X

Resources and guidance for supporting employees, customers, and partners during this unprecedented health crisis.

  • September 24, 2020

Forecasting the Impact of Sudden Change in Demand

Tom McDonough
Sr. Director, Product Marketing, Supply Chain Planning
This is a syndicated post, view the original post here

The rise of the COVID-19 virus in 2020, as well as increasingly regular ‘black swan’ events which can be weather or geo-politically driven makes it increasingly important for Demand Planners to develop agile and resilient responses to adjust forecasts to the new normal.

In discussion with customers through the first six months of the pandemics spread, we’ve heard stories of independent demand fluctuating wildly.  Many companies saw a pause or initial drop in demand in March 2020 as customers began to try to understand the impact of the virus on their lives and businesses, followed in some cases by significant increases in demand which led to retail stockouts (view the BISSELL webinar with Oracle from July 2020).  

Many companies manage a diverse portfolio of products and sales channels and, while demand may have increased for some product lines, others saw demand drop to a very small percentage of expected sales. An industry analyst relayed to us the story of one large candy manufacturing company which saw retail grocery sales surge while simultaneously watching cinema sales, 30% of their channel sales, fall to zero.

Improved forecast accuracy enables planners to reduce end-to-end variability and optimize inventory and production efficiency. The question on many planners minds however is “how do we adapt to events like the COVID-19 pandemic and maintain the relevance of forecasting?”. One approach is to develop models which include the influences and causal factors as insights to these factors grows.

Modern demand forecasting programs, such as Oracle Cloud Supply Chain Planning, typically use machine-learning algorithms to predict future demand based on historical demand and specified patterns like seasonality, holiday and price changes. In addition, the analytical engine takes into account long-term trends and levels such as an increase or decrease in demand over time. The recent demand shock from the COVID-19 pandemic does not fit the predefined patterns but the machine learning algorithms in Oracle Cloud Demand Management can be guided to capture these new impacts with additional tuning and causal factors.

Oracle Cloud Demand Management facilitates the approach to adapting forecasts through

  1. Enabling new causals to be defined that extend the machine learning models search for new patterns
  2. Enabling causal decomposition that provides insight into the cause/effect information and visualization of the impacts
  3. Allowing customers to identify which items were impacted by demand shock
  4. Providing new cross-validation feature which builds more robust models and so improves the overall quality of the results when multiple causals are used

Oracle Cloud customers can leverage this white paper, Forecast Impact of Sudden Change in Demand, as a start. The whitepaper along with a wealth of other information such as user discussion forums, product roadmaps, events and learning assets are available to Oracle Cloud customers via the Customer Connect Portal (requires registration and login credentials). 

The paper provides step by step instructions and recommended configuration modifications to capture sudden demand changes in forecasts. The demand forecasting service uses a supervised machine learning algorithm that can identify given patterns to predict the forecast. Supporting files are provided which already contain the measures and data examples of how to fill in the demand shock values and while you will need to modify the files based on your source system, the white paper provides the level of detail required to begin modeling.

You can find details of Oracle Cloud Supply Chain Planning products via our website or contact Oracle sales.

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.