Abstract: A network-based service is provided that utilizes a machine learning model (MLM) trained on a variety of data from disparate systems of a retailer to generate predictive guidance/parameters for a price markdown process. The predictive guidance includes a markdown prediction that indicates whether an item should or should not be marked down. For an item designated for markdown, the MLM also generates markdown parameters including a markdown level for the item and a quantity of the item to be marked down. The predictions of the MLM are optimized to reduce item shrink, reduce item spoilage, increase item sales, and increase item margins. The service can be integrated into existing retailer systems and services to provide optimal markdown instructions for perishable items that are data-driven and objective.

Abstract: Item vectors representing transaction contexts for items are mapped to multidimensional space. A request is received for an alternative to a given item from a resource. The multidimensional space is evaluated to identify closest candidate items to the given item based on the corresponding item vectors. An optimal candidate item is selected from the candidate items based on the request. The association between the given item and the optimal candidate item is injected within a process workflow associated with the resource.

Abstract: Item codes are mapped to multidimensional space as item vectors based on each item codes context relevant to other item codes in a product catalogue. A transaction history for a given customer is obtained and each item vector associated with a corresponding item purchase made by that customer is obtained. All item vectors per customer are summed to create an aggregated and single vector representing the purchase history of each customer. The aggregated customer-item vectors for the customers are plotted in the multidimensional space. The plotted customer-item vectors are then clustered into groupings based on their distances from one another in the multidimensional space; the groupings representing data-driven customer segments. The data-driven customer segments along with customer identifiers for the customers comprising each segment are provided as input to promotional engines and/or loyalty systems.

Abstract: An item price that is noted in a transaction for an item is identified. Any discount or item price override is identified for the transaction. A catalogue price for the item is obtained. Similar items associated with the item are determined based on mapped transaction contexts for the item and the similar items within a multidimensional space. Similar item prices are obtained and a median price for the item and similar items is calculated. A real-time price alert is sent to a resource that is associated with processing or handling the transaction when the item price, adjusted for any discount or item price override, deviates (above or below) from the median price by a threshold amount.

Abstract: A cross-entity and cross-retailer platform is provided that captures transaction data (and/or browser history data associated with online browser-based transactions), indexes, and stores the data in a cloud-accessible data store. A cloud service is provided that custom processes retailer and entity-defined workflows based on purchase transactions using the data store. The service discovers and updates trends and patterns associated with item sales for a given retailer or for a given entity across channels associated with in-store and online item sales. The trends and patterns are dynamically reported to the corresponding entity or the corresponding retailer. The entities may comprise manufacturers of an item, a supplier of the item, a distributor of the item, and a Consumer Packaging Goods (CPG) company of the item.

Abstract: A cross-entity and cross-retailer platform is provided that captures transaction data, indexes, and stores the data in a cloud-accessible data store. A service is provided that custom processes retailer and entity-defined workflows based on purchase transactions using the data store. The service also generates and maintains correlations between items, the transactions, geographical locations, retailers, and stores of the retailers. The correlations are provided as timely retail recommendations to the retailers and entities for suggested changes that are likely to optimize purchase transactions. The entities may comprise manufacturers of an item, a supplier of the item, a distributor of the item, and a Consumer Packaging Goods (CPG) company of the item.

Abstract: An image of a transaction receipt is obtained and optical character recognition (OCR) is processed on the image to identify a transaction identifier, a transaction total, and a handwritten amount associated with a transaction. When a confidence value associated with the OCR is above a threshold value, the handwritten amount is processed automatically through a transaction interface for the transaction to update the transaction total for the transaction.

Abstract: A real-time and dynamic media baseline for a transaction terminal is calculated when requested. The baseline is optimized to minimize media activities on the terminal while at the same time to minimize a total media volume level associated with each of the terminals for an enterprise as a whole. Real-time and dynamic conditions at the terminals are accounted for in any calculated baseline at the time the baseline is requested to optimally minimize the media activities of the corresponding terminal and to optimally minimize the total media volume level of the terminals as a whole.

Abstract: Historical transaction data, events, replenishment activities, and terminal statuses for terminals of a given store are obtained. The data is preprocessed to include calculated terminal traffic patterns, cash usage patterns, and elapsed time until a replenishment activity for a given interval of time on per terminal. A machine-learning model (MLM) is trained on the modified data to configure itself for predicting the elapsed time per terminal. Real-time transaction data, events, and terminal statuses for the terminals of the store are obtained at the interval of time, and current traffic and usage patterns are inserted into the real-time data and fed as input to the MLM. The MLM returns terminal identifiers for the terminals and corresponding elapsed periods of time until a projected replenishment activity is likely to occur. The terminal identifiers and corresponding elapsed periods of time are rendered within an interface accessible to the store for media management.

Abstract: A machine-learning algorithm is trained with features relevant to transaction exceptions, distributions of items in transaction mapped to product hierarchies, and operator data. The trained algorithm is trained to predict whether a given transaction requires a transaction exception for potential fraud or for management approval. The trained algorithm is then provided a set of in-progress input data for an in-progress transaction being processed on a transaction terminal. Output from the trained algorithm is used to determine whether the in-progress transaction is allowed to continue processing unabated or whether the in-progress transaction is to be suspended with a transaction exception requiring a manager override or security credential to continue processing.

Abstract: A prescriptive data model for stores of a retailer is maintained. The data model comprises clusters of benchmarks and benchmark values for successful stores and unsuccessful stores. A machine-learning model (MLM) is trained on the data model to predict Key Performance Indicator (KPI) values. An interface is provided that permits an end user to override a given current benchmark value with a changed value. The changed value along with unchanged current benchmark values are provided as input to the MLM and the MLM produces as output a set of current predicted KPI values. The set of current predicted KPI is rendered within the interface to the end user as a predicted impact the changed value will have on a given store or a given department of the given store.

Abstract: Metrics are captured from a variety of systems associated with stores of a retailer. Values for factors or benchmarks are calculated per store from their corresponding metrics. Each of the stores are labeled as successful or unsuccessful. Factors for which high values are correlated with successful stores and low values are correlated with unsuccessful stores are clustered together. Similarly, factors for which low values are correlated with successful stores and high values are correlated with unsuccessful stores are clustered together. A set of clustered factors associated with the success, or the failure of stores are reported to the retailer in a data model that also comprises the various degrees to which the various clusters of the factors relate to or correlate with both the successful stores and the unsuccessful stores. Prescriptive recommendations are derived from the data model to improve metrics associated with successful factors.

Abstract: Transaction items for a transaction are received during a transaction. Any non-barcoded items are identified and processed within a retailer-specific vector space to identify a most-similar barcoded item offered by a corresponding retailer to the non-barcoded item. The transaction items are revised to include the most-similar barcoded item as a replacement for the non-barcoded item. The revised transaction item list is used to identify a recommended item based on a segment-specific vector space associated with a segment assigned to the transaction. The recommended item is provided in real time to a transaction service that processes the transaction for delivery to a customer associated with the transaction.

Abstract: Item codes for items are mapped to multidimensional space as item vectors based on transaction contexts. Similarities between item codes are based on distances between the item codes within the multidimensional space. Substitute items for out-of-stock items are automatically identified based on the item similarities and based on collected feedback from transactions. The substitute items are provided in real time to customers during transactions, item picking services during item fulfillment, and shelf management services for item shelf stocking. In an embodiment, the substitute items are further determined based on a specific transaction history for a given customer and specific feedback collected for the given customer from the specific transaction history.

Abstract: Average processing times associated with components of a transactions are maintained as performance metrics for the transactions. Ratios associated with the average processing times are normalized based on calendar dates, days of week, and times of day. A given transaction is analyzed in view of the averages and ratios and a bottleneck is identified within a time slice of the given transaction. A type of bottleneck detected within the time slice is further identified based on identifiers associated with resources of the time slice and a pattern associated with one or more of the identifiers. The transaction, time slice, type of bottleneck, and resource identifiers are reported to a retailer associated with a transaction terminal that processed the transaction through an interface.

Abstract: An ecommerce application (app) or a transaction interface of a transaction terminal are enhanced to call a Purpose of Purchase (POP) service during a user session with the app or the terminal. The POP service integrates a question posed to the user regarding the user's POP for the session. A machine-learning model is trained on customer answers, transaction data, history data, and/or loyalty data to predict a customer's POP profile classification. When another customer is engaged in a session and fails to provide an answer, the model predicts a POP profile for the customer and the session. When the customer provides an answer, the appropriate POP profile that corresponds to the answer is assigned. During the session, the POP service provides the assigned or predicted POP profile to a recommendation service to use as a factor in making a product recommendation to the user during the session.

Abstract: A checkout state for different types of transaction terminals of a store as a whole is determined during a given analyzed interval of time based on transaction logs for transactions processed during the interval. Each transaction is classified into a basket size based on that transaction's total number of items during the interval. Metrics are recorded for the checkout state by basket size and terminal type relative to the total number of transactions within the interval for all terminal types. Trends are derived from the metrics between adjacent analyzed intervals of time. The metrics and trends are custom provided by checkout state per basket size of each terminal type over configurable periods of time within graphs, charts, and tables rendered within a user interface.

Abstract: An unrecognized item code is identified during a transaction at a transaction terminal. When the unrecognized code is determined to be associated with a typographical error, alternative corrected item codes are supplied to the transaction terminal to replace the unrecognized code. When the unrecognized code is determined to be missing from a product catalogue, an ordered list of most likely corrected item codes for the unrecognized code is provided to the transaction terminal for selection of one of the corrected item codes by an operator.

Abstract: A machine-learning algorithm is trained with features relevant to basket data for items of transactions. The trained algorithm is trained to predict whether a given transaction is more or less likely to be associated with theft being engaged in by a transaction operator for the transaction. The trained algorithm is then provided basket data for a given transaction and produces as output a theft prediction value. When the theft prediction value exceeds a configured threshold value, the transaction is flagged for manual intervention or the transaction is flagged for subsequent manual verification.

Abstract: The probabilities of transitioning between item states for a given item sequence of a given transaction are calculated and item non-fraud scores are calculated from the probabilities for each item of the given transaction. The item non-fraud scores for the items of the transaction are provided to a fraud-detection system for determining whether any of the item non-fraud scores is more likely or less likely to be associated with sweethearting fraud by a cashier that performed the transaction.