Accelerate open source forecasting with SAS (Part 1 of 3)


Welcome to the first of a 3-part series by guest bloggers Jessica Curtis and Andrea Moore:


Forecasting is core to many different business decisions across virtually every industry.  Whether you’re a retailer planning a compelling assortment of SKUs or improving labor planning for distribution centers and stores, or a consumer product goods company revamping a demand planning process.  You could be a media company leveraging forecasting for digital ad inventory planning and pricing, or a communications company forecasting network utilization for optimal resource allocation… and the list goes on.  The impact of having a better forecast is far reaching and fundamental for better business decisions across any organization.  

For over 44 years, SAS has been improving large-scale forecasting processes for thousands of companies globally.  Over the years, SAS has developed and enhanced robust forecasting software that drives bottom line value through more accurate statistical forecasts and efficient forecasting processes.  You could say we wrote the book on forecasting.  In fact, we’ve written several.   

SAS’s newest forecasting technology, SAS Visual Forecasting, offers unique capabilities to solve enterprise forecasting challenges quickly and automatically.  SAS Visual Forecasting not only includes cutting edge algorithms – machine learning and time series and ensemble, oh my! – but also built-in best practices for diagnosing historical data, automatically forecasting across complex hierarchies, and managing forecast exceptions.  One of the core tenets of SAS Visual Forecasting is that it offers an open ecosystem to run open source models and deploy models at scale 

Open source is widely used to develop forecasting models.  Many organizations begin with an open source strategy, leveraging Python or R to build forecasts, and face challenges when trying to scale across many different forecasting use cases across the enterprise.  There are benefits in running open source forecast models with SAS that allow you to build upon existing open source strategies in an agile, efficient way.  You no longer have to choose between SAS and open source – it is truly a complementary relationship. 


Many organizations struggle with creating robust forecasts and face challenges when trying to scale across different forecasting use cases.  With each forecasting challenge comes incremental data and increased complexity. 

For example, if you are a telecommunications company, you need to forecast demand for data bandwidth to guide decisions about where to invest additional infrastructure (such as cells).  To plan new infrastructure investments, you want to understand how demand for bandwidth will change over time.  You build an open source forecast solution to gauge the incremental overall demand in a market. This forecast is the basis of the annual planning for how much to grow the network.  Your use of analytics for planning is well regarded.  Next cycleinstead of just one overall forecast for planning, you are asked for analysis that incorporates additional data and to create a forecast for each individual piece of equipment.   

 In Figure 1, this request implies a forecast not just for the market, but also for each city, head end, node and premise location.  The analysis must be scaled to work with large data volumes and large number of time series to produce thousands, or even millions of forecasts.  Using statistical forecasts for each granular network component rather than allocating a high-level value down to other levels will result in increased forecast precision.  With this additional forecast precision, the capital planning process will be even sharper, and resources will be allocated to the exact location they are most needeat the right time.   

This story isn’t unique to network planning, the theme is repeated in all industries.  What is needed is a paradigm shift.  You can no longer manually tune each individual forecast; the volume is prohibitive.  

You need a solution that enables all user personas (e.g. coder and non-coder) to be productive in building models and producing high quality forecasts, as well as process the data in a distributed fashion for speed.  Automation helps produce forecasts with a high degree of accuracy. 

Figure 1. Illustrative example of network complexity

Forecasting diagram

At this point, it is not just about building the forecasting model.  It’s so much more than that.  It’s also about integrating the forecasts into the business planning process.  A solution is needed that can enable many users to interact with the results and add incremental processing as needed.  The ability to share the results and the ability for business users to consume them is also key.   

Companies are challenged because they don’t want to lose the effort they’ve put in and continuity may be a concern, but they know that they’ve hit limitations of their current open source approach.  


This is where SAS comes in: when you have hit the limits of your open source approach in an enterprise forecasting journey.  Understandably, you don’t want to lose all the great work you’ve invested in building open source forecast models.  And you don’t have to.  SAS extends open source models by integrating them into SAS Visual Forecasting.   

Fundamental to open source integration with SAS Visual Forecasting is the TSMODEL procedure and the EXTLANG package.  TSMODEL is the underlying procedure to SAS Visual Forecasting.  The EXTLANG package enables seamless integration of external languages, including Python and R. 

TSMODEL and EXTLANG improve upon an open source forecasting strategy.  Through these techniques, SAS offers a distributed, scalable, and resilient way to run open source models.  In every step of the analytics life cycle, from data preparation to model development to model deployment, SAS enhances open source forecast models.  SAS is not only open to Python or R models, SAS accelerates open source by automatically distributing the workload for you.   We’ll pause for a moment and let that sink in…  In other words, SAS improves open source models by making them run faster.  Plus, you get a lot of other proven, valuable functionality that comes with it.  Intrigued?  Let’s read on. 

(Go to Part 2)


About Author

Mike Gilliland

Product Marketing Manager

Michael Gilliland is a longtime business forecasting practitioner and formerly a Product Marketing Manager for SAS Forecasting. He is on the Board of Directors of the International Institute of Forecasters, and is Associate Editor of their practitioner journal Foresight: The International Journal of Applied Forecasting. Mike is author of The Business Forecasting Deal (Wiley, 2010) and former editor of the free e-book Forecasting with SAS: Special Collection (SAS Press, 2020). He is principal editor of Business Forecasting: Practical Problems and Solutions (Wiley, 2015) and Business Forecasting: The Emerging Role of Artificial Intelligence and Machine Learning (Wiley, 2021). In 2017 Mike received the Institute of Business Forecasting's Lifetime Achievement Award. In 2021 his paper "FVA: A Reality Check on Forecasting Practices" was inducted into the Foresight Hall of Fame. Mike initiated The Business Forecasting Deal blog in 2009 to help expose the seamy underbelly of forecasting practice, and to provide practical solutions to its most vexing problems.

Back to Top