For years, the application of predictive analytics in supply chain management has been described as “transformative,” a “big opportunity,” the
“new business intelligence,” and even “the holy grail.”
However, in conversations, there is often confusion on
where and how predictive analytics can be applied.
The basic premise of supply chain management is
to control the manufacture, storage, transportation
and sales of goods and services to meet customer demand. Have it when they want it, keep nothing else.
Predictive analytics is about using a large amount of
data to gain insight into possible future scenarios and
their potential outcomes. I will focus on two major
areas where predictive analytics can be applied to
supply chain management—customer demand forecasting and inventory optimization.
CUSTOMER DEMAND FORECASTING
Demand forecasting can be defined as the prediction
of demand for a product or service, by customers. In the
results from their third annual supply chain survey—
Supply Chain Talent of the Future—Deloitte Consulting notes that demand forecasting was listed as one of
the top “fast-evolving technical capabilities” that is currently in use or expected to be used in the supply chain.
In fact, demand forecasting is second only to optimization tools, with 53 percent of respondents noting they
currently use demand forecasting and an additional 43
percent expect to use demand forecasting in the future.
In the article, Demand Forecasting: The Key to
Better Supply-Chain Performance, The Boston Consulting Group notes that traditional methods are often
powered by limited data, time intensive, and use outdated forecasting models that are typical of the one-size-fits-all variety.
THE IMPROVEMENTS TO DEMAND FORECASTING USING PREDICTIVE ANALYTICS ARE FAR REACHING:
Insight: The machine learning employed in predictive
analytics models allows for large amounts of struc-
tured ERP and supply chain management data to be
processed with seemingly disparate data such as con-
sumer sentiment data which derives from Facebook,
Twitter, Pinterest, Instagram and macroeconomic in-
dicators such as GDP, unemployment, Leading Indica-
tors Index, etc. Other disparate data include iOT device
data, demographics, weather and other domain-specif-
ic factors like production lines, engineering changes,
etc. However, predictive analytics is not just about col-
lecting data, but enabling actionable insight into how,
when, or why customers make purchases.
Speed: Near real-time collection of data combined
with machine learning allows predictive analytics to
be calculated within minutes or a few hours rather
than a 24/48-hour batch processing cycle that is often
used today. Microsoft has a great case study showcasing how Pier 1 Imports used predictive analytics
to understand and act on the customer shopping habits across their e-commerce site and more than 1000
“Pier 1 is a very data-rich company,” says Eric Hunt-
A Primer on Predictive Analytics
and Supply Chain Management