time in Patent Office: 2,077 Days
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities

Abstract

Techniques for predicted performance may be provided. For example, a computing service may be implemented to analyze information about past performances of merchants associated with providing items to consumers. Based on the analysis, the computing service may generate a performance prediction model. Further, the computing service may use the performance prediction model to determine a predicted performance for a particular merchant. Information about the predicted performance may be provided to various users including, for example, consumer and merchants.

Claims

1. A computer-implemented method, comprising

storing, by a computer system, performance metrics for item providers offering an item at an electronic marketplace, the performance metrics indicative of past performances of the item providers associated with providing units of the item to destinations and of past contexts and past conditions corresponding to the past performances, a past context associated with a particular source and a particular destination of one of the units of the item, a past condition associated with a past route between the particular source and the particular destination, the past context and the past condition stored in one or more databases;

generating, by the computer system, a performance prediction model based at least in part on the performance metrics, the performance prediction model comprising a machine learning algorithm trained to output, for an item provider of the item providers, expected performances associated with providing a unit of the item, the expected performances varying based at least in part on potential contexts and potential conditions, wherein the machine learning algorithm is trained based at least in part on a tuple that comprises a nested hierarchy of elements, wherein the elements comprise merchant identifiers, source locations, destination locations, contexts, and conditions, wherein, upon completion of training, the machine learning model outputs the expected performances corresponding to levels of the nested hierarchy;

receiving, by the computer system from a computing device of a consumer, a web search request for information about the item based at least in part on a web site hosted by the computer system and on an access of the computing device to the web site;

determining, by the computer system, a context associated with providing the unit of the item from a source location associated with the item provider to a destination location associated with the consumer, the destination location determined as a geo-location of the computing device based at least in part on the access to the web site;

determining, by the computer system from the one or more databases, a condition associated with a route for providing the unit of the item to the destination location;

generating, by the computer system based at least in part on input to the machine learning algorithm, a performance prediction associated with the item provider, the performance prediction comprising a predicted delivery time for providing the unit of the item and a likelihood for meeting the predicted delivery time, the input comprising an identifier of the provider, the context, and the condition, the performance prediction corresponding to a particular level of the nested hierarchy, wherein the particular level comprises the elements of the tuple;

sending, by the computer system to the computing device of the consumer, a web page of the web site for presentation in response to the web search request, the web page comprising the performance prediction, the identifier of the item provider, the predicted delivery time, and the likelihood for meeting the predicted delivery time;

initiating, by the computer system, the providing of the unit of the item via a carrier based at least in part on a selection of the item provider, the selection received from the computing device of the consumer based at least in part on the web page;

sending, by the computer system to the computing device of the consumer, notifications about progress of the providing of the unit of item, the notifications generated based at least in part on location tracking of the unit of the item, the location tracking comprising receiving a time and a location from a subscription service based at least in part on a scan of a barcode associated with the unit of the item, the subscription service available based at least in part on a subscription with the carrier, the notifications generated based at least in part on times and locations;

detecting, by the computer system based at least in part on the location tracking, a deviation between at least one of;

the context and an updated context, or the condition and an updated condition, the updated context and the updated condition associated with the providing of the unit of the item;

generating, by the computer system, an update to the performance prediction based at least in part on inputting the deviation to the machine learning algorithm, the update comprising at least one of;

an updated predicted delivery time or an updated likelihood for meeting the predicted delivery time; and

sending, by the computer system to the computing device of the consumer, a notification about the update to the performance prediction, the notification sent based at least in part on the deviation being detected and comprising a link to an updated web page, the updated web page comprising the updated predicted delivery time or the updated likelihood for meeting the predicted delivery time.

2. The computer-implemented method of claim 1, wherein the performance metrics are based at least in part on a historical delivery parameter of a second item provider, wherein the historical delivery parameter comprises one or more of:

a second source location associated with the second item provider, a second route between the second source location and the destination location associated with the consumer, a historical condition associated with the second route, or historical delivery timing information.

3. The computer-implemented method of claim 1, further comprising:

determining whether the item provider initiated the providing of the unit of the item to the consumer;

determining a change to the context and a change to the condition; and

updating the predicted delivery time and the likelihood for meeting the predicted delivery time based at least in part on a second input to the performance prediction model, the second input describing the change to the context and the change to the condition.

4. The computer-implemented method of claim 1, further comprising:

determining an actual time for when the unit of the item was received at the destination location associated with the consumer; and

updating the performance prediction model based at least in part on the actual time.

5. A computer-implemented method, comprising

identifying, by a computer system, metrics associated with offering an item by item providers through a service provider, the metrics indicative of past performances of the item providers associated with providing units of the item to destinations and of past contexts and past conditions corresponding to the past performances, a past context associated with a particular source and a particular destination of one of the units of the item, a past condition associated with a past route between the particular source and the particular destination;

generating a performance prediction model based at least in part on the metrics, the performance prediction model comprising a machine learning algorithm trained to output, for an item provider of the item providers, expected performances associated with providing a unit of the item, the expected performances varying based at least in part on potential contexts and potential conditions, wherein the machine learning algorithm is trained based at least in part on a tuple that comprises a nested hierarchy of elements, wherein the elements comprise merchant identifiers, source locations, destination locations, contexts, and conditions, wherein, upon completion of training, the machine learning model outputs the expected performances corresponding to levels of the nested hierarchy;

receiving, from a computing device of a user of the service provider, a request for information about the item, the request received as a web search request based at least in part on an access of the computing device to a web site of the service provider;

determining a context associated with providing the unit of the item from a source associated with the item provider to a destination associated with the user, the destination determined as a geo-location of the computing device based at least in part on the access to the web site;

determining a condition associated with a route for providing the unit of the item to the destination;

generating, based at least in part on input to the machine learning algorithm, a prediction of a performance for providing the unit of the item from the source associated with the item provider to the destination associated with the user, the input comprising an identifier of the provider, the context, and the condition, the prediction corresponding to a particular level of the nested hierarchy, wherein the particular level comprises the elements of the tuple;

providing an indication of the prediction and the identifier of the item provider to the computing device of the user, the indication causing the computing device to present the prediction and the identifier in a web page of the web site;

tracking progress of the providing of the unit of the item via a carrier based at least in part on location tracking of the unit of the item, the location tracking comprising receiving a time and a location from a subscription service based at least in part on a scan of a barcode associated with the unit of the item, the subscription service available based at least in part on a subscription with the carrier;

providing, to the computing device of the user, updates about the progress, the updates generated based at least in part on the location tracking;

detecting, based at least in part on the location tracking, a deviation between at least one of;

the context and an updated context, or the condition and an updated condition, the updated context and the updated condition associated with the providing of the unit of the item;

generating an update to the prediction based at least in part on inputting the deviation to the machine learning algorithm; and

providing, to the computing device of the user, the update to the prediction, the update is provided based at least in part on the deviation being detected, the providing of the update causing the computing device to present the update in an updated web page of the web site.

6. The computer-implemented method of claim 5, further comprising:

prior to receiving the request, generating a plurality of pre-computed predictions of performances for providing the unit of the item to locations of other users;

determining, based at least in part on the request, the destination associated with the user;

selecting, based at least in part on the destination, a particular pre-computed prediction from the plurality of pre-computed predictions; and

generating the prediction of the performance based at least in part on the particular pre-computed prediction.

7. The computer-implemented method of claim 6, wherein individual pre-computed predictions of the plurality of pre-computed predictions are generated based at least in part on one or more of:

a source location, a destination location, a lead time, a method for providing the item, the route for providing the unit of the item, or the condition associated with the route.

8. The computer-implemented method of claim 6, wherein individual pre-computed prediction of the plurality of pre-computed predictions are based at least in part on historical data associated with a second item provider, wherein the historical data comprises timing performances associated with providing the item to the other users.

9. The computer-implemented method of claim 5, wherein generating the prediction of the performance comprises:

determining the destination associated with the user based at least in part on the request;

determining the source based at least in part on the item provider;

providing the destination and the source as an input to the performance prediction model; and

setting an output of the performance prediction model as the prediction of the performance, wherein the output is based at least in part on the destination and the source location.

10. The computer-implemented method of claim 9, further comprising adjusting the output of the performance prediction model based at least in part on the route for providing the item and the condition associated with the route.

11. The computer-implemented method of claim 5, further comprising receiving, prior to receiving the request, data identifying that the item providers have provided items to the destinations, wherein the items are of a same type as a type of the item, wherein the past performances comprise timing performances associated with previous deliveries of the items to the destinations, and wherein the previous deliveries are associated with routes and route conditions.

12. The computer-implemented method of claim 11, wherein generating the prediction of the performance is based at least in part on the timing performances, the routes and the route conditions.

13. The computer-implemented method of claim 12, wherein generating the prediction of the performance further comprises:

determining the destination associated with the user based at least in part on the request for the information about the item;

determining the source of the unit of the item associated with the item provider;

determining a delivery method, the route, and a route condition associated with providing the unit of the item from the source to the destination; and

updating the prediction based at least in part on the delivery method, the route, and the condition.

14. A system of a service provider, comprising:

a memory that stores computer-executable instructions; and

a processor configured to access the memory and to execute the computer-executable instructions to collectively at least;

identify a plurality of item providers based at least in part on offers of an item by the plurality of item providers;

determine past performances of the plurality of item providers, the past performances associated with providing units of the item to a plurality of destinations within a region;

determine past contexts and past conditions corresponding to the past performances, a past context associated with a particular source and a particular destination of one of the units of the item, a past condition associated with a past route between the particular source and the particular destination;

generate a performance prediction model based at least in part on the past performances of the plurality of item providers, the past contexts, and the past conditions, the performance prediction model comprising a machine learning algorithm trained to output, for an item provider of the plurality of item providers, expected performances associated with providing a unit of the item within the region, the expected performances varying based at least in part on potential contexts and potential conditions, wherein the machine learning algorithm is trained based at least in part on a tuple that comprises a nested hierarchy of elements, wherein the elements comprise merchant identifiers, source locations, destination locations, contexts, and conditions, wherein, upon completion of training, the machine learning model outputs the expected performances corresponding to levels of the nested hierarchy;

at least in response to a request for information about the item from a computing device of a user received over a data network, identify the item provider, the request received from the computing device as a web search request based at least in part on an access of the computing device to a web site of the service provider;

determine a context associated with providing the item, the context comprising address information of a source associated with the unit of the item provider and of a destination that is within the region and that is associated with the user;

determine a condition associated with a route for providing the unit of the item to the destination associated with the user, the destination determined as a geo-location of the computing device based at least in part on the access to the web site;

generate, based at least in part on an input to the machine learning algorithm, an expected performance for the item provider given the address information of the context and given the condition, the input comprising an identifier of the item provider, the context, and the condition, the expected performance corresponding to a particular level of the nested hierarchy, wherein the particular level comprises the elements of the tuple;

provide, over the data network, an indication of the expected performance and the identifier of the item provider to the computing device of the user, the indication causing the computing device to present the expected performance and the identifier in a web page of the web site;

at least in response to an updated performance associated with the providing of the unit of the item from the source to the destination via a carrier, monitor a deviation between at least one of;

the context and an updated context, or the condition and an updated condition, the updated context and the updated condition associated with the updated performance, the monitoring comprising receiving a time and a location from a subscription service based at least in part on a scan of a barcode associated with the unit of the item, the subscription service available based at least in part on a subscription with the carrier;

generate an update to the expected performance based at least in part on the deviation being input to machine learning algorithm; and

provide, over the data network based at least in part on the deviation, a second indication of the update to the expected performance to the computing device of the user, the second indication causing the computing device to present the update to the expected performance in an updated web page of the web site.

15. The system of claim 14, wherein the expected performance comprises an expected time for when the unit of the item is expected to be provided to an item recipient, a likelihood for providing the unit of the item to the item recipient at the expected time, an expected route for providing the unit of the item to the item recipient, and an expected condition associated with the expected route.

16. The system of claim 15, wherein the computer-executable instructions further comprise computer-executable instructions that, when executed by the processor, cause the system to:

determine that the item provider initiated a providing of the unit of the item to the item recipient;

track the unit of the item before the unit of the item is received by the item recipient; and

update, based at least in part on the tracking, the expected time and the likelihood for providing the unit of the item at the expected time.

17. The system of claim 16, wherein the tracking comprises receiving one or more of:

data indicative of a scan at a location on the actual route of a tag associated with the unit of the item, location coordinates of a sensor associated with the unit of the item along an actual route, or data indicative of a load of the unit of the item on a carrier and location coordinates of the carrier along the actual route.

18. The system of claim 15, wherein the expected performance is generated based at least in part on a location of the item recipient and a resolution associated with the location, wherein the resolution is selected from a set of resolutions based at least in part on the past performances, and wherein the set of resolutions comprises one or more of:

a street address, location coordinates, a zip code, an area, a city, or a state.

22. The system of claim 14, wherein

the performance prediction model comprises a multiple input multiple output prediction model configured to output an expected performance per item provider, context, and condition based at least in part on the machine learning algorithm that regionalizes the past performances, the past contexts, and the past conditions in regions and finds a best fitness region out of the regions.

23. The system of claim 14, wherein the update to the expected performance is generated as a second output from the performance prediction model based at least in part on inputting the updated context and the updated condition to the performance prediction model.

24. The system of claim 14, wherein indications about expected performances are automatically provided based at least in part on an update rate that increases with a decrease of remaining time until an expected delivery time, and

wherein the second indication is provided automatically based at least in part on the deviation and independently of the update rate.

19. One or more non-transitory computer-readable storage media storing computer-executable instructions that, when executed by one or more computer systems of a service provider, configure the one or more computer systems to perform operations comprising:

identifying a plurality of providers of an item;

receiving unit providing data associated with the plurality of providers providing units of the item to recipients, the unit providing data indicative of past performances of the plurality of providers for providing the units of the item and of past contexts and past conditions corresponding to the past performances, a past context associated with a particular source and a particular destination of one of the units of the item, a past condition associated with a past route between the particular source and the particular destination;

generating a performance prediction model based at least in part on the unit providing data, the performance prediction model comprising a machine learning algorithm trained to output, for a provider of the plurality of providers, expected performances associated with providing a unit of the item, the expected performances varying based at least in part on potential contexts and potential conditions, wherein the machine learning algorithm is trained based at least in part on a tuple that comprises a nested hierarchy of elements, wherein the elements comprise merchant identifiers, source locations, destination locations, contexts, and conditions, wherein, upon completion of training, the machine learning model outputs the expected performances corresponding to levels of the nested hierarchy;

receiving, from a computing device of a user, a request for information about the item, the request received as a web search request based at least in part on an access of the computing device to a web site of the service provider;

determining a context associated with providing the unit of the item via a carrier from a source associated with the provider to a destination associated with a user, the destination determined as a geo-location of the computing device based at least in part on the access to the web site;

determining a condition associated with a route for providing the unit of the item to the destination;

determining, based at least in part on input to the machine learning algorithm, a performance metric associated with providing the unit of the item by the provider to the destination associated with the user, the input comprising an identifier of the provider, the context, and the condition, the performance metric corresponding to a particular level of the nested hierarchy, wherein the particular level comprises the elements of the tuple;

determining, based at least in part on the performance metric, an expected delivery time for providing the unit of the item to the destination associated with the user;

tracking progress of the providing of the unit of the item, the tracking comprising receiving a time and a location from a subscription service based at least in part on a scan of a barcode associated with the unit of the item, the subscription service available based at least in part on a subscription with the carrier; and

providing, to the computing device of the user, updates about the progress, the updates generated based at least in part on the tracking.

20. The one or more non-transitory computer-readable storage media of claim 19, wherein the operations further comprising instructing a computing device of the user to display an indication of the expected delivery time in response to a request for information about the item, wherein the indication comprises a comparison of the expected delivery time associated with the provider to a second expected delivery time associated with a second provider of the plurality of providers.

21. The one or more non-transitory computer-readable storage media of claim 19, wherein the expected delivery time is associated with a first method for providing the item by the provider to the user, and wherein the indication is based at least in part on a comparison of the expected delivery time of the provider to a second expected delivery time associated with a second method for providing the item to the user.