3 ways to rethink retail forecasting and demand planning

The pandemic has profoundly changed consumer shopping behaviors and experiences. Buy online, pick up in-store, curbside pickup, same-day and contactless delivery are no longer amenities – they're necessities. Consumers expect a consistent, seamless experience, whether in-store or online. And if expectations are not met, a sea of retail competitors is only a few phone swipes away.

The increasing pressure has retailers scrambling to improve their ability to precisely predict and plan for demand. Because it’s now crystal clear: An accurate and efficient forecasting and demand planning process is essential for growing profitability and improving customer service.  

If that sounds overwhelming and you don’t know where to start, here are three questions to ask as you rethink your forecasting and demand planning.

1.  Forecasting and demand planning: Can you automate and scale across the enterprise?  

Enhanced forecasting and demand planning affect multiple key decision points across every retail organization. Retailers require in-depth, accurate forecasts to:  

• Plan a compelling assortment of SKUs with the right choice count, depth and breadth. 
• Make profitable inventory investment decisions.
• Ensure inventory productively meets customer demand across a complex supply chain.
• Allocate labor effectively to each distribution center and store.
• Build optimized price and promotional strategies.

An effective cross-channel approach requires consistency and connectivity in demand planning across all of these essential business decisions. Having one platform that can solve many different forecasting challenges helps ensure alignment across merchandising, supply chain and operations.  

At a time when doing more with less is vital, automation and scalability are key to any large-scale forecasting and demand planning challenge. Automated statistical forecasting drives efficiencies in business forecasting processes and facilitates true exception-based management. This then frees up time to focus on business planning and expanding to new areas of forecasting.  

With automation, retail organizations can scale across many different forecasting use cases and down to the lowest levels of detail – to the level at which forecasts are executed and business decisions are made.

2. Forecasting and demand planning: Could you do more with advanced statistical forecasting methods?

The foundation of every enterprise forecasting journey is data-driven analytics. Retailers must be able to forecast any variable – from sales units and sales dollars to traffic at the store and the number of cases flowing through distribution centers. Accurate long-term, midterm and short-term demand forecasts are required to balance strategic and tactical decisions.  

Advanced statistical forecasting methods capture the effects of unlimited causal variables and quantify the key drivers of demand. Causal variables reflect both internal and external factors, such as price, promotions, weather, epidemiological models and social media data. Quantifying the impact of these variables provides insight for better decision making and improves the transparency of forecast models. This helps you:

• Proactively shape demand with multiple what-if scenarios and quickly understand the effect of potential initiatives or unexpected disruptions.
  
  
• Transform discussions with the business by better explaining what's in the forecast – and what's not in the forecast – to ultimately drive improved forecast adoption.

Each incremental forecasting challenge expands data volumes and increases complexity. A wide range of analytical techniques and algorithms – time series, machine learning and ensemble – is essential for improving accuracy.  

Retailers no longer need to rely on a high-level forecast that's manually disaggregated to lower levels. Now you can independently generate forecasts at each level of the product/location hierarchy – using unique models for each time series – to capture nuances of demand as the level of detail increases. Then forecasts can be automatically reconciled and aligned up and down the planning hierarchy.  

When forecasting at a detailed level like SKU/store, there's inevitably a lot of variation in the characteristics of demand across the product portfolio. For example, some products are seasonal while others only sell sporadically with intermittent demand. Demand patterns can be characterized across seasonal, intermittent, short history, etc., in order to apply a unique forecast modeling approach to each demand segment. 

A one-size-fits-all approach to forecast modeling isn’t enough. To produce accurate results, you need best-fit models using a variety of modeling techniques and intentional modeling strategies.  

3. Forecasting and demand planning: Can you establish a unified, repeatable workflow?

Implementing effective enterprise forecasting and demand planning is not just about the modeling – that’s only one part of the equation. It’s also about establishing a unified, repeatable workflow for data scientists, forecast analysts and business users.

Forecasts don’t matter if they aren’t used in decision making, so statistical forecasts need to easily integrate into downstream planning and execution systems. Statistical forecasts serve as barometers for identifying risk and opportunities against sales plans while offering an unbiased starting point for planners to incorporate their business domain knowledge. The result? Higher planner productivity, better inventory management, and an improved understanding of demand drivers and consumer behavior.

What’s next? 

The future of retail holds both opportunity and uncertainty. Retailers must adopt a resilient approach to forecasting and demand planning for better decision making across the enterprise. That approach needs to be automated and agile, bring visibility to underlying business drivers, help manage complexities and drive process efficiencies.  

With repeatable forecasting processes, retailers can respond quickly as new data comes in. With optimized forecast models, retailers can make collaborative decisions and have an immediate understanding of the downstream impact. This puts retailers on a path to long-term profitability and growth.  

About the Author

Jessica Curtis has been helping organizations drive business value through forecasting, demand planning and optimization for more than 13 years. At SAS, she empowers retail and CPG customers to solve their business challenges with data-driven analytics.