Professional Documents
Culture Documents
Forecasting Technological Change
Forecasting Technological Change
02/19/17 1
Sessions
Introduction
Trend Analysis Techniques
Expert Opinion Techniques
Integrative Techniques
Closing
02/19/17 2
Sessions
Introduction
Trend Analysis Techniques
Expert Opinion Techniques
Integrative Techniques
Closing
02/19/17 3
2. Trend Analysis Techniques
Shape of S-Curve
Dependent on utility function
Based on historical data If you dont know where you
(time or experience) are, and how you got there,
how can you possibly
Not causal know where you are going?
Reproducible Abe Lincoln
02/19/17 7
Future Forecast
First Industrial Revolution (1760 - 1860)
Second Industrial Revolution (1860 -
1950)
Third Industrial Revolution (1950 -
2020)
02/19/17 8
Nature of Work
Source: Snyder, et
al, The Strategic
Context of
Education
in America 2000 -
2020
02/19/17 9
Information Technology
Source: Jeremy
Greenwood, The
Third Industrial
Revolution
02/19/17 10
Third IR Technologies
DP HP CP NP
Cell Phones
PC
Internet
Sattelite Communications
IC
Technology
Fiber Optics
Programming
Electronic Computer
Transistor
TV
Telephone
02/19/17 14
Precursor Analysis
Identify lead-lag relationship
Obtain lead-lag data
Decide whether lead-lag relationship
will continue
Causal connections are best
Similar control factors are next best
Accidental relationships are misleading
02/19/17 15
Techniques
Analogy
Precursor Developments
Trend Extrapolation
Limit Curve
Learning Curve
Substitution Analysis
Multiple Substitution Analysis
02/19/17 16
Trend Extrapolation
If
Technical parameter has utility
Market continues to value utility
Driving forces remain same
Not approaching limit
Then
Most technologies follow a pattern of constant
percentage increase
y = y0emt i.e.(log y = log y0 + m t ) (natural log)
02/19/17 17
Trend Analysis
02/19/17 18
First Technology Trend Forecast
100
B-58
90
Number of Electronic Component Parts
80
70
(Thousands)
60
50
B-52
40
30
20 B-47
10
B-17 B-29 Glocal Vantage, Inc.
B-50
0
1940 1945 1950 1955 1960 1965
Year
02/19/17 19
Internet Users
900
800
eTForecasts
700 (2000)
Computer
Internet Users (WW, Millions)
600 Industry
Almanac
(2000)
500 Computer
Industry
Almanac
(1998)
400
Tapscott
(1996)
300
US
DOC(1999)
200
100
0
1980 1985 1990 1995 2000 2005
Year
02/19/17 20
Internet Users
100000
10000
World's Population
Internet Users Worldwide (Millions)
1
Trend Analysis (2001)
World's Population
0.1
0.01
02/19/17 21
Internet Users Analysis
Growth
114% increase per year
doubling time = 11 months
Surpass worlds population in 2005
Is it a utility function?
Limits to growth?
Whats a user?
02/19/17 22
IC Example
www. intel.com/research/silicon/mooreslaw.htm
02/19/17 23
Microprocessor Speed
02/19/17 26
Genetic Sequencing
Source:
www.ncbi.nih.gov/Genbank/
genbankstats.html
02/19/17 27
Genetic Sequences
100,000,000
10,000,000
Number of Sequences (millions)
1,000,000
100,000
10,000
1,000
100
1982 1987 1992 1997 2002
Year
Source: Enriquez Doubling every 17 months
02/19/17 28
Chromosome Mapping
Source: www.ornl.gov/hgmis/posters/chromosome
02/19/17 29
Trend Extrapolation
Specify assumptions
Include quantitative evidence
Follow logical approach
Use prior rates of improvement unless
There is a limit
Utility is changing
Approach is changing
Driving forces are changing
Document reasons
Useful to develop alternative projections
Provide a basis for rational discussion
02/19/17 30
Techniques
Analogy
Precursor Developments
Trend Extrapolation
Limit Curve
Learning Curve
Substitution Analysis
Multiple Substitution Analysis
02/19/17 31
Limit Curve
Organic growth model
Often used when technology approaches a
natural limit
Limit may not be real
Use limit and make forecast
Review limit - real?
Shifts usually occur as limit is approached if
utility still exists
02/19/17 32
Growth Curve
350
300
Eggs/Chicken Year
250
200
150
100
1930 1940 1950 1960 1970 1980 1990 2000
Year
02/19/17 34
Limit Curve
400
350
300
Eggs/Chicken Year
250
200
150
100
1930 1940 1950 1960 1970 1980 1990 2000
Year
02/19/17 35
Limit Curve
400
350
300
Eggs/Chicken Year
250
200
150
100
1930 1940 1950 1960 1970 1980 1990 2000
Year
02/19/17 36
Limit Curve
400
350
300
Eggs/Chicken Year
250
200
150
100
1930 1940 1950 1960 1970 1980 1990 2000
Year
02/19/17 37
Limit Curve
Pearl curve
y = L / (1+a e -bt)
L = 350
a = 1.71
b = .038
t = 0 in 1937
02/19/17 38
Internet Users
10000
1000
100
Number (millions)
10
Data
Forecast
1
1980 1985 1990 1995 2000 2005 2010 2015
0.1
0.01
0.001
Year
02/19/17 39
Limit Curves
Pearl
Fisher-Pry
Gompertz
Bertalanffy
02/19/17 40
Limit Curves
Pearl
y = L / (1 + a exp (- b t))
Fisher-Pry y = L / (1 + 10(A - B t))
Gompertz
Bertalanffy L = limit
a = controls location
b = controls shape
02/19/17 41
Limit Curves
Pearl f = 1/2 (1 + tanh a (t - t0))
Fisher-Pry
Gompertz f = fraction of applications
Bertalanffy in which the new technology
has been substituted for the old
t0 = time for 50% substitution
a = controls shape
02/19/17 42
Limit Curves
Pearl
Fisher-Pry Y = L exp (- b exp ( - k t))
Gompertz
Bertalanffy Non symmetrical
02/19/17 43
Limit Curves
Pearl
y = L + (y0 - L) exp (- b t)
Fisher-Pry
Gompertz L = limit
Bertalanffy y0 = initial value
b = controls shape of curve
02/19/17 44
Envelope Curve
02/19/17 46
Envelope Curve
02/19/17 47
Techniques
Analogy
Precursor Developments
Trend Extrapolation
Limit Curve
Learning Curve
Substitution Analysis
Multiple Substitution Analysis
02/19/17 48
Learning Curve
Some improvements can be forecast
better based on experience rather than
time.
Frequently production costs follow this
type of relationship
y = a x- b
log y = log a - b log x
02/19/17 49
Learning Curve
Frequently experience is equivalent to
cumulative production quantity
Usually expressed as percent
improvement for each doubling of
experience
5% cost reduction in 2x
95% learning curve
02/19/17 50
Learning Curve Example
1915
1.4
1925
1.2
1935
1 1920 1930
1940
0.8 1945
Data
0.6
1955
1950
0.4 1965
1960
0.2
1970
0
0.000 0.500 1.000 1.500 2.000 2.500 3.000 3.500 4.000
Log (cumulative residential consumption in terawatt hours)
02/19/17 52
Residential Electric Power Learning Curve
1.6
Log (cents per kilowatt hour in 1967 dollars)
1915
1.4
1925
1.2
1935
1 1920 1940
1930
Data
0.8
Forecast
66% Learning Curve 1945 1955
0.6
1950
1965
0.4
1960
0.2 1980
1970
1975 1985
0
0.000 0.500 1.000 1.500 2.000 2.500 3.000 3.500 4.000 4.500
Log (cumulative residential consumption in terawatt
hours)
02/19/17 53
Residential Electric Power Learning Curve
1.6
Log (cents per kilowatt hour in 1967 dollars)
1915
1.4
1925
1.2
1935
1 1920 1940
1930 Data
0.8
Forecast
1945 1955
0.6
1950
1965
0.4
1960
0.2 1980
1970
1975 1985
0
0.000 0.500 1.000 1.500 2.000 2.500 3.000 3.500 4.000 4.500
Log (cumulative residential consumption in terawatt
hours)
02/19/17 54
Techniques
Analogy
Precursor Developments
Trend Extrapolation
Limit Curve
Learning Curve
Substitution Analysis
Multiple Substitution Analysis
02/19/17 55
Fisher-Pry Substitution Analysis
0.9
0.8
a=1
0.7
Ratio (new / old)
0.6
0.5 t0 = 3
0.4
0.3 a = 0.5
0.2
a = 1.5
0.1
0
0 1 2 3 4 5 6 7 8
Time Units
02/19/17 57
Substitution of Steam for Sail
100%
90%
80%
Steam Gross Tonnage
70%
60%
Sail
50%
40%
Steam
30%
20%
10%
0%
1780 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980
Year
02/19/17 58
Steam/Sail Fisher-Pry Substitution
100
10
Steam/Sail
Data
1 Fisher-Pry Model
0.1
0.01
Year
02/19/17 59
CATV Substitution in US
10
Homes with CATV/Homes without CATV (US)
1
1950 1960 1970 1980 1990 2000
Data
0.1
Forecast
0.01
0.001
Year
Source: Martino, 1983
02/19/17 60
CATV Substitution (US)
10
Homes with CATV/Homes without CATV (US)
1
1950 1960 1970 1980 1990 2000
0.01
0.001
Year
02/19/17 61
Telecommunication Technology Substitution
1000
100
Stored Program/Ectromechanical
10
Data
Forecast
0.1
0.01
Year
Source: Lenz & Vanston, TFI 1986
02/19/17 62
Telecommunications Substitution Analysis
140
Electromechanical Systems (millions of
120
Total Market Annual Growth Rate
100 100%
subscribers)
80 80%
Electromechanical
60 60% Market
Total Market
40 40%
Growth Rate
20 20%
0 0%
0% 20% 40% 60% 80%
Stored Program % of Total Market
02/19/17 63
Techniques
Analogy
Precursor Developments
Trend Extrapolation
Limit Curve
Learning Curve
Substitution Analysis
Multiple Substitution Analysis
02/19/17 64
Multiple Substitution Analysis
02/19/17 65
Multiple Substitution Example
Source: Porter
02/19/17 67
Summary
Trend analysis has strong historical
validation
Models have been accepted but not
explained
Quantitative
Does not explain how change will occur
Often misused and misunderstood
02/19/17 68
Summary (cont.)
First noticed by Henry Adams in 1918
Knowledge begets knowledge as
money begets interest. - Arthur Conan
Doyle
02/19/17 69
IC Line Width
100
10
Ine Width (micrometers)
Data
1
Forecast
1965 1970 1975 1980 1985 1990 1995 2000 2005
0.1
0.01
Year
02/19/17 70
Trend Analysis
Analogy
Precursor
Developments
Trend Extrapolation
Limit Curve
Learning Curve Trend Change Formal
Analysis
Substitution Analysis Material
Surveillance Data, Insight
Multiple Substitution
Analysis
02/19/17 71
Glocal Vantage, Inc.
PO Box 161475
Austin, TX 78716
(512) 632-6586
paul.schumann@glocalvantage.com
www.glocalvantage.com
http://incollaboration.com
Twitter: innovant2003
02/19/17 72
Paul Schumann
Futurist and innovation consultant
Application of web 2.0 to market & strategic
intelligence systems
Web 2.0 tools & technologies
Application of web 2.0 to democratic processes
Broad perspectives on the future
Services
Strategic market research & technology forecasting
Intelligence systems consulting
Seminars, webinars & presentations
02/19/17 73
This work is licensed under
the Creative Commons
Attribution license. You may
distribute, remix, tweak,
and build upon this work,
even commercially, as long
as you credit me for the
original creation as Paul
Schumann, Glocal Vantage
Inc,
02/19/17 www.glocalvantage.com. 74