Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy
-
Upload
yinghui-liu -
Category
Documents
-
view
220 -
download
2
Transcript of Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy
![Page 1: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/1.jpg)
October 18, 2012 Webinar
Nada R. Sanders, Ph.D.
Iacocca Chair and Professor Department of Management
Lehigh University
[email protected] [email protected]
TAILORING YOUR FORECASTING METHODS TO IMPROVE ACCURACY
![Page 2: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/2.jpg)
October 18, 2012 Webinar
On-demand Webinars and Handouts
Today’s Webinar along with a PDF version of the slide set will be posted on our website next week.
All previously presented Webinars are available for viewing on www.forecastpro.com
Participants will receive an e-mail confirming availability after the Webinar and slide set are posted.
![Page 3: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/3.jpg)
October 18, 2012 Webinar
BIOGRAPHY
Nada R. Sanders, Ph.D.
Dr. Sanders is an internationally recognized expert on business forecasting, holding the Iacocca Chair at Lehigh University and is Principal at NRS Consulting. She has a Ph.D. and MBA from the Ohio State University and is author of over one hundred scholarly publications, including three books. She has served on the board of the International Institute of Forecasters (IIF) and was Associate Editor and co-founder of Foresight: The International Journal of Applied Forecasting. Dr. Sanders has extensive business experience offering consulting and training services. She has worked with companies including Nike, Mattel, AT&T, IDG, Schottenstein Corp., Cognitive TPG, MTC Corp., Ciba Corning, and numerous others.
![Page 4: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/4.jpg)
October 18, 2012 Webinar
Principles of Forecasting
Combining Disparate Data
Segmenting & Dissecting Data
Key Take-Aways
TAILORING YOUR FORECASTING METHODS TO IMPROVE ACCURACY
AGENDA
![Page 5: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/5.jpg)
October 18, 2012 Webinar
WHY FORECASTING?
Accurate forecasting provides a
significant competitive advantage
Good forecasts
correctly estimate demand
anticipate disruptions
identify new markets & trends
Poor forecasts
excess inventory
poor customer service
missed opportunity
![Page 6: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/6.jpg)
October 18, 2012 Webinar
Technology & software alone are not the
answer Need understanding of basic forecasting
principles
Rules to improve accuracy: I. Follow established forecasting principles II. Combine data from disparate sources
III. Segment & Dissect Data
HOW TO IMPROVE ACCURACY?
![Page 7: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/7.jpg)
October 18, 2012 Webinar
I. PRINCIPLES OF FORECASTING
1. FORECASTS ARE MORE ACCURATE FOR GROUPS OF ITEMS RATHER THAN FOR INDIVIDUAL ITEMS:
- Low level data are usually very irregular.
- Forecast errors at the low level are often many times greater than errors from forecasting aggregate series.
- Example: you cannot have the same accuracy for global PC sales versus sales of Media Tablets in one low-income market.
![Page 8: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/8.jpg)
Illustration of a Time Series
10
20
30
40
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
Legend
BENJERRY
Ben & Jerry’s Quarterly Sales
![Page 9: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/9.jpg)
A Monthly Time Series: Diamond Ring Sales
300
400
500
600
700
800
900
2003 2004 2005 2006 2007
DiamondRings2502 - 530772502
History
![Page 10: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/10.jpg)
A Weekly Time Series: Soft Drink Product Sales
5000
10000
15000
2005 2006 2007 2008
BRAINTREE_10Super_12oz12pkCan - None
History
![Page 11: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/11.jpg)
Use “Top Down” forecasting approach to leverage this principle:
Forecast group as a whole and each product
separately, then allocate the group total to the individual products proportionally.
Higher level: countrywide or product family
Lower level: individual region or item level Use same technique by location
I. PRINCIPLES OF FORECASTING
![Page 12: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/12.jpg)
October 18, 2012 Webinar
I. PRINCIPLES OF FORECASTING
2. CLEAN DIRTY DATA
▪ Missing Data Values
▪ Trading Day Adjustments
▪ Outliers (or Special Event) Adjustments
▪ Constant versus Current Price Data
![Page 13: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/13.jpg)
October 18, 2012 Webinar
I. PRINCIPLES OF FORECASTING
3. SOMETIMES THERE IS NO DATA
▪ Look to historical analogies
▪ Identify “like” products and study their patterns
▪ Use “structured analogy process”
![Page 14: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/14.jpg)
October 18, 2012 Webinar
I. PRINCIPLES OF FORECASTING
4. MATCH METHOD TO DATA
DATA = PATTERN + RANDOMNESS Type of Data Pattern
- Level – data fluctuates around a mean
-Trend – data exhibits increasing or decreasing pattern
-Seasonality – pattern that regularly repeats
-Cycle – patterns caused by politics, economy, etc.
-Noise – Randomness
![Page 15: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/15.jpg)
October 18, 2012 Webinar
1. QUALITATIVE (MANAGERIAL) FORECASTS
Judgmental, subjective, based on opinions.
2. QUANTITATIVE FORECASTS (Analytics)
Objective, based on mathematics and statistics.
They have complementary strengths
II. COMBINE DISPARATE DATA
![Page 16: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/16.jpg)
October 18, 2012 Webinar
Highly responsive to latest changes in
environment.
Can include “inside” and “soft”
information difficult to quantify.
Can compensate for “one-time” or unusual
events.
MANAGERIAL FORECASTS - STRENGTHS
![Page 17: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/17.jpg)
October 18, 2012 Webinar
Human cognitive limitations:
limited attention span
short-term memory
difficulty in understanding causal relationships
Biases:
lack of consistency
optimism
wishful thinking
political manipulation
MANAGERIAL FORECASTS - WEAKNESSES
![Page 18: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/18.jpg)
October 18, 2012 Webinar
Objective
Consistent
Can process large amounts of data
Can consider many variables and complex
relationships
QUANTITATIVE FORECASTS - STRENGTHS
![Page 19: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/19.jpg)
October 18, 2012 Webinar
Slow to react to changing environments.
Only as good as the model and data it is based upon.
Can be costly and time consuming to model ‘soft’ information.
Requires technical understanding.
QUANTITATIVE FORECASTS - WEAKNESSES
![Page 20: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/20.jpg)
October 18, 2012 Webinar
Each method has its strengths and
weaknesses
Best forecasting method is one that integrates both approaches
Combining
MANAGERIAL VERSUS QUANTITATIVE FORECASTS
![Page 21: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/21.jpg)
October 18, 2012 Webinar
1. Mechanical combining
METHODS OF COMBINING
Quantitative Forecast
F t = f (x1 ….xn)
Combine Forecasts
Managerial
Forecast
Final Forecast
F t+k
![Page 22: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/22.jpg)
October 18, 2012 Webinar
Provides improved accuracy
Objective
May not provide user with sense of
“ownership”
MECHANICAL COMBINING
![Page 23: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/23.jpg)
October 18, 2012 Webinar
2. Managerial Input to Model Building
METHODS OF COMBINING
Managerial
Opinion
Identify patterns,
Select variables,
Define parameters
Quantitative Model
Ft = f (x1…xn)
Final Forecast
F t+k
![Page 24: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/24.jpg)
October 18, 2012 Webinar
Least subject to bias
Objective
Time consuming and slow to react
to change
MANAGERIAL INPUT TO MODEL BUILDING
![Page 25: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/25.jpg)
October 18, 2012 Webinar
3. Managerial Adjustment
METHODS OF COMBINING
Quantitative Forecast
Ft = f (x1……xn) Adjust Forecast Final Forecast
F t+k
Managerial
Opinion
![Page 26: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/26.jpg)
October 18, 2012 Webinar
Most popular method in practice.
Enables rapidly reacting to change.
91% of companies report using it.
Has high potential for introducing bias.
Provides user with sense of “ownership.”
MANAGERIAL ADJUSTMENT
![Page 27: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/27.jpg)
October 18, 2012 Webinar
1. ADJUST IF THERE IS DOMAIN KNOWLEDGE
Managerial adjustment with domain
knowledge can improve accuracy.
Managers with higher understanding of market
conditions generate better adjustments.
Adjustment without domain knowledge can reduce
accuracy.
WHEN TO ADJUST QUANTITATIVE FORECASTS
![Page 28: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/28.jpg)
October 18, 2012 Webinar
DOMAIN KNOWLEDGE IS KEY
■ Blind extrapolation of past data can be misleading:
If 165,000 people were living in Las Vegas in l980
260,000 in l990
480,000 in 2000
Can you assume a trend?
In 1910 a Bell telephone statistician: “Every woman in
America will be employed as a switchboard operator.” It
clearly missed automated switching.
![Page 29: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/29.jpg)
October 18, 2012 Webinar
2. ADJUST IN SITUATIONS WITH HIGH DEGREE OF UNCERTAINTY
Quantitative models cannot deal with discontinuities or pattern changes
Experienced managers found superior to quantitative models when estimating:
onset, duration and magnitude of change
WHEN TO ADJUST QUANTITATIVE FORECASTS
![Page 30: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/30.jpg)
October 18, 2012 Webinar
3. ADJUST WHEN THERE ARE KNOWN CHANGES IN THE ENVIRONMENT
Compensate for specific events not captured by the model (e.g. advertising campaign or delayed shipment).
Adjust to compensate for past events that are not expected to re-occur in the forecast horizon (e.g. labor strike).
WHEN TO ADJUST QUANTITATIVE FORECASTS
![Page 31: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/31.jpg)
October 18, 2012 Webinar
1. STRUCTURE THE ADJUSTMENT PROCESS
Decompose process into subtasks
(e.g. decompose a time series into trend, seasonal, and random components.)
Can be computer aided or pencil and paper.
Today’s software enables adjustment
HOW TO ADJUST QUANTITATIVE FORECASTS
![Page 32: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/32.jpg)
October 18, 2012 Webinar
2. CONSIDER METHOD OF DATA PRESENTATION
Graphical presentation is better for trend data, tabular otherwise.
Graphical is better for short term forecasts, tabular for long term.
Graphical is better for macro data, tabular for micro data.
HOW TO ADJUST QUANTITATIVE FORECASTS
![Page 33: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/33.jpg)
October 18, 2012 Webinar
3. DOCUMENT ALL ADJUSTMENTS
Keep record of adjustments made and
reasons for adjusting.
Process serves as feedback and aids
learning.
HOW TO ADJUST QUANTITATIVE FORECASTS
![Page 34: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/34.jpg)
October 18, 2012 Webinar
4. MONITOR FORECAST ACCURACY OF ALL ADJUSTMENTS
Measure adjustments with formal forecast
accuracy measures
Process provides feedback
HOW TO ADJUST QUANTITATIVE FORECASTS
![Page 35: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/35.jpg)
October 18, 2012 Webinar
III. SEGMENT & DISSECT DATA ▪ All forecasts are not the same ▪ Best strategy is to segment series and use
differential approaches to each segment ▪ Segmentation can be based on: - ABC Analysis - Sales - Promotions - Degree of Importance
![Page 36: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/36.jpg)
October 18, 2012 Webinar
III. SEGMENT & DISSECT DATA
▪ Segment series based on “forecastability”
▪ Consider measuring “forecastability” of series
by coefficient of variation.
CV = Standard Deviation/ Mean
Rule of Thumb: CV< 20% stable;
CV>50% unstable; CV>100%
unforecastable
![Page 37: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/37.jpg)
October 18, 2012 Webinar
WHICH MODEL FOR WHICH DATA?
Low CV High CV Stability - CV
Degree of
Importance
High
Low
Combination Models
(Adjustment)
Multivariate Models (Model
Building)
Automate (Time Series)
Managerial Forecasts
Combination Models
(Adjustment)
Combination (Mechanical)
Automate (Time Series)
![Page 38: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/38.jpg)
October 18, 2012 Webinar
III. SEGMENT & DISSECT DATA
▪ Consider segregating series by stability and
data pattern
Segregate by stable and unstable; type of data pattern
Process tends to increase overall accuracy
▪ Consider disaggregating the variable to be
forecast.
Separate variable to be forecast and forecast them separately
![Page 39: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/39.jpg)
October 18, 2012 Webinar
TO IMPROVE FORECAST ACCURACY:
Do not ‘blindly’ apply technology and software
Understand & follow established forecasting principles
Understand & work with today’s software capabilities
KEY TAKE-AWAYS
![Page 40: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/40.jpg)
October 18, 2012 Webinar
KEY TAKE-AWAYS
Combine data from disparate sources
- Qualitative & Quantitative Forecasts are not mutually
exclusive: don’t use one to the exclusion of the other.
- Combining often gives best forecast accuracy.
Segment & dissect data for best results
- Time permitting, experiment with a variety of
forecasting models and a variety of segmentation strategies.
![Page 41: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/41.jpg)
October 18, 2012 Webinar
Remember: One size does not fit all.
THANK YOU
KEY TAKE-AWAYS
![Page 42: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/42.jpg)
October 18, 2012 Webinar
BFS offers forecasting Webinars and product training workshops
On-site, and remote-based (via WebEx) classes are available
Visit www.forecastpro.com to sign up!
TRAINING AND WORKSHOPS
![Page 43: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/43.jpg)
October 18, 2012 Webinar
Examples from today’s Webinar used Forecast Pro
To learn more about Forecast Pro:
– Request a live WebEx demo for your team (submit your request as a question right now)
– Visit www.forecastpro.com
– Call us at 617-484-5050
FORECAST PRO
![Page 44: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/44.jpg)
October 18, 2012 Webinar
What Now? How to Proceed When Automatic Forecasting Doesn’t Work
Date: Thursday, January 17 at 1:30 pm EST (GMT/UTC -5hrs)
Presented by Eric Stellwagen, CEO of Business Forecast Systems, Inc.
Visit www.forecastpro.com to sign up!
OUR NEXT WEBINAR
![Page 45: Forecast Pro Tailoring Your Forecasting Methods to Improve Accuracy](https://reader036.fdocuments.us/reader036/viewer/2022062412/577ccf041a28ab9e788eaaa4/html5/thumbnails/45.jpg)
October 18, 2012 Webinar
QUESTIONS?