The brain behind retail planning — transforming historical data and statistical models into actionable forecasts, financial plans, and assortment strategies.
What is RPAS?
Oracle Retail Predictive Application Server (RPAS) is the platform that powers all of Oracle Retail's planning and forecasting applications. It is not a single application — it is a multi-dimensional database engine that hosts multiple planning applications (called "solutions") on top of it.
Think of RPAS as the "Oracle Database" of planning. Just as Oracle Database hosts tables and PL/SQL procedures, RPAS hosts planning domains containing hierarchies, measures, and workbook templates. Business planners interact with these domains through Excel-like grid interfaces called workbooks.
Domain Architecture
RPAS organizes data into domains — self-contained databases that store hierarchies, measures, and planning data.
Hierarchies
RPAS uses the same hierarchies that exist in RMS, but extends them for planning purposes:
- Product Hierarchy: Company → Division → Group → Department → Class → Subclass → Style → SKU
- Location Hierarchy: Company → Region → District → Store (or Channel → Web/Stores)
- Calendar Hierarchy: Year → Season → Month → Week → Day
Measures
A measure is a data point at a specific intersection of the hierarchies. Examples:
- Sales Units at SKU/Store/Week = 47 (actual)
- Sales Units at SKU/Store/Week = 52 (forecast)
- Beginning of Period Inventory at Class/District/Month = $1.2M (plan)
The Cube
All measures across all hierarchy intersections form a multi-dimensional cube. Planners navigate this cube by rolling up (aggregating from SKU to Class) or drilling down (from Month to Week).
CALENDAR
┌─────────┐
/ /│
/ Week / │
PRODUCT ┌─────────┐ │
/ /│ │
/ Class / │ │
┌─────────┐ │ │
│ │ │ /
│ MEASURE │ │/ LOCATION
│ (Sales │ │
│ Units) │ /│
│ │/ │
└─────────┘ │
/
/
Store
Merchandise Financial Planning (MFP)
MFP is the top-down planning process where senior merchants set financial targets for the business.
The Planning Cycle
-
Pre-Season Planning: 6–9 months before the season, the planning team sets targets:
- Total sales target by department for the season
- Target gross margin percentage
- Target inventory turns
- Markdown budget as a percentage of sales
-
Open to Buy (OTB) Calculation: MFP calculates how much the buyer is authorized to spend:
Open to Buy Formula
OTB = Planned Sales + Planned Markdowns + Planned Shrinkage + Planned End-of-Month Inventory − Beginning-of-Month Inventory − On Order − In Transit
If OTB is positive, the buyer can place new orders. If negative, the buyer has already committed more inventory than the plan allows.
- In-Season Tracking: As the season progresses, MFP compares actual sales against the plan. If sales are trending 15% below plan, the planner may reduce OTB (stop buying) or increase the markdown budget (clear excess inventory faster).
Demand Forecasting (RDF)
Oracle Retail Demand Forecasting (RDF) is the statistical forecasting engine within RPAS. It generates weekly sales forecasts at the item/store level.
Forecasting Methods
RDF uses multiple statistical models and automatically selects the best one for each item:
| Model | Best For | Technique |
|---|---|---|
| Simple Exponential Smoothing | Stable items with flat demand | Weighted average of historical sales |
| Holt | Items with a trend (growing or declining) | Exponential smoothing with trend component |
| Winters | Seasonal items | Exponential smoothing with seasonality component |
| Causal | Items influenced by external factors | Regression against promotions, holidays, weather |
| Bayesian | New items with no history | Uses similar item profiles to bootstrap forecast |
Forecast Consumption
The forecast is consumed by downstream systems:
- Replenishment (RMS): Dynamic and Time Supply replenishment methods use the RDF forecast to calculate order quantities
- Allocation: Initial allocation quantities can be based on forecast sales for the first few weeks
- MFP: In-season forecasts update the plan vs. actual comparison
- RPAS Replenishment Optimization: Uses forecasts with service level targets to set optimal safety stock
Assortment Planning
Assortment Planning (AP) answers the question: "Which items should we carry in which stores for next season?"
Not every store should carry every item. A ski resort store doesn't need swimwear. A budget-focused suburban store doesn't need luxury brands.
The Assortment Planning Process
- Define Store Clusters: Group stores with similar customer demographics and sales patterns
- Select Items per Cluster: Choose which items (or item categories) each cluster should carry
- Set Space Constraints: Each store has limited shelf/rack space — assortment must fit within the physical space available
- Finalize Assortment: The approved assortment becomes the basis for buying (PO creation in RMS) and allocation
Workbook Templates & Workflow
Planners interact with RPAS through workbooks — Excel-like grid interfaces that display measures across hierarchy intersections.
Workbook Lifecycle
- Build: The planner creates a new workbook from a template, selecting the hierarchical scope (e.g., Department 100, all stores, Spring 2027)
- Edit: The planner enters or modifies plan values in the grid
- Calculate: RPAS runs calculation rules (e.g., spreading a department-level target down to class/subclass levels)
- Submit: The planner submits the workbook for managerial approval
- Approve: The manager reviews and approves the plan
- Commit: The approved plan is committed to the master domain, making it the official plan
Workbook Types
- MFP Workbook: Financial targets (sales, margin, inventory) at Department/Month level
- Item Plan Workbook: Unit-level plans at Item/Store/Week level
- Forecast Review Workbook: System-generated forecasts for planner review and override
- Assortment Workbook: Item selection per store cluster per season
Data Flow & Integration
RPAS has bidirectional integration with RMS and other Oracle Retail modules:
Into RPAS (From RMS):
- Historical Sales: Weekly sales units and revenue by item/store — the foundation for forecasting
- Inventory Positions: Current SOH, on-order, and in-transit quantities
- Item Master: New items, hierarchy changes, and item attributes
- Price History: Regular price changes, clearance events, and promotions (for causal forecasting)
Out of RPAS (To RMS/Allocation):
- Sales Forecasts: Weekly item/store forecasts consumed by Dynamic and Time Supply replenishment
- OTB Budgets: Open-to-Buy limits that constrain buyer purchasing in RMS
- Assortment Plans: Item/store range lists that feed into RMS item-location ranging
- Allocation Targets: Store-level unit targets that feed into the Allocation module
RPAS Cloud (Planning Cloud Service)
Oracle is migrating RPAS from on-premise to cloud under the brand Oracle Retail Planning Cloud Service. Key changes:
- Web-Based Workbooks: The legacy Java Fusion Client is being replaced by browser-based HTML5 workbooks
- AI/ML Integration: The Retail Science Platform integrates machine learning models for demand sensing, new item forecasting, and markdown optimization
- Shared Infrastructure: Cloud deployments share compute resources, reducing total cost of ownership
- Automated Batch: ETL and forecast generation batches are managed by Oracle, not the retailer's IT team
Planning Best Practices
Important Gotchas
- !Forecast accuracy degrades dramatically beyond 8 weeks for fashion items. Don't force long-horizon forecasts on highly volatile, trend-driven categories — use shorter planning windows with more frequent replanning.
- !Always cleanse promotional lifts from historical data before generating baseline forecasts. If last year's sales included a 40% off promotion, this year's baseline forecast (without the promotion) should not assume the same demand level.
- !MFP plans must align with RPAS forecasts. If the financial plan says $10M in February but the forecast says $7M, the buyer will either over-buy (creating excess inventory) or the plan needs to be revised.
- !New items with no sales history are the hardest to forecast. Use the "Like Item" (Bayesian) profile feature to bootstrap forecasts from similar items that have established history.
Key Takeaways
- RPAS is the multi-dimensional planning platform that hosts MFP, RDF, Assortment Planning, and Replenishment Optimization.
- Domains organize data into hierarchies (Product, Location, Calendar) and measures (Sales, Inventory, Margin) forming a navigable cube.
- MFP sets top-down financial targets; the Open to Buy (OTB) calculation controls how much buyers can spend.
- RDF generates statistical forecasts using multiple models (Exponential Smoothing, Holt, Winters, Causal) and selects the best fit per item.
- Forecasts feed into RMS replenishment (Dynamic/Time Supply methods) and Allocation for initial distribution.
- Oracle is migrating RPAS to cloud with browser-based workbooks and AI/ML-powered Retail Science Platform integration.
