Amazon Technologies, Inc.

Jeffrey B Maurer

Devesh Mishra

Eric Mathew Mack

Jason W Murray

Deepak Bhatia

Jingchen Wu

Nicholas John Morren

Ali Sadighian

Tolga Han Seyhan

Robert Howard Sawers

Nicholas Deming Sherman

Edward Michael Leonard

Arjun Krishna Subramaniam

Onur Özkök

Zhihao Cen

Jingqiao Zhang

Gordon Mitchell Goetz

Minshu Xu

Chandrasekaran Murugan

Kip Christopher Larson

Techniques for generating a forecast associated with an item may be described. For example, the forecast may be generated based on simulations of supply and demand variables associated with the item over a planning horizon. The forecast may include at least one of: a labor forecast or a capacity forecast associated with inventorying units of the item in an inventory over the planning horizon. Further, real-time data associated with realizations of the supply and demand variables during the planning horizon may be monitored. At least a portion of the real-time data may be available from a management system. The management system may be configured to manage orders for the units of the item from the inventory. The forecast may be updated during the planning horizon based on the real-time data. The update may include changing the at least one of: the labor forecast or the capacity forecast.

Updating a forecast based on real-time data associated with an item

Techniques for computing a feature based on variables may be provided. For example, historical realizations associated with a first variable may be accessed. The first variable may be associated with an item. Realizations of the first variable may be based on one or more factors associated with the item. Historical realization of a second variable associated with the item may also be accessed. The historical realizations of the first and second variables may be analyzed to generate an expected realization of the first variable as a function of the second variable. The feature may be computed based on the generated function.

Efficiently computing a feature based on a plurality of variables

Techniques for determining a decision to acquire units of an item to be inventoried may be provided. For example, a demand for an item may be simulated to determine a consumption of a capacity for inventorying the item. A discrepancy between the consumption of the capacity and the capacity may be determined. An opportunity cost associated with the capacity may be updated based at least in part on determining that the discrepancy fails a convergence criterion. The opportunity cost may indicate a value associated with using the capacity. The consumption of the capacity may be simulated based at least in part on the updated opportunity cost. A resulting discrepancy may be determined. If the resulting discrepancy meets the convergence criterion, the decision to acquire the units of the item may be generated based at least in part on the updated opportunity cost.

Adaptive control of an item inventory plan

Systems and methods for supply risk detection are disclosed. In some embodiments, a method includes identifying fulfillment-related features corresponding to an item to be stored in inventory. The method also includes selecting a subset of the fulfillment-related features that is correlated with a supply constraint associated with the item. For example, in some cases the correlation is based on historical supply constraint data. The method further includes building a supply risk early detection model based, at least in part, upon the subset fulfillment-related features and evaluating the model to determine a probability that a third-party vendor will suffer a shortage of the item, as well as an expected duration of the shortage. Upon evaluation of the model, the method may include creating a purchasing plan for the item that takes into account the probability of the shortage and the expected duration of the shortage.

Supply risk detection

Systems and methods for predictive drop shipping are disclosed. In some embodiments, a method may include maintaining inventory of an item that is offered for sale to customers via a webpage. A ratio may be calculated between a number of views of the webpage and a number of units of the item purchased over a first time period when a fast shipping or delivery option is available, and over a second time period when such option is not available. The method may identify the item as a candidate for drop shipping in response to the ratio being larger during the first time period than during the second time period. Upon receipt of an order, the received order may be fulfilled via drop shipping, in some cases, despite the inventory being sufficient to fulfill the order, and despite a cost of fulfillment via drop shipping being higher than fulfillment from inventory.

Predictive drop shipping

Techniques and systems provide real-time return processing before a retailer receives a returned item, such as during a return process where the retailer (seller) and customer (buyer) are separated from one another (e.g., electronic commerce, etc.). The retailer may collect information about the order and items to be returned from the remotely located customer. The retailer may optionally confirm an identity of the customer such as by verifying the original payment type and/or payment instrument number. The customer may provide collateral in exchange for the real-time return of a refund or an exchange. The retailer may then provide shipping details, calculate the refund (if applicable), and finalize processing. If the retailer does not receive the item in the allotted time for the return (e.g., for ground transport, etc.), then the retailer may charge the customer a value of the item using the collateral.

Real-time return processing

Growing community of inventors

Port Orchard, WA, United States of America

Jeffrey B Maurer