2 Implementation Considerations

RDF is part of the Retail Analytics Platform (RAP) and imports the forecast generated by the Retail Science Engine (RSE). The RDF application is tailored towards the workflow of a Forecast Analyst. Whereas the data scientist would interact with the RSE UI to set low level forecast tuning parameters. The following sections describe the overall implementation flow and whether it is configured on the RDF side or RSE side.

The following information must be considered before configuring Demand Forecasting Cloud Service:

• Configuration Considerations

• RDF Cloud Service Hierarchies

• RDF Cloud Service Input Data

• Integration

• User Roles and Securities

• Internationalization

Configuration Considerations

Before implementing RDF Cloud Service, an implementor should first answer the following questions listed in Table 2-1

Table 2-1 Configuration Considerations and Applications

Consideration	Applications
Is my forecasted item Long Lifecycle (LLC) or Short Lifecycle (SLC)?	RSE
What is the purpose of my forecast? To drive replenishment, allocation, or others?
Are there any promotions that impact my forecast? If yes, how can I define the promotions?	RSE
Based on the purpose of my forecasting, which level should the forecast be generated on (sku/stor/week)? How many escalation levels are needed for the forecasting? Which level should the forecast be exported to?	RDF, RSE
What data is available to use for forecasting: rsal, psal, csal, Promotions, or Price?
What kind of preprocessing is needed: Outage, Outlier, Depromote, or Deseasonalize Smooth?	RSE
How do I want to handle New Items? Is there any product attribute information?	RDF
Do I want to integrate RDF Cloud Service with other Applications?	RDF, RSE
If I want to use grouping in my escalation levels, how do I group my item/stores?	RDF, RSE
Do I need to generate daily forecast, and/or both weekly and daily forecasts?	RDF, RSE
Do I have a foundation system to provide foundation (hierarchy) data?

Depending on the answers to the previous questions, the implementor can use the RDF Cloud Service plug-in to generate RDF Cloud Service configurations. For details about how to generate RDF Cloud Service configuration, refer to RDFCS Configuration. The generated RDF Cloud Service configuration can be customized to satisfy client specific requirement. For details about how to customize RDF Cloud Service configuration, refer to RDF Cloud Service Extensibility.

Note:

In order to implement planning applications on RAP, you should ensure their foundation data such as Product and Organization hierarchies, align with RMFCS (Merchandising Systems) so that

the foundation and transactional data can be used by all applications in RAP. They can have additional alternate dimensions than available in RMFCS (Merchandising Systems), if it is needed for their planning solution.

Customers can use the flex fields available in RAP Foundation files to interface this additional data. Also, if multiple planning applications like MFPCS, APCS, or RDFCS reside in the same PDS, then the common hierarchies should have the same dimension names to share the data interfaced from RAP. However additional non-shared dimensions can be present in each application.

RDF Cloud Service Hierarchies

There are four type of hierarchies in RDF Cloud Service:

• Standard RPAS Hierarchies Files

• User Managed RDF Cloud Service Hierarchies

• Plug-in Generated RDF Cloud Service Hierarchies

• RHS Hierarchies

Standard RPAS Hierarchies Files

This is the foundation data to build any RPAS solution. Demand Forecasting Cloud Service requires the standard three hierarchy files, Calendar, Product, and Location. Additional sets of hierarchy files specific to different solutions may also be needed. The standard hierarchy files for Calendar, Product, Location and Product Attributes need to be loaded into the RI interface. Refer to the Data Requirements section in the Oracle Retail Analytics Platform Implementation Guide.

Note:

The following format only shows the hierarchy structure used by RDF. The file to be loaded needs to conform to the RI interface. The user provides standard RPAS hierarchy files.

For information on the hierarchy files, see the following sections:

• Calendar Hierarchy File (CLND)

• Product Hierarchy File (PROD)

• Location Hierarchy File (LOC)

• Products Attributes Hierarchy File (PATR)

Calendar Hierarchy File (CLND)

Name	Label	Hierarchy Type	Parent
DAY	Day	Main	None
WEEK	Week	Main	DAY
MNTH	Month	Main	WEEK
QRTR	Quarter	Main	MNTH
HALF	Half	Main	QRTR
YEAR	Year	Main	HALF
DOW	DAY OF WEEK	Alternate	DAY
WOYR	Week of Year	Alternate	WEEK
STDB	STD/BTA	UDA	WEEK

Product Hierarchy File (PROD)

Name	Label	Hierarchy Type	Parent
SKU	Item	Main	None
SKUP	Style/Color	Main	SKU
SKUG	Style	Main	SKUP
SCLS	Sub-Category	Main	SKUG
CLSS	Category	Main	SCLS
DEPT	Department	Main	CLSS
PGRP	Group	Main	DEPT
DVSN	Division	Main	PGRP
CMPP	Company	Main	DVSN
VNDR	Vendor	ALT	SKU
PAT1	Prod Attribute 1	UDA	SKU
PAT2	Prod Attribute 2	UDA	SKU
STA1	Style UDA 1	UDA	SKUG

Location Hierarchy File (LOC)

Name	Label	Hierarchy Type	Parent
STOR	Location	Main	None
DSTR	District	Main	STOR
REGN	Region	Main	DSTR
CHNL	Area	Main	REGN
CHAN	Chain	Main	CHNL
COMP	Company	Main	CHAN
SFMT	Store Format	Alternate	STOR
STCL	Store Class	Alternate	STOR
SAT1	Store Attribute 1	UDA	SAT1
SAT2	Store Attribute 2	UDA	SAT2

Products Attributes Hierarchy File (PATR)

The product attributes hierarchy represents attributes associated with products. These attributes are used to group products within categories.

This hierarchy is intended to capture all product attributes for all product types. The attributes are then assigned to individual products. This assignment is used to calculate similarity scores between items.

The retailer needs to load the Product Attributes into the RI interface. Refer to the Attribute Files section of the Oracle Retail Analytics Platform Implementation Guide.

In RDF, the Product Attribute hierarchy structure is described in this section.

Name	Label	Hierarchy Type	Aggs
PATV	Prod Attribute Value	Main	None
PATT	Prod Attribute	Main	PATV

The following table describes the fields in this file.

Field	Description
Prod Attribute Value	The various values that an attribute might have. For example, the package type attribute might take the values bag, box, or convenience.
Prod Attribute	The name of a product attribute, such as brand, family type, flavor, grain, package type, size, or temperature.

Note:

In RAP, the product attributes would flow from RI to RDX to PDS. When imported this way into PDS, the product attribute name is concatenated with the product attribute values using ‘_’ to make the product attribute values unique. The Product Attribute name for Supplier (W_PDS_SUPPLIER_D) is used as ‘supp’ and Brand (W_PDS_BRAND_D) is used as ‘brnd’.

User Managed RDF Cloud Service Hierarchies

The implementor or retailer can update these RDF Cloud Service hierarchy files. The GA RDF Cloud Service package contains these hierarchy files.

For information on the hierarchy files, see the following sections:

These hierarchies can be classified into two categories:

• Hierarchies that are configured in Retail Science and RDF:

  • Group Level Hierarchy (GLVH)

  • Group Hierarchy File (GRPH)

  • Offer Hierarchy (OFFH)

  • Reward Type Hierarchy (RDTH)

• Hierarchies that are configured only in RDF:

  • Business Rule Attribute Hierarchy File (ATTH)

  • Business Rule Hierarchy File (RULH)

  • Condition Hierarchy File (CONH)

Group Hierarchy File (GRPH)

The group hierarchy is an internal application-specific hierarchy to divide item/stores into different grouping to use during parameter estimation and forecasting. You can customize this hierarchy during implementation and use the GA dataset hierarchy as a reference. Users can add or change how many groups are allowed in the domain through modifying the group hierarchy data file.

File name: grph.csv.dat

File format: comma-separated values file

The following table describes the fields in this file.

Name	Description
GRPD	This is the grouping to use during estimation and forecast.

Example:

111,Time Series Group 111 112,Time Series Group 112 113,Time Series Group 113 114,Time Series Group 114 115,Time Series Group 115

Business Rule Attribute Hierarchy File (ATTH)

This is attribute hierarchy used in the Business Rule Engine functionality. Note that this is different from the PATR (Product Attribute) hierarchy. This hierarchy file is a hybrid between user-loaded attributes and RDF GA calculated attributes generated by the plug-in. Refer to the Appendix: RDFCS Business Rule Engine for detailed information on Business Rule Engine.

File name: atth.hdr.csv.dat

File format: comma-separated values file

The following table describes the fields in this file.

Name	Label	Hierarchy Type	Parent
ATTD	Attribute	Main	None
ATDT	Attribute Data Type	Alternate	ATTD
ATSC	Attribute Source	Main	ATTD
ATST	Attribute Source Type	Main	ATSC
ATTP	Attribute Type	Alternate	ATTD

Example:

attd,attd_label,atdt,atdt_label,atsc,atsc_label,atst,atst_label,attp,attp_labelbrand,Brand,3,String,ldprdattstr01,loaded Product Attribute 01,load,Loaded Attribute,prod,Product Attributeregn,Region,3,String,regn,region of location Hierarchy,hier,Hierarchy Attribute,loc,Location Attribute

Business Rule Hierarchy File (RULH)

This hierarchy structure is used to associate Business Rules within a Business Rule Group in the Business Rule Engine functionality. The GA hierarchy file loads five placeholder Business Rules per Rule Group. The implementor can customize this file, based on the maximum number of rules the retailer would like to define per Business Rule Group. Refer to the Appendix: RDFCS Business Rule Engine for detailed information on Business Rule Engine.

File name: rulh.hdr.csv.dat

File format: comma-separated values file

The following table describes the fields in this file.

Name	Label	Hierarchy Type	Parent
RULD	Business Rule	Main	None
RULG	Business Rule Group	Main	RULD

Example:

ruld,ruld_label,rulg,rulg_labelr001,Rule 1,g001,Rule Group 1r002,Rule 2,g001,Rule Group 1 r003,Rule 3,g001,Rule Group 1 r004,Rule 4,g001,Rule Group 1 r005,Rule 5,g001,Rule Group 1

Condition Hierarchy File (CONH)

This is the Condition hierarchy used to organize the strategies (conditions) in defining a business rule. Refer to the Appendix: RDFCS Business Rule Engine for detailed information on Business Rule Engine.

File name: conh.hdr.csv.dat

File format: comma-separated values file

The following table describes the fields in this file.

Name	Label	Hierarchy Type	Parent
COND	Rule Condition	Main	None

Example:

cond,cond_labelcond01, condition 01cond02, condition 02cond03, condition 03cond04, condition 04

Group Level Hierarchy (GLVH)

This hierarchy contains the Group Sets used for the Grouping functionality. It is configured in RSE and needs to be imported from Retail Science.

The following table describes the fields in this file.

Name	Label	Hierarchy Type	Parent
GLVL	Group Level	Main	None

Example:

glvl,glvl_labelsrate01,clss/regn/week for Rate of Sales level 1ptier01,scls/chnl/week for Price Tier level

Offer Hierarchy (OFFH)

This hierarchy represents the offer/promo fields. If promo has been enabled in Retail Science this hierarchy will be imported during the batch. Refer to the RAP Integration for detailed information. If promo is not enabled in Retail Science, then the retailer can upload the GA offer hierarchy file. Note that the offer hierarchy needs to be populated for the Forecast Review workbook to build.

File name: offh.hdr.csv.dat

File format: comma-separated values file

The following table describes the fields in this file.

Name	Label	Hierarchy Type	Parent
offd	Offer	Main	None
prmd	Promo	Main	offd
camd	Campaign	Main	prmd
offa	Offer Alias	ALT	offd
cama	Campaign Alias	ALT	camd

Reward Type Hierarchy (RDTH)

This hierarchy represents the reward type that can be associated per offer. If promo has been enabled in Retail Science this hierarchy will be imported during the batch.Refer to the RAP Integration for detailed information. This hierarchy is needed to build the Offer Analysis workbook. This workbook is useful only if promo is enabled.

File name: rdth.hdr.csv.dat

File format: comma-separated values file

The following table describes the fields in this file.

Name	Label	Hierarchy Type	Parent
rdtp	Reward Type	Main	None

Plug-in Generated RDF Cloud Service Hierarchies

The RDF Cloud Service plug-in generates this hierarchy files based on the RDF Cloud Service configuration:

• Final Level Hierarchy File

Final Level Hierarchy File

This hierarchy file defines the Business Rule Group Types associated with a Final Level. Please refer to the RDF Business Rule Engine Appendix of this guide for detailed information on Business Rule Engine.

File name: flvh.hdr.csv.dat

File format: comma-separated values file

The following table describes the fields in this file.

Name	Label	Hierarchy Type	Parent
RGTP	Business Rule Group Type	Main	None
FLVL	Final Level	Main	RGTP

Example:

rgtp,rgtp_label,flvl,flvl_labelaprv01,approve 01,01,01 Weekly Units Forecastnavi01,navigation 01,01,01 Weekly Units Forecast

RHS Hierarchies

The PROR and LOCR internal hierarchies are mirrored hierarchies of the PROD and LOC hierarchies. They are also referred as PROD RHS and LOC RHS. In the RPAS Cloud Edition versions 19.0 and later, PROR and LOCR are considered as virtual hierarchies. Refer to the Oracle Retail Predictive Application Server Cloud Edition Configuration Tools User Guide for information on Virtual Hierarchies.

Since these hierarchies are virtual, you do not have to load the hierarchy files. All of the other operations remain the same. You can register measures on PROR and LOCR and include them in workbooks.

Notes about these virtual hierarchies:

• PROR and LOCR hierarchies have been marked as virtual in the GA configuration.

• We cannot define security dimension on a virtual hierarchy or make them translatable.

• Virtual hierarchies cannot have user defined dimensions.

• If a retailer is upgrading from a pre-19.0 RDF version, then RDF will automatically mark them as virtual and conform to the virtual hierarchy requirements.

RDF Cloud Service Input Data

A detailed data set is required to use the capabilities of RDF Cloud Service to its fullest. Some of the data required is relatively easy to obtain, for example, information about sales. To simplify the data integration, all measure files are configured to be loaded as one measure per file. Each measure's data must be present in a separate file and the file name must be the same as the measure name with the .csv.ovr extension. All files must be in csv format. During the initial domain build, all data files marked as required are needed with historical data to build the domain.

Measure Name and Intersections

Because many RPAS measure names and intersections are dynamically generated by RDF Cloud Service plug-in. Tokens will be used to represent the RDF Cloud Service level names. The labeled intersection were also listed for measure intersection

Table 2-2 lists the Token names.

Table 2-2 Token Names

Token Name	Description
_CF_	Forecast Final Level Name, such as 07
#LLC_frcst_L_#	Forecast final level intersection generated by plug-in based on user specified plug-in input parameters
#LLC_frcstTS_L_#	Forecast final level timeseries intersection = LLC final level intersection – clnd dim generated by plug-in based on user specified plug-in input parameters
#LLC_seascurve_L_#	Forecast level intersection – clnddim + woyr generated by plug-in based on user specified plug-in input parameters
#SLS_INTX#	Sales History intersection. This labeled intersection is user defined
#SLSNC_INTX#	Sales History intersection -clnd dim This labeled intersection is user defined
#NIT_ATT_WGT#	Attribute weight intersection, generated by plug-in based on user specified plug-in input parameters
#NIT_SKU_ATT#	Product attribute intersection, generated by plug-in based on user specified plug-in input parameters
#NIT_SKUSTR_INTX#	New Item assignment intersection, generated by plug-in based on user specified plug-in input parameters
#PRESLS_INTX#	Forecast Preprocessing data source input intersection

Measure Names and Descriptions

Table 2-3 lists the measure names and descriptions. The measure field descriptions include:

Module Used

This field explains which solution is using the file. The possible values can be: All, New Item, and Forecast.

Required or Optional Required

This field means the data is necessary. Optional means that during data load and, if not loaded, certain functionality which uses those measures cannot be used. All administration measures are marked as Optional for data load, since those can be directly set in the Admin workbooks as well.

Load Frequency

This specifies the suggested frequency for the data load. It uses the following values:

• W - Weekly

• A - Anytime as needed or when the values change in source system; it can be weekly, monthly, quarterly, or yearly

Data Source

This specifies the typical data source to get that measure data:

• RI - Oracle Retail Insights or equivalent Data Warehouse solutions

• Admin - Data can be set by Administrator based on customer data referencing sample data in GA domain.

• RDF, IPCS - Oracle Retail Planning Cloud Service or equivalent. Can be readily loaded from RMS or derived from data loaded from RMS.

• ORASE - Oracle Retail Advanced Science. Those are the derived measure files extracted from ORASE integration files.

• RMS - Oracle Retail Merchandising System or equivalent. Can be readily loaded from RMS or derived from data loaded from RMS.

• 3P - Third-party data aggregator such as Nielsen or Symphony IRI.

Load Intersection

Most of the time, the load intersection of the measure is the same as the base intersection of the measure. When the field is empty, the load intersection is the same as base intersection.

Table 2-3 RPAS Measure Names and Intersections

Measure Name	Measure Description	Base Intersection	Measure Type	Module Used	Required or Optional	Load Frequency	Data Source	Load Intersection
rsal	Regular Sales	#SLS_INTX#	Real	all	Required	Weekly	RMS/RI	#DAYSLS_INTX#
psal	Promotion Sales	#SLS_INTX#	Real	all	Required	Weekly	RMS/RI	#DAYSLS_INTX#
csal	Clearance Sales	#SLS_INTX#	Real	all	Required	Weekly	RMS/RI	#DAYSLS_INTX#
ldactivefcstitem	Active Forecast Item Indicator	#SLSNC_INTX#	Boolean	all	Optional	Weekly
outlierind_CF_	Loaded OutLier Indicator	#PRESLS_INTX#	Boolean	Preprocess	Optional	Weekly
outageind_CF_	Loaded Outage Indicator	#PRESLS_INTX#	Boolean	Preprocess	Optional	Weekly
prdattT	Product Attribute	#NIT_SKU_ATT#	String	New Item	Optional	Weekly	RMS/RI
nitdattwgt	Attribute Weight	#NIT_ATT_WGT#	Real	New Item	Optional	Weekly
nitfcststovr	New Item Forecast Start Date	#NIT_SKUSTR_ INTX#	Date	New Item	Optional	Weekly
nisros	New Item Base Rate of Sales	#NIT_SKUSTR_ INTX#	Real	New Item	Optional	Weekly
likeitemexmask	Like Item Exclusion Mask	#NIT_SKUSTR_ INTX#	Boolean	New Item	Optional	Weekly
promoaggprof_ CF_	Promotion Aggregation profile for Forecast	User provided during configuration time (Promo Aggprof Intx)	Real	LLC	Optional	Weekly/ Anytime
basespreadprof_ CF_	Profile to spread forecast from week to day	User provided during configuration time (Baseline Spread Prof Intx)	Real	LLC	Optional	Weekly/ Anytime
The following measure can be edited in RDF Cloud Service workbooks. It can also be loaded if a data file is provided.
grpassignpos_CF_	TimeSeries Grouping membership for LLC. It shall contain group dimension position names.	#llc_frcstTS_L_#	String	LLC	Optional	Weekly/ Anytime

Historical Data

It is recommended that you have at least two full years of historical data for long life cycle forecasting and one full year of historical data for short life cycle forecasting.

Loading and Extracting Data

Data is loaded into RDF Cloud Service using the Online Administration Tools, which in turn use standard RPAS utilities. For more information on loading and extracting data using Online Administration Tools, see the Oracle Retail Demand Forecasting Cloud Service Administration Guide

Loading Image Based Data

RDF Cloud Service is pre-configured to support the display of images for items and product attributes in the Forecast Review and New Item workbooks. Table 2-4 lists the dimension attribute measures used to load images.

Table 2-4 Labeled Intersections

Measure	Hierarchy	Dimension
skuimage	PROD	sku
skupimage	PROD	skup
skugimage	PROD	skug
skurimage	PROR	skur
skprimage	PROR	skpr
skgrimage	PROR	skgr
patvimage	PATR	patv
pattimage	PATR	patt

The Content Server exposes the client's image files placed into a particular directory as HTTP URLs. The images must be defined in the load file in an xml format. The images are available at:

http://{content server url}/imgfetch/{sub directory if defined}

Sample File for skuimage.csv.ovr

The first field represents the SKU ID followed by the required image location. At a minimum, a thumb size image file must be loaded to show in the pivot table. However, both the thumb and full size images can be loaded.

10000010,"<image id=""main"" label=""Front View"">\<url size=""thumb"">http://msp00alq.us.oracle.com:9001/contentserver/imgfetch/sku_10000010_main_thumb.jpg</url></image>"

Example File for skuimage.csv.ovr

10000010,"<image id=""main"" label=""Front View"">\
<url size=""thumb"">http://msp00alq.us.oracle.com:9001/contentserver/imgfetch/sku_10000010_main_thumb.jpg</url>
<url size=""full"">http://msp00alq.us.oracle.com:9001/contentserver/imgfetch/ sku_10000010_main_full.jpg</url></image>"

Integration

RDF Cloud Service is part of the Retail Analytics Platform (RAP). The foundation data needs to be loaded into the RI data interface. Any hierarchy or data specific for RDF can be loaded via the File Transfer Service (FTS).

User Roles and Securities

To define workbook template security, the system administrator grants individual users, or user groups, access to specific workbook templates. Granting access to workbook templates provides users the ability to create, modify, save, and commit workbooks for the assigned workbook templates. Users are typically assigned to groups based on their user application (or solution) role. Users in the same group can be given access to workbook templates that belong to that group alone. Users can be assigned to more than one group and granted workbook template access without belonging to the user group that typically uses a specific workbook template. Workbook access is either denied, read-only, or full access. Read-only access allows a user to create a workbook for the template, but the user cannot edit any values or commit the workbook. The read-only workbook can be refreshed.

For more information on security, see the Oracle Retail Predictive Application Server Cloud Edition Administration Guide. For more information on data security in a cloud environment, see the Hosting Policy documents for the cloud solution.

Internationalization

Internationalization is the process of creating software that can be translated more easily. Changes to the code are not specific to any particular market.

Oracle Retail applications have been internationalized to support multiple languages.

The RPAS platform supports associated solution extensions and solution templates.

• A solution extension includes a collection of code and generally available configurations. Typically, solution extensions are implemented by a retailer with minimal configuration.

• A solution template does not include code. A solution template is most typically implemented as a retailer configuration.

Oracle Retail releases the translations of the RPAS server and client, as well as strings from the solution extensions.

Translations of the solution templates are released. All templates have the ability to support multi-byte characters.

For more information on internationalization, see the Oracle Retail Predictive Application Server Cloud Edition Administration Guide.

Translations are available for RDF Cloud Service for the following languages:

• English (United States, Great Britain, Canada, Australia)

• Chinese (Simplified)

• Chinese (Traditional)

• Croatian

• Dutch

• French

• German

• Greek

• Hungarian

• Italian

• Japanese

• Korean

• Polish

• Portuguese (Brazilian)

• Russian

• Spanish

• Swedish

• Turkish

Note:

For information about adding languages for the first time or for translation information in general, see the Oracle Retail Predictive Application Server Cloud Edition Administration.