1.

Introduction to business
forecasting and predictive analytics

Dr. Pradeep Kumar Dadabada

Asst. Professor, Information Systems & Analytics
[email protected]
Contents at glance…?
1.1. All about business forecasting
1.2. All about predictive analytics
1.1. All about business forecasting
What is Forecasting?
 Forecasting refers to the practice of predicting what will happen in the future by taking into
consideration events in the past and present.
https://corporatefinanceinstitute.com/resources/knowledge/finance/forecasting/

 Forecasting in today’s business world is becoming increasingly important as firms focus on increasing
customer satisfaction while reducing the cost of providing products and services. In virtually every
decision they make, executives today consider some kind of forecast.

 Part of successful business leadership comes from an ability to foresee future developments and to make
the right decisions. Forecasting can be used as a tool to guide such business decisions, even though
some level of uncertainty may still exist.

 Top management is generally interested in making decisions based on forecasting economic factors that
are critical in strategic planning and action. While forecasters will not be completely certain of what will
happen in the future, they can reduce the range of uncertainty surrounding a business decision.

 But, remember:
 Forecasts are seldom perfect
 Most techniques assume an underlying stability in the system
Importance of Forecasting
 A good forecast must be meaningful, reliable, accurate, and timely
 Business decisions almost always depend on some forecast about
the course of events. Virtually every functional area of business
makes use of some type of forecast. For example:
1. Accountants rely on forecasts of costs and revenues in tax planning.
2. The personnel department depends on forecasts as it plans
recruitment of new employees and other changes in the workforce.
3. Financial experts must forecast cash flows to maintain solvency.
4. Production managers rely on forecasts to determine raw-material
needs and the desired inventory of finished products.
5. Marketing managers use a sales forecast to establish promotional
budgets.
How much serious is a bad prediction?
 As reported by ABC, the BBC, the Wall Street Journal, the New York Times, and
elsewhere, an Italian court has convicted seven earthquake experts of failing to
appropriately sound the alarm bell for an earthquake that wound up killing more
than 300 people in L’Aquila in 2009. The experts received long prison sentences and
fines of more than $10 million.
https://freakonomics.com/2012/10/23/whats-wrong-with-punishing-bad-predictions /
 The meteorologist, who spotted a storm cell just north of Rio and another to the
south, predicted a horrendous storm on Dec. 31. He went so far as to warn Rio
residents, known as Cariocas, to stay at home, batten down the hatches and avoid the
beaches. His forecast was dead wrong. As it turned out, a subtropical jet stream
whisked the threatening storm out to sea. The skies that night were crystal clear.
 "I've been forecasting for 35 years and I've made some mistakes. But I've also been
right a lot of the time. The records show I'm accurate nearly 90 percent of the time,"
Austin said. "I just wanted people to be safe — and now I'm being sued?“ Accused
of ‘Sounding a False Alarm’ https://
abcnews.go.com/WNT/story?id=130551&page=1
Types of Forecasting methods
 Qualitative Forecasting methods
Type Characteristics Strengths Weaknesses
Executive A group of managers Good for strategic or One person's opinion
opinion meet & come up with new-product can dominate the
a forecast forecasting forecast

Market Uses surveys & Good determinant of It can be difficult to

research interviews to identify customer preferences develop a good
customer preferences questionnaire

Delphi Seeks to develop a Excellent for Time consuming to

method consensus among a forecasting long-term develop
group of experts product demand,
technological
changes, and
Quantitative forecasting methods
 Time Series Models:
 Assumes information needed to generate a forecast is
contained in a time series of data
 Assumes the future will follow same patterns as the past

 Causal Models or Associative Models

 Explores cause-and-effect relationships
 Uses leading indicators to predict the future
Forecasting Time Horizons
 Short-term forecast
 Up to 1 year, generally less than 3 months
 Purchasing, job scheduling, workforce levels, job
assignments, production levels
 Medium-term forecast
 3 months to 3 years
 Sales and production planning
 Long-term forecast
 3+ years
 New product planning, facility location, R&D
Example 1: Krispy Kreme
 During summer 1937 the first Krispy Kreme doughnuts were sold in Winston- Salem,
North Carolina. Since that time the company has grown and spread well beyond the
borders of North Carolina.
 Krispy Kreme’s operations had expanded to a point that it recognized the need for a
new multiple forecasting system to provide information related to production
requirements based on demand forecasts and to provide financial forecasts.
 It identified three major drivers of its business: new stores, new off-premises
customers that make Krispy Kreme products available through retail partners, and
seasonal factors.
 For new stores, forecast models were developed for the opening week through sales 18
months out. Sales are related to such factors as general population growth, brand
awareness, foot traffic, and display locations.
 Each month a series of conference calls with market operators are used to gather
information for the forecasting models.
 Meetings with executive-level managers are also held on a monthly basis to
communicate forecast information. This process has led to forecasts with errors of only
plus or minus 1 percent.
Example 2: Bell Atlantic
 At Bell Atlantic, the forecasting process begins with the
collection of historical data on a monthly basis. These data are
saved for both service classifications and geographic regions.
 The Demand Forecasting Group at Bell Atlantic developed a
data warehouse so that the data can be shared and integrated
across the entire corporation.
 In preparing forecasts, subjective forecasting methods are used
along with time-series methods, and regression modeling based
on economic, demographic, and other exogenous variables.
 The forecasts are continually monitored and compared with
actual results monthly and annually to ensure that Bell Atlantic
meets customer needs.
Example 3: Columbia Gas
 Columbia Gas of Ohio (Columbia) is a large natural gas utility that
delivers over 300 billions of cubic feet (BCF) of natural gas annually.
 Columbia develops two kinds of forecasts, which it refers to as the Design
Day Forecast and the Daily Operational Forecast. The former is used to
determine gas supply, transportation capacity, storage capacity, and related
measures. This forecast is used primarily for supply and capacity
planning. Over a seven-year period the average mean absolute percentage
error in its Design Day Forecast was 0.4 percent.
 The Daily Operational Forecast is used primarily to ensure that supplies
are in balance with demand over five-day spans. As would be expected,
the average errors for these shorter term forecasts have been higher at
about 3 percent. The forecasts are based to a large degree on regression
models in which demand is a function of such variables as current-day
temperatures, previous-day temperatures, wind speed, and day of the
week.
Example 4: Segix Italia
 Segix Italia is a pharmaceutical company in Italy that produces
products that are sold domestically and are exported to countries in
Europe, such as Belgium, Holland, Germany, and England, as well
as to African, South American, Asian, and Middle Eastern countries.
 The forecasting function at Segix is housed within the marketing
group, and forecasts are reviewed by the marketing director and the
sales director, both of whom may make subjective adjustments to
the forecasts based on market forces not reflected in the original
forecasts.
 The forecasts are prepared monthly for seven main prescription
drug products. The monthly forecasts are then aggregated to arrive
at annual forecasts. These forecasts are used to develop targets for
sales representatives.
Example 5: Fiat Auto
 Top management at Fiat considers the forecasting function as an
essential aspect of its decision-making process. Midway through the
1990s Fiat was selling over 2 million vehicles annually and employed
some 81,000 people in Italy and about another 38,000 overseas.
 All functional areas in the company make use of the forecasts that are
prepared primarily in the Planning, Finance, and Control Department
and in the Product Strategy Department.
 Macroeconomic data such as gross domestic product, the interest rate,
the rate of inflation, and raw-material prices are important inputs in
Fiat’s forecasting process.
 At Fiat forecasts are first prepared for total sales of vehicles, engines,
and gears, and then broken down to specific stock-keeping units
(SKUs). Sales are measured by orders rather than shipments because
its system is customer-driven.
Example 6: Brake Parts, Inc.
 Brake Parts, Inc. (BPI), is a manufacturer of replacement brake parts for both foreign and
domestic cars and light trucks.It has nine manufacturing plants and seven distribution
centers in the United States and Canada. Overall, BPI has roughly 250,000 SKUs at
various distribution locations (SKULs) to forecast.
 The development and implementation of a multiple forecasting system (MFS) has saved
BPI over $6 million per month, resulting from sales not being lost due to stockouts.
 The MFS at BPI uses up to 19 time-series forecasting techniques, such as a variety of
exponential smoothing methods, and causal regression models in tandem. Forecasts are
first developed with a time-series method, and then the errors, or residuals, are forecast
using regression. The two forecasts are then added together and provided to management
in a form that allows management to make subjective adjustments to the forecasts.
 Forecasts are evaluated using three measures: percent error (PE), mean absolute percent
error (MAPE), and year-to-date mean absolute percent error (YTD MAPE). The first two
of these are common error measures, but the third is somewhat unique. The YTD MAPE is
used to give management a feeling for how each forecast is performing in the most current
time frame. The PE and MAPE contain errors that may have occurred at any time in the
historical period and thus may not reflect how well the method is working currently.
Forecasting research
 In 1982, the International Institute of Forecasters set up
the Journal of Forecasting, followed in 1985 by
the International Journal of Forecasting. 
 The primary aim of their foundation, laid out in the first
issues, was to take a “multi-disciplinary perspective”; all
types of forecasting methods were of interest. Of
particular importance were “papers that compare[d]
different approaches to actual forecasting situations”.
1.2. All about Predictive Analytics
Data Analytics
 The word “analytics” has replaced the previous individual
components of computerized decision support technologies that
have been available under various labels in the past. Indeed, many
practitioners and academics now use the word analytics in place of
BI.
 Although many authors and consultants have defined it slightly
differently, one can view analytics as the process of developing
actionable decisions or recommendation for actions based upon
insights generated from historical data.
 The Institute for Operations Research and Management Science
(INFORMS) has created a major initiative to organize and promote
analytics.
 According to INFORMS, analytics represents the combination of
computer technology, management science techniques, and
statistics to solve real problems.
Data Analytics
Data Analytics
Predictive Analytics
 In its simplest form, predictive analytics, or advanced data analytics, is the
process of coupling historical data with statistical algorithms/machine
learning/deep learning to make predictions about future events.
 Predictive analytics is heavily used in a variety of ways by our largest
corporations and industries.
 Some predictive analytics examples include:

 Netflix leverages a complex web of predictive analytics algorithms to drive their

content recommendations.
 Zillow leverages various forms of quantitative methods to estimate house listing
prices. This data provides useful market comparisons to their users.
 Financial services use big data and real time statistical models to identify fraud
and combat money laundering.
 Utility companies forecast the amount of energy their users will consume to help
with load management and ensure they keep their grids sufficiently supplied.

 A good reading: https://hbr.org/2014/09/a-predictive-analytics-primer

Predictive Analytics
Predictive Analytics
Conclusion
 Understanding of Business Forecasting
 Understanding of Predictive Analytics