AWS for Industries

Modernizing retail demand planning with SAS Intelligent Planning on AWS

Modernizing retail demand planning with SAS Intelligent Planning on AWS

About 100 years ago, a retailer with over 10,000 locations across the United States had a problem. How could they balance the need for goods with available stock for each store in their chain, forecast the need for replenishment, and coordinate logistics to ensure each location received the stock they needed for the customers they served? They used two things: data and gut feel.

Let’s examine the first thing they used: data. Data comes in many forms, and the Great Atlantic and Pacific Tea Company, also known as A&P, had data. Lots of it. However, it was on paper, lagged in time (store orders were mailed by way of the post office into the corporate office), and manually sorted and processed. Yet, remarkably, every store got stock on a regular basis, and every store was able to serve its customers.

If A&P could do it 100 years ago, what’s so tough about demand planning, logistics, price optimization, and fulfillment today? Glad you asked. First, assortment for an A&P store at the time may have been a few hundred items. Compare that to today’s modern grocery chain with north of 90,000 UPCs (not counting some 30,000 UPCs for greeting cards alone). Second, products moved a bit slower back then, so an order of 10 crates of tomato soup might have taken a month to get to a store. Other differences today include much lower margins and significantly more efficient consumer packaged goods (CPG) suppliers. What separates the 2,500 square foot A&P location in 1924 from a 75,000 square foot grocery store is, quite simply, scale.

Times have changed. Data transitioned from paper to digital. Today, store orders are transmitted electronically to host systems, and computer-aided ordering (CAO) automatically handles store-level replenishment requests based on item movement captured from the point of sale system. And economies of scale for manufacturers have driven costs down, so prices can be kept low, and margins can remain manageable.

Industry problem: Yesterday’s demand planning needs modernization

So, why can’t we just keep doing what A&P did 100 years ago? Sounds easy, and for many retailers, demand planning hasn’t changed that much. Yes, it’s true that the data moves faster today, but even 100 years ago, prior sales were used as an indicator of what future demand might look like. Therefore, it shouldn’t surprise you to hear that demand planners still use historical shipment data; but would it surprise you to learn that they combine that with 100% judgement to forecast demand? Even for today’s demand planner, gut feel and intuition take center stage.

It’s also true that judgement and experience make for some of our best demand planners—all of whom are essential to every retailer in the world. However, even the best demand planners are confronted with so much information that they become overloaded. At some point, the most experienced demand planner resorts to their “best guess” estimates for demand.

Consider an average day for a demand planner working for a grocery chain. National brands like Kraft or Procter & Gamble have shorts predicted for shipments destined to half of the chain’s stores. An imminent snowstorm is heading toward the northeast region, signaling demand for winter supplies, while supplier shorts in baby formula have left Miami store shelves empty. Drayage increases in other parts of the country are forcing price changes for some of the chain’s private label products. Add in other factors, such as micro-changes in customer demand for items used for in-store cooking demonstrations, wine tastings, and other local events. In the Midwest, an extremely contagious strain of influenza is spreading, inundating in-store pharmacies with flu shot appointments—for vaccines that are quickly running out. You get the picture.

With data coming at them from every direction, how could the demand planner possibly be expected to consume, analyze, and estimate demand for nearly 90,000 items in each location while making certain prices are optimized? Well, they can’t. So, they use judgement and intuition—they give it their best guess.

Demand planning, replenishment, price optimization—they all must work in concert, so that the right product gets to the right location at the right time. From the wayward customer who needs snowmelt because of a nor’easter coming their way in Connecticut, to the parents needing formula for their newborn in Miami, and to the family needing flu vaccines in the Midwest, customer demand is—quite literally—all over the map. Demand planning from the gut just doesn’t scale. And guessing isn’t good enough anymore.

Stacking up: How mature is your demand planning?

By now, you might be thinking about your own organization. You might cite the use of various analytics, advanced models, and tools put in place to facilitate better forecasting. Perhaps having these systems at the demand planner’s fingertips gives you the confidence that you’re doing it right. But are you? How does your demand planning stack up to that of other retailers? Is there some kind of measure you can apply so that you know if you’re looking toward the future, or possibly stuck in the past?

When we look at demand planning, both current and future states, it becomes apparent that there is a continuum along which we can plot the level of demand planning maturity within an organization. Although these levels might have somewhat blurry edges with some overlaps, demand planning falls into the four general stages outlined in this post.

Stage 1: Yesterday’s demand planning

In the first stage of demand planning, both the demand planner from 1920s A&P and many of today’s demand planners still use prior trends and 100% judgement. For many demand planners today, that process works well enough, especially for chains with small footprints.

Stage 2: Systematic, yet not well measured

In this stage, demand planning becomes more collaborative and formalized. Shipment history is captured in an enterprise resource planning (ERP) system, and planning becomes more centered on supply rather than the consumer (what’s available, rather than what’s wanted). The demand planner can forecast only what is available from suppliers, and analytical outputs are based on system-generated simple time series statistical methods, many of which aren’t transparent. Data, especially from various suppliers, is inconsistent and requires cleansing efforts (many times manual) prior to use. Other factors influence demand planning, such as certain suppliers or stores favored over others, and manual overrides made by teams with differing priorities.

Stage 3: Automation and metrics bring more accurate forecasts

The third stage of demand planning maturity is becoming more democratized, subscribing to customer wants, with an even approach to distribution.

Sophisticated data management is in-place, where data is cleansed and stored in repositories such as data lakes. Demand signal repositories help harmonize and normalize data across the enterprise, supporting fully automated analytics. Predictive analytics has replaced simple time series methods, complementing ERP systems. Now, sales orders and shipment data create short-term demand sensing forecasts. Automatic outlier detection and predictive models enhance forecast accuracy. More rigor is infused into the sales and operational business planning process.

Stage 4: Fully integrated and supplemented by AI

The fourth stage of demand planning maturity looks to integrate all planning, whether strategic, operational, or tactical, into one analytics platform. Now, cloud-native demand planning solutions support continual optimization across all teams. Artificial intelligence (AI) and machine learning (ML) now drive automation, providing “assisted demanding” to boost planner’s forecast value added (FVA). Causal factors, such as promotions, pricing, social media trends, and local economic data, improve forecast accuracy. Data signals from new and existing sources boost statistical capabilities. Finally, data scientists take their seat alongside demand planners, helping fine-tune recommendations from predictive analytics models for highly accurate forecasts.

Now that you have a clear understanding of the different stages of demand planning maturity, where does your organization land? If you see that your capabilities span the early stages, you’re not alone. The question is, how do you get to Stage 4?

Modernize demand planning with SAS Intelligent Planning on AWS

If you are a retailer operating at different stages of demand planning maturity, would it help if you had a single platform where you can bring everything together in one place? If so, we have something that might help. SAS and AWS have worked together to offer SAS Intelligent Planning, running on AWS. Built on the SAS Viya platform, which enables ML, AI, and visualization for the SAS Retail Solutions, SAS brings a completely integrated set of solutions to you; and it’s now offered on AWS.

Other SAS Retail Solutions, highlighted in green in the following figure, include Intelligent Pricing, Intelligent Inventory Management, and Intelligent Performance Management—all of which can take advantage of models created in SAS Intelligent Planning.

Figure 1: SAS Retail Solutions powered by SAS ViyaFigure 1: SAS Retail Solutions powered by SAS Viya

If you use SAS Customer Intelligence 360 (also running on AWS) as your Customer Data Platform (CDP), it works seamlessly with SAS Intelligent Planning and the rest of the SAS Retail Solutions.

SAS Intelligent Planning benefits

With SAS Intelligent Planning, AI-assisted demand planning ensures on-shelf availability through a hyper-accurate demand planning service that is available for both CPGs as well as grocery, drug, and convenience retailers. SAS Intelligent Planning includes the following capabilities:

SAS Demand Planning – Offers large-scale automated statistical modeling, forecasting, and optimization that includes out-of-the-box modeling strategies with predefined models. It’s cloud-ready, with an open API design, and it integrates with SAS Visual Forecasting and SAS Visual Data Mining and ML.

SAS Assortment Planning – Supports unlimited dimensions, allows predictive recommendations on most influential attributes, uses scenario planning with AI- and ML-based assortments, and accommodates demand forecasts for both seasonal and non-seasonal planning.

SAS Financial Planning – Provides a common user interface across modules, uses an advanced in-memory planning engine, and provides seamless integration with embedded analytics.

SAS Intelligent Planning helps you by detecting forward-looking demand signals and providing recommendations for balanced and profitable commercial plans across all your channels, categories, and customers. SAS Intelligent Planning is fully automated, providing you with insights for solving business problems without having to worry about complicated cloud operations. Planning ahead for shopper needs in uncertain times just got easier.

SAS Intelligent Planning on AWS

The engine that powers SAS Intelligent Planning and the other SAS Retail Solutions, SAS Viya, provides a suite of products and capabilities for data management, analytics, data visualization, and ML. First released in 2016, SAS Viya builds on years of experience and makes analytics straightforward and more accessible while helping to bind the SAS Retail Solutions together into one cohesive stage. End-users access SAS Retail Solutions through web browsers without needing to install anything locally. Additionally, the interface is customizable for different roles.

SAS Viya uses in-memory processing and distributed computing to handle large, complex data workloads. SAS Viya includes SAS Studio for coding, SAS Visual Analytics for reporting and visualization, SAS Data Management for data preparation and governance, and SAS Model Manager for ML model management. SAS Viya also integrates with popular open source technologies including Python, R, Lua, and Java for additional programming flexibility. APIs are provided to embed analytics into other applications.

Figure 2: SAS Viya on AWSFigure 2: SAS Viya on AWS

SAS Viya uses a variety of AWS services, including Amazon Elastic Compute Cloud (Amazon EC2), Amazon Elastic Kubernetes Service (Amazon EKS), Amazon Elastic Container Registry (Amazon ECR), and Amazon Relational Database Service (Amazon RDS). SAS Viya takes advantage of the flexibility AWS provides with on-demand compute, storage, and networking capabilities that can scale up or down automatically to meet changing needs. This makes it effortless to scale SAS Viya quickly. And with AWS, you pay only for the resources you use, which can lead to significant cost savings compared to buying and maintaining your own hardware. AWS also allows right-sizing resources to match workloads and optimize costs. So, no matter where you are in the world, you can run SAS Viya and SAS Intelligent Planning in any AWS Region.


SAS provides a GitHub repository that you can use to get started in your own AWS account. The SAS Viya 4 Infrastructure as Code (IaC) for Amazon Web Services project contains a Terraform script that provisions the AWS resources required to deploy SAS Viya platform product offerings, including SAS Intelligent Planning. SAS Viya is deployed from a single EC2 instance, which you create in your preferred AWS account. From there, you’ll run the Terraform script (from the GitHub project) that creates and launches the AWS infrastructure components required for SAS Viya.

When the SAS Viya deployment is complete, you’ll apply the license for the SAS Intelligent Planning (obtained from SAS) and any of the other SAS Retail Solutions you plan to use. To get you started, you can install sample data that SAS provides to learn how SAS Intelligent Planning works before you import your own data.


Demand planning has grown up in the 21st century, and if you’re stuck forecasting as A&P did in the 1920s, it may be time to make the move to a fully integrated and AI-supplemented solution. SAS Intelligent Planning on AWS eliminates the guesswork from demand planning. Subscribing to the fourth stage of demand planning maturity, SAS Intelligent Planning includes AI-assisted demand planning, self-tuning demand plans to ensure the right products are stocked at the right time in the right locations. The suite helps retailers collaborate planning activities, balance and reconcile commercial goals across every department, and scale capacity according to a retailer’s unique business cycle. With SAS Intelligent Planning on AWS, retailers have secure, always-on analytics, which allows them to focus on the results instead of the complexities of computer infrastructure.

Additional reading

Related resources:
AWS industrial cloud services

Cody Shive

Cody Shive

Cody Shive is the Global Partner Solutions Architect for Grocery, Drug, and Convenience at AWS, where he works with both cloud and physical store retail partners. Cody has 20+ years in Retail as an independent consultant, a technical lead for IBM/Toshiba Global Commerce Solutions, and as a Retail Transformation architect for NCR. Cody specializes in deep data analytics and keeps himself involved in self-service solutions such as Self-Checkout and Dash Cart technologies. He is passionate about the retail industry, stemming from his very first job at Albertsons in Florida. Cody is a graduate from the University of North Florida with a degree in Computer and Information Sciences and minor in Business Management.