Oracle Retail AI Foundation Cloud Services Implementation Guide Release 22.2.301.0 F59895-02 |
|
![]() Previous |
![]() Next |
The Oracle Retail AI Foundation Cloud Services combines AI, machine learning, and decision science with data captured from Oracle Retail AI Foundation Cloud Services SaaS applications and third-party data. The unique property of these learning-enabled applications is that they detect trends, learn from results, and increase their accuracy the more they are used, adding massive amounts of contextual data to obtain a clearer picture on what motivates outcomes.
The Oracle Retail AI Foundation Cloud Services are comprised of the following Cloud Services:
Oracle Retail AI Foundation Platform Cloud Service
Oracle Retail AI Foundation Assortment and Space Optimization Cloud Service
Oracle Retail AI Foundation Promotion and Markdown Optimization Cloud Service
Oracle Retail AI Foundation Offer Optimization Cloud Service
The Oracle Retail AI Foundation Platform Cloud Service provides retailers with a data science toolkit that supports specific use-cases in planning, operations and execution and can be expanded to support broader retail uses. This includes Advanced Clustering, Customer Segmentation, Demand Transference, and Customer Decision Tree capabilities, as well as the recently introduced Attribute Extraction/Binning and Innovation Workbench capabilities.
The Oracle Retail AI Foundation Assortment and Space Optimization Cloud Service is used to determine the optimal selection and arrangement of products within stores by optimizing the product assortment and product placement on a virtual planogram.
The Oracle Retail AI Foundation Promotion and Markdown Optimization Cloud Service and Oracle Retail AI Foundation Offer Optimization Cloud Service reflect the evolution of our price and promotion optimization capabilities into an integrated life-cycle price optimization offering that enables retailers to engage their customers in an omnichannel environment while maximizing profits. The modular approach to offering life cycle pricing for promotions and markdowns separate from targeted offers enables retailers to innovate at the speed of their customer, while also accounting for the maturity of loyalty data necessary for targeted offers. The combined capabilities provide the following benefits to retailers:
Drive optimal promotion and pricing decisions for the entire product life cycle
Engage customers with targeted and contextual offers
Execute consistently, incorporating price and promotion plans, projected receipts, and returns.
Simplify decision-making through high-automation, exception-driven processes and what-if optimizations
Maximize accuracy and scale using artificial intelligence, machine learning, and optimization on the Oracle AI Foundation data science infrastructure
AI Retail Foundation Cloud Services are hosted in the Oracle Cloud with the security features inherent to Oracle technology and a robust data center classification, providing significant uptime. The Oracle Cloud team is responsible for installing, monitoring, patching, and upgrading retail software. Included in the service are continuous technical support, access to software feature enhancements, hardware upgrades, and disaster recovery. The Cloud Service model helps to free customer IT resources from the need to perform these tasks, giving retailers greater business agility to respond to changing technologies and to perform more value-added tasks focused on business processes and innovation.
Oracle Retail AI Foundation Software Cloud Service is acquired exclusively through a subscription service (SaaS) model. This shifts funding from a capital investment in software to an operational expense. Subscription-based pricing for retail applications offers flexibility and cost effectiveness.
Oracle Retail AI Foundation Cloud Services Advanced Clustering is an enterprise-specific clustering solution that leverages data mining capabilities to create store groupings at various product levels using multiple inputs. These inputs include performance data, product attributes, store attributes, third-party data such as demographic data as well as consumer segments. Using embedded science and automation capabilities, retailers are able to identify patterns within available data to create the necessary customer-centric and targeted clusters to be used by downstream assortment planning, allocation/replenishment, pricing, and promotions planning processes.
The store clustering process enables the creation, review, and approval of store clusters for downstream solution use, while also providing the ability to define and use clustering templates that can be specific to given product/location combinations.
Oracle Retail AI Foundation Cloud Services Advanced Clustering provides retailers with multiple clustering generation approaches and methods. These include the creation of simple, nested, and mixed attribute clusters using multiple methods, including those that support discrete and non-discrete attributes.
The types of clusters include the following:
Performance-based clusters (Sales Revenue, Sales Units, Gross Profit%, and so on)
Product attribute-based clusters (Brand, Color Family, Price Band, and so on)
Location attribute-based clusters (Store Size, Climate, Population Size, and so on)
Consumer profile-based clusters (Consumer Segment Profiles)
In addition to the above, users have the ability to create multiple clustering scenarios within a single cluster run. This enables the ability to leverage embedded rankings, scoring logic, as well as solution recommendations to define and approve the most appropriate clusters for use in intended planning or execution processes.
Users can access and review the following reporting information to drive decisions related to the clustering process.
Users can perform the following:
Determine what categories or merchandise classifications benefit most from clustering; determine the level of product or location hierarchy at which to cluster; and determine what attributes should be leveraged.
Analyze details related to the available cluster recommendations, assessing areas such as cluster composition, performance, and attributes, as well as store level scores (in relation to total clusters).
Review cluster scenario comparison features, visually assessing differences between the respective store cluster details.
Oracle Retail AI Foundation Assortment and Space Optimization Cloud Service can help maximize return on space, sales, revenue, and profits while improving customer satisfaction by optimizing assortments and facings to available space.
Leveraging key inputs such as optimization goals, demand transference science, and visual guidelines as well as inventory and replenishment factors, retailers are presented with a recommended shelf/fixture layout that can be leveraged in downstream execution processes.
Leveraging available fixture data, Oracle Retail AI Foundation Assortment and Space Optimization Cloud Service dynamically groups stores (known as space clusters) with common fixture dimensions, enabling retailers to optimize and refine their assortments at the planogram or store level.
Oracle Retail AI Foundation Assortment and Space Optimization Cloud Service provides retailers with the ability to conduct 'what-if' analysis by adjusting fixture lengths during an optimization run. The solution allows for a visual review, comparison, and validation of the results. This provides the ability to dynamically manage and assess the impacts of adding or removing fixture space from a particular store (or store group). The solution can help plan for and conduct store projects by recommending the re-allocation of space to planograms with an optimal return on space.
Prior to approving optimization results for downstream execution, retailers are able to review shelf previews, assessing variation from current or historical planograms as well as confirming that recommended results align with expectations. Updates to the respective shelf preview may be made in near real-time, with forecasted results updated in a real-time manner.
This section describes CDT Science and DT Science.
Oracle Retail AI Foundation Customer Decision Tree Science and Demand Transference Science enable retailers to create customer segment-specific decision trees using available transaction level data. These customer decision trees are specific to their customer segments and the respective geographies they operate within, and retailers are provided a better understanding of their most important products and product attributes. Using this detailed information, the retailer is able to effectively analyze assortment coverage and identify the duplication of item types as well as prevent the removal of core items that would cause a loss of customers.
Using Customer Decision Tree and Demand Transference Science, retailers can analyze a significant number of households (for example, in the thousands) to identify and rank which products are truly unique and whose sales are incremental, as opposed to those that can be discontinued because they are repetitive in nature and can be substituted with other products.
Understanding the incremental and substitutable sales associated with each item within an assortment, category managers can optimize the breadth of their assortments, as experienced by their customer's purchase preferences, with the optimal number of SKUs, given space constraints or financial goals.
Attribute Extraction (AE) is an enterprise-specific solution that uses machine learning to extract product attributes from free-form product description strings.
The application's embedded science and automation helps you to extract the attributes (such as brand, color, flavor, and so on) of each product in a particular category and to normalize the attribute values by correcting short forms, mis-spellings, and other inconsistencies. The product attributes can be used by Demand Transference, Customer Decision Trees, Advanced Clustering, and other retail applications that require product attributes in a structured format.
The AE application consists of the following tabs: Overview, Edit Labels, Annotation, Errors, Normalization, and Results. You use the Overview tab to select one of the previously added product categories or to add a new category. You use the Edit Labels to define category-specific attributes that you want to extract. In the Annotation and Errors tabs, you follow an iterative process to extract attributes and correct any mislabeled attributes. In the Normalization tab, you can use the embedded List of Values (LOV) or create your own LOV to standardize the attribute values. You use the Results tab to review and export the table of attributes.
Affinity Analysis (AA) lets retailers review the analysis about their customer market baskets. The system calculates association rules from the provided sales transaction data, which provides insight into customer shopping patterns. The process examines sales transaction data and identifies associations between different types of products. Such information can help a retailer understand that promoting one product is sufficient to help drive sales of another product, given the sales associations they exhibit.
The AIF solutions have a similar workflow and user interface (UI). The workflow lets users implement new science applications using similar techniques. For example, a retailer who uses Demand Transference Science and Customer Decision Tree Science may then be able to more easily learn and use Advanced Clustering and other aspects of demand modeling. This approach lowers the future total cost of implementing various science applications.
The Oracle Retail AI Foundation Cloud Services User Guide provides details about using these applications.
Two connection channels are used for interaction with the AIF:
Browser-Based
The application is accessed through a URL. The user is authenticated in order to gain access to the application. Access rights are controlled by the customer administrator through a Web application (Oracle Access Manager). Note that a role-based security policy is used. This allows the administrator to specify which applications and the tasks associated with those applications are accessible to which users.
Bulk Data Movement
A scheduled ETL extraction process must be used to extract the required data on the customer side and send it to the application through SFTP. Similarly, a schedule-based set of processes must be set up to process data coming in the opposite direction: from the application to on-premise. Note that this connection is still initiated from on-premise. Data is made available by the application for the download to on-premise location and processed further. All the necessary processes and credentials are set up during implementation.
Web Services
RASE web services are REST-based. RASE web services provide access to some of RASE application data and functionality but do not fully mirror the user interface or the export and import features of the backend. They are not a replacement for bulk data export, which must still be done at a scheduled time as part of batch processing. However, access to the configuration can be used during implementation and upgrade time, while AC and ASO export web services can serve as the means of obtaining incremental update data from a specified point in time (driven by a query parameter) as means of intra-day processing.
In the Cloud
The application processes are hooked up to a cloud scheduler to work in concert with what is sent (uploaded) from on-premise and what must be published to the outgoing SFTP directory for on-premise download.
The interactions with the application are illustrated in Figure 1-1.
Figure 1-1 Interacting with Oracle Retail AI Foundation Cloud Services
Here are the major steps.
Table 1-1 Interacting with Oracle Retail AI Foundation Cloud Services
Step # | Protocol | Direction | External/Internal | Description | Type of Data Sent |
---|---|---|---|---|---|
1 |
https/443 |
Inbound |
External (Internet) |
Used by customer to communicate with the application UI, OAM (login), OIM. |
Cluster, DT, CDT, space optimized results parameters |
2 |
sftp/2222 |
Inbound |
External (Internet) |
Data synchronization content (from RMS); Import/export files uploaded or downloaded by customer scripts (ODI for RMS-sourced data). |
Includes hierarchies, calendar, sales transactions, and so on. Clusters generated, like entities, CDT xml, similarity files, space optimized results. |
3 |
nfs/2049 |
Copy to/from local NFS mount |
Internal |
Copy files uploaded via SFTP to application server for processing with the data processing jobs. Copy exported result files to SFTP server for customer download. |
Clusters generated, like entities, CDT xml, similarity files, and so on. |
4 |
nfs/2049 |
Copy from local NFS mount |
Internal |
Pull files uploaded to SFTP server to application server for processing by Oracle Data Integrator (ODI) data load jobs. |
Hierarchies, calendar, sales transactions, and so on. |
5 |
SQLnet/1521 |
App->DB |
Internal |
DB create, update and delete operations from the application. |
Cluster parameters, query parameters, and so on. |
Oracle Retail AI Foundation Cloud Services has the following requirements:
Client System Requirements
The following operating system is supported:
Microsoft Windows 10 with Microsoft Office 2013
Note: AIF assumes that the retailer has ensured its Operating System has been patched with all applicable Windows updates. |
The following browsers are supported:
Mozilla Firefox
Microsoft Edge
Google Chrome (Desktop)
Microsoft has deprecated Internet Explorer 11 in Windows 10 and recommends using Edge as the default browser. Refer to the Oracle Software Web Browser Support Policy (/www.oracle.com/middleware/technologies/browser-policy.htmlhttps:/
) for additional information.
Other Requirements
The user's source IP address must be communicated to the application cloud administration team for security purposes.
The SFTP client used for uploading and downloading data must be compatible with the SFTP protocol used by the application. Examples include:
Putty Command line client
Win SCP
WS_FTP Pro Version 9
Note that all file exchange must be carried out in binary format.