M,UT.
UBRARIE8 - DEWEY
J
tJewey
HD28
1^
MIT LIBRARIES
The International Center for Research, on the
Management of Technology
3 9080 00932 7732
Metrics for Managing Researcli and
Development
Marc H. Meyer*
Peter Tertzakian
James M. Utterback
WP # 124-95
April 1995
Sloan
WP # 3817
*Northeastern University
©
1995 Massachusetts Institute of Technology
Sloan School of Management
Massachusetts Institute of Technology
38 Memorial Drive, E56-390
Cambridge,
02139
MA
Abstract
The paper proposes methods
product development.
to
measure the performance of research and development
in
new
We base the work in the context of evolving product families in the
technology-based firm with the goal of more clearly understanding the dynamics and consequences
of platform renewal and product generation for long term success.
We test the proposed methods with data gathered from a large measurement systems manufacturer.
We find that the methods and the resulting measures provide management with a technique to
identify both the technological
and market leverage achieved from the
firm's present
and past
product platforms. This provides a foundation for transforming single product, single period
planning processes into a multi product, multi period form that embraces the product family and the
renewal of product architecture. The research also shows the need to integrate data from
engineering, manufacturing, and sales organizations to produce information for managing the
growth of the
firm's product families.
;MSSACnus£;'rrs institute
OF TECHNOLOGY
JUN 2
8 1995
LIBRARIES
When
should a firm renew the underlying technologies and designs of
will these efforts cost
and how long may they be expected
commercial benefits can the firm expect
its
to gain
essential characteristics of products
planning and management of
improvement
in the
way
in
What
from product redesign?
approaches and strategies for product development?
sustained success in product development that
to take?
we seek
How much
types of engineering and
How can
a firm improve
We
to address in this article.
creation. Systematic
which a firm creates new products
commercially successful product development. In
products?
These are fundamental questions for
and propose measurement methods
new product
its
is
turn, learning
will define the
to guide the firm in
its
and continuous learning and
the basis for
more rapid and
cannot be achieved without clear
and purposeful measurement.
1
.
Measurement
for Research
and Development
What measures
of performance can help firms better manage the development of
What might be
the purpose and goal of these measures ?
desired outcomes of effective
new product development.
of generating a continuous stream of
competitive position and die.
It
new
has been
shown
that the
1991
;
commonly
lie in
products ?
examining the
Productive, innovative firms are capable
products. If firms do not
the individual products, can be efficiently constructed
product architectures,
The answer may
new
make new
elements of
this
products, they lose
stream of products,
ie.
on successive generations of underlying
referred to as product platforms. (Sanderson and Uzumeri,
Wheelwright and Clark, 1992; Meyer and Utterback, 1993). Modular platforms (Uhrich
and Tung, 1991) with clear interfaces between embodied modules (Smith and Reinertsen, 1992)
can allow a firm
to rapidly
and
efficiently derive follow-on products
platforms. Clear understanding of customers' needs that
is
based on
its
product
then incorporated into platform and
follow-on product designs can generate high levels of sales to engineering resource productivity.
(Hauser and Clausing, 1988). As a product family evolves, platform architectures must be
continuously renewed and
new component technology
firm's offerings freshly competitive
The
integrated into these designs to keep the
(Meyer and Utterback, 1993;
lansiti, 1993).
purpo.se of measures of the effectiveness of research and development should be to
demonstrate firm performance
in this critical
way towards improvement. How do
dimension of new product development, pointing
firms measure
R&D performance observed in Moser's (1985)
R&D performance now
indu.stry
?
the
The measures of
survey were short-term in nature and
focused on single products or projects. These measures either qualitatively assessed goal
attainment or computed the variance between project plans and actual outcomes along the
dimensions of cost and time. Mo.ser found
that the
most commonly used
R&D metric
was
/'(/1,'C
/
"slip."
ie.
the
gap between expected and actual project time and budget. Slip
management's schedule predictions. Some firms
rate
measures the quality of
that use a stage-gate product generation process
(Cooper 1990) compute slippage between the development phases for respective products,
determining which phases encounter the greatest estimation
found
that
difficulty.
Gupta and Wilemon 1990)
(
poor or unclear product definition was the primary cause of delayed projects. The
measurement systems manufacturer whose data
successive phase slip for
market definition
to be
its
projects.
will
Management
be used
later in this
has found the
initial
paper currently tracks
stage of technical and
most problematic.
Accurate project estimation should not necessarily be construed as a predictor of successful
technical and commercial outcomes in product development. Mansfield's (1972) study of a
successful pharmaceutical
company showed
the firm experienced overruns
on
80%
of
its
new
products efforts, with an average slip of 1.78 over planned costs, and 1.61. on planned schedules.
(Mansfield, et
al.
1971) Yet, the firm prospered from these "late" products.
aluminum company's
efforts to create industrial products outside
its
A study of an
traditional smelting business
revealed that "tight schedules" for novel products led to poor financial outcomes (Utterback et
1992) Slip rate
is
also
products). This has
commonly used
little
to
to
examine
do with what occurs
the experiences of individual projects (or
in the
aggregate across the stream of
products generated by the typical product-making company.
the
al.
development of base architectures may take a long time
From
to
new
a product family perspective,
complete, but once finished,
may
serve as the foundation for rapid development of derivative products.
Numerous measures of R&D
in total
based on our
effectiveness have been proposed in the literature (some seventy-five
literature search
of the topic). Appendix
typology based on the purpose of measurement (Cordero
The purposes of various measurement may be seen
the sequential stages of the product
life
1
I
places these measurements in a
990) and the method of measurement.
as seeking to understanding performance in
cycle: measuring the effectiveness of product planning
before actual development, controlling projects during development, measuring the technical
outcomes of
the
completed development, measuring the commercial outcomes associated with
completed products, and
full life
cycle.
As shown
lastly,
in
understanding overall performance across the span of a product's
Appendix
the literature to
measure performance
assessments of
new products
in
1
in
,
the techniques of
measurement
that
in
these various stages included: (I) comparative
terms of success, (2) scoring models on product features relative
to competitors' products, (3) benefits contribution
methods based on the sales-related outcomes of
products. (4) various forms of performance to schedule analy.ses (the most
today being
we found proposed
total project slip or slip at different
commonly used form
key phases of a project), and (5) analyses of
l',i
w-
individual and organization technical accomplishment, including the counting and comparison of
patents awarded to
The impact of
are
that
members of a group
these various measures
is
for measuring
commercial outcomes on a quantitative
found
that
35%
that
basis.
A recent survey of 248 R&D executives by the Industrial
"measuring and improving
R&D productivity and effectiveness"
problem" for respondents (Industrial Research
(1994)
in his
Another survey of 700 manufacturing firms also
of the sample failed to systematically assess the performance of
and Panicelli, 1982).
showed
suspect, however. Schainblatt (1982) found that "there
R&D productivity that are not flawed." Of the firms
59% did not measure the R&D function and only 20% compared R&D costs to
no currently used systems
he studied,
or publications emerging from them.
Institute, 1993), a finding
R&D (Rockwell
Research
was
Institute
the "biggest
confirmed by Roberts
survey of corporate executives.
We believe that the limited utility of existing measures of R&D performance is that by their design
and implementation, these measures do not help management understand the longer term dynamics
of evolving product
lines, the
renewal of their underlying architectures, and the leverage that these
architectures provide in derivative products. Addressing this need
it
is
the key point of comparison between
literature.
It
market introduction and
to
the purpose of this paper, and
what we propose and the existing
should also be clearly noted that our primary focus
development as opposed
is
is
the
R&D performance
R&D phase of product
performance measurement for the manufacturing (Kaplan 1990),
distribution.
A truly
comprehensive picture would incorporate
these
all
dimensions.
We will
propose measures of
R&D performance compatible with the management model of
evolving product families. Data from a more than twenty year period within a large measurement
systems manufacturer
will then
be used to
product family and then measures of
test
and explain these measures.
R&D effectiveness suited for
it.
We
will first define the
These metrics
are then
applied to a sample of industrial measurement systems and interpreted within the management,
marketing, and technological factors influencing the evolution of these product families.
We then
conclude the paper with a discussion of the benefits, costs, and qualifications of the proposed
methods and suggest areas
make
a contribution to
for
complementary research. The overriding goal of the paper
management
practice in the understanding of the effectiveness of
product development that encompasses groups of individual products related by
is
lo
R&D for
common
architecture.
P</vc
.<
Meta- Metrics for the Product Family
3.
3.
1
Defining the Product Family
The
effectiveness of the firm's
new product
generation activity
lies in:
(1) its ability to create a
continuous stream of successful new products over an extended period of time; and
(2) the
attractiveness of these products to the firm's chosen markets.
Streams of new products generated by firms may be thought of as evolving product families.
A product family
is
defined as a set of products that share
related set of market applications. This
common
technology and address a
commonality may be validated
for a
group of products by
applying the technology and market newness criteria proposed by Meyer and Roberts (1986).
Successive products within a product family can be assessed by virtue of newness of embodied
An evolving
core technology and market applications.
technologies into product designs and target
change
will be
bounded.
"related" to older ones:
New
new
product family will introduce
new customer
new
requirements. However, the scope of
technologies and market applications of a product family will be
technologies will be combined with and enhance existing core
technologies and the resulting products will target existing and related market applications.
commonality of technologies and markets leads
distribution,
and service, where the firm
tailors
to efficiency
and effectiveness
in
The
manufacturing,
each general resource or capability to the needs of
specific market niches.
The
is
technological foundation of the product family
we
define as the product platform.
A
platform
the physical implementation of a technical design that serves as the base architecture for a
of derivative products. The efficiency of the platform
derivatives
is
logically
The concept of
in
.series
serving as a foundation for such
one key aspect of product family management.
platforms, product derivatives, and renewal
is
not new. Wheelwright and Clark
(1992) differentiated between platforms, their derivative products, and platform "extensions
renewals for vacuum cleaners; Meyer and Utterback (1993) did the same
electronic imaging systems and peripherals, as did Sanderson and
of portable cassette players, and Lehnerd
(
in their study
Uzumeri (1993)
strategy.
its
or
of
in the
evolution
1987) for power tools. These examples reflect more
recent industrial experience and practice. Chrysler's newly achieved success with
automobiles, and more recently,
"
Neon automobiles, was based on
Timely platform renewal has been
its
LH
.series
of
a clear platform development
essential to Intel's continuing success with
family of microprocessors. As soon as a particular microproces.sor platform
is
its
X8086
completed, and
I'lii^c
4
derivative products start to be generated from
the current platform with
new
it,
resources are immediately allocated to obsoleting
designs. Planning looks out into the future for several or
more
platform generations.
These cases can be generalized
Figure
new
1.
It
into an evolutionary
model
shown
for the product family as
in
represents a single product family that undergoes successive platform extensions, a
platform development, and the respective follow-on product developments of these platform
versions.
To
achieve sustained success, a firm must seek to continuously renew
architectures. This approach generates a set of engineering activities focused
that coexist with specific follow-on product
in
platform design
product family.
modules and
is
1
on platform renewal
here
and success of a
centrally important to sustaining the quality
A platform design consists of a basic
the interfaces
base product
developments based on existing product platforms.
Insert Figure
Excellence
its
architecture,
comprised of subsystems or
between these modules. Smith and Reinertsen (1992) made the analogy
of the subsystems being bricks, and the interfaces the mortar that binds them together. In an
assembled product such as a camera, subsystems include the shutter mechcinism, the lenses, the
various operator controls and focus mechanisms, the flash, the power source, and the camera
The
housing.
internal
fittings
subsystem
and electronics by which these subsystems are integrated serve as the
interfaces.
The mechanisms
for controlling shutter speed
and the aperture, for
rewinding and changing film, or for using a flash are the "user interfaces". The
one puts into the camera can also be viewed as having subsystems
which the film
is
roll
that include the spool
wrapped, the cover over the film, and the end caps. Even the film
comprised of various layers
(the subsystems)
of film that
combined through chemistry
in
itself
around
may be
a continuous
production process. The width of the spool, the length of the film leader, and the spacing of the
edge perforations on the film are external interfaces
that allow
many
types of films to be used
in
different cameras.
The notion of subsystems and
shown
o
in
interfaces leads to a typology of product family evolution that
Figure 2 and which mirrors the approach of Wheelwright and Clark (1992)
The
initial
is
:
plarfonn of a product family consists of the subsystems and subsystem
interfaces of the basic product design.
These subsystems are incrementally reengineered or
refined to generate specific product offerings.
Pdi-c
5
A platform extension occurs when (a) particular
o
subsystems within the existing platform
design are substantially changed and enhanced, and/or (b)
new subsystems
are added to the
design without disturbing the primary subsystems and interfaces in the existing design.
For example, a computer manufacturer
faster processor offered
by a chip
may
supplier.
replace a particular processor subsystem with a
Both the
initial
platform and
its
extensions
may
be referred to as successive platform versions.
A platform renewal occurs
o
when
the product design
new subsystems and new subsystem
specific
interfaces.
The
platform,
dimensions of value and cost competition
rearchitected to incorporate major
The new
subsystems from older product platforms.
as well as subsystem interfaces.
is
if
architecture
may
will also incorporate
It
carry forward
new subsystems
architected well, should provide
new
to the firm.
Insert Figure 2 here
Specific derivative products can be based any these versions of platform architectures.
refer to derivatives as {he follow-on products of a platform.
events
may
We
will
Appendix 2 shows how these key
be formed as an identifying records for subsequent data gathering.
3.2 Defining Metrics for the Product Family
We propose the following measurements of R&D performance
follow-on products within a product family:
(a)
focused on platforms and their
platform efficiency, being the degree to which a
platform allows economical generation of derivative products; and (b) platform effectiveness,
being the degree to which the products ba.sed on a product platform produce revenue for the firm
relative to the cost of
developing those products.
R&D productivity is the degree to which an organization can translate relevant technological
made by
discoveries,
firm.
itself
or others, into functioning products that reap financial benefits for the
As argued above, platform
efficiency indicates the degree to
which a product platform serves
as an economical base for generating follow-on products. Platform effectiveness
the
commercial productivity of a
used
to create products with the
Freeland. 1975; Foster et
of firms into
al.,
series of related products,
new
a measure of
comparing the engineering resources
commercial outcomes derived from those products (Baker and
1985;
R&D are represented by
with the development of
is
Brown and Svenson.
1988; and Cordero. 1990).
the engineering budget
products.
We
The
inputs
amounts and project times associated
believe that product sales
is
a better measure of
P(ii;c
6
commercial outcomes than
profits
due
to the lack of consistency with
which firms compute and
maintain profits for specific products (Patterson, 1983). Profits from product sales are not
consistently gathered at the product level by firms, and the definition of "profit"
business units. In comparison, sales figures are absolute.
may
vary
We have also found them to be
among
more
accessible than profit figures in our field research.
These measures, when combined with
management understand
visual interpretation of product family
the timing of platform renewal
and the frequency of follow-on product
generation from existing platforms. Further, these measures can be applied
product family
maps, can also help
at three levels
of the
(1) individual products within a platform version of product family, (2) at an
:
aggregated level for the product family as a whole for successive platform versions, and (3)
comparatively across different platform versions for different product families.
measures as meta-metrics because
refer to these
their
purpose
is
to help
We have come to
understand the dynamics of
innovation across sets of products rather than individual ones.
3.2.
1
Variables and Indices
This section presents the symbolic notation for the mathematical description of the metametrics. These notations are the building blocks for creating the measures of
performance
to
R&D
be presented thereafter.
Variables:
S =
Sales attributable to a platform or derivative product within a product family.
C=
Costs attributable to a platform or derivative product within a product family.
D-
The
deflator factor relative to a base year.
Used
to correct for inflationary effects for very
long product cycles.
Indices:
Np =
Number of platforms
in
a product family excluding the
initial
platform version or
architecture.
Ny =
Number
of platform extensions, or generations created off the base platform
architecture, excluding the initial platform version.
Nf =
Number
initial
Nj[) =
The
of follow-on products in a platform, or platform extension, excluding the
platform version.
total
number of follow-on products
in a
family excluding the
initial
piatfonn
version.
T[) =
The
last
time period number prior to entering the commercial cycle.
T^ =
The
last
time period for which sales were recorded
in the
commercial cycle.
I'diiv
7
p-
platform index;
v=
platform version index; v = base,
p =
base,
/=
follow-on product index;
/=
time period index;
For example, the sales
would be given
(1)
in
period
r,
=
t
1,
/=
2, 3,
base,
1, 2, 3,
...,
2, 3,
1,
1,
2, 3,
...,
Ny
...,
Nf
Tj) or T^;
of the second follow-on product, for the
initial
platform version
by:
ijl.'xur.;.!
3.2.2 Aggregating the Sales and Costs of a Product
From
Np
....,
Over Time
these basic variables, sales and engineering costs attributable to products can be aggregated
as part of performance measurement. Ideally, sales and costs are recorded at regular time intervals.
t,
throughout the entire product cycle.
If so, the
corresponding aggregated sales and costs are
given by:
T,
(2)
5..-.,
=5^(5.,./),
and,
Ti>
G..,=X(<^"")'
(3)
where
sales
accumulate from introduction to the end of product
from research
initiation to the
Over long product
deflator
is
cycles,
computed
life,
and research costs accumulate
time of commercial entry of a product.
it is
desirable to correct for inflationary effects in the
for each time period relative to the desired base year.
The
economy.
A
deflators are
applied to the formulas above to yield the adjusted aggregate sales and costs. Adjusted metrics are
denoted with a superscript 'primed' symbol
n
(4)
.
S'r..,^'^{S,..,).D.
1=1
(5)
C',..r=X(C,
•')£>'
These formulas represent the aggregated
the specific platform version within the
sales
and costs for individual /o//on'-o/j products within
same product
family.
To accumulate
costs at the platform
version level, the summations are extended. For example, the accumulated inflation adjusted sales
for an entire platform, p. including
all its
platform extensions, would be given by:
Pa^c S
N.
(6)
5'.
=
Ni
£
£s'.....
Similarly, the accumulated inflation adjusted costs for an entire platform
the cost variable, C.
With these
definitions,
it
is
now
would be computed using
possible to define the proposed meta-
metrics.
3.2.3
Platform Efficiency
At the follow-on product
level,
R&D leverage shall be defined as the costs incurred in
average
developing a follow-on product, divided by the costs incurred
platform upon which the products are derived.
Costs for Follow-on Product
R&D
Costs for Platform Version
Mathematically, the notation for efficiency shall be
developing the version of the
generally defined as:
R&D
=
(7) Platform Efficiency
It is
in
E The
.
indices defined above shall apply
accordingly. At the individual follow-on product level, the mathematical formulation for E, the
efficiency of developing an individual follow-on product relative to
E„...
(8)
=
that this
fraction of what
r"
\^
(9)
i.
is
given by:
Ihue
measure seeks
was allocated to
The average platform
base platform,
^'^''
p.
The question
its
to
answer
is:
"How much did
the product cost to develop as a
the base platform architecture?"
efficiency
is
defined
as:
E,.^—^,
R&D costs associated with
developing all the follow-on products of a platform version divided by the R&D costs of
The average platform
efficiencies
developing the platform
itself.
is
therefore the average of the
By grouping
versions, the question being addressed
is:
follow-on products within their base platform
"How much
did
it
cost on average to develop follow -on
products relative to what was spent on developing the base plarfonn versions within the product
family?"
Piii-c
y
One can
also
compare
product families
the corporation
.
the platform efficiency of different platform versions across different
The question being asked
compare with one another in
development of several or more
the respective engineering
groups within
which provide
their abilities to build robust platforms
the basis for efficient product development?"
the
"How do
is:
This might be useful for managers responsible for
An
distinct product lines.
aggregate
R&D cost leverage value
can be computed for an entire product family by summing across follow-on products, platform
versions, and
(10)
between successive new platform
E =
N/
Nv
Nf
f=hitie
\=hiue
f =\
'X
-";:
S
p = Itase
What
is
architectures:
lev.
V = huie
We
a reasonable platform efficiency value?
strongly suspect that the answer
is
industry
specific for particular types of products within assembled, nonassembled, information system
pure information product categories.
Our
application of this metric to the product families of the
firm studied (described in the next major section of the paper)
platforms yielded values for
E as
low as
produce streams of follow-on products
and
.
at
10,
meaning
10%
showed
that strong product
that efficient platforms allow the firm to
per product of the cost of developing their
respective base platform architectures.
If
resources are being effectively used, and learning
E should therefore
efficiency should improve (that
of a product family.
been added
in the
E values of
or close to
new product and
corrected. Conversely, the closer
follow-on product generation.
If
a product platform, the platform
posit that platform
may
indicate that
little
problems
or no
the firm
much
spends as
poorly designed.
its
in earlier
to
consistent with the S-curve
generation performance
is
indicate
done
weakness
phenomenon where,
in
in
as
it
does on
conjunction with assessments of
in the
An
later in the paper.
increase in
after a certain point,
management
E over
underlying product architecture. This
little
additional product
achieved from incremental investment. (Foster 1986)
platform efficiency can also signify a change
proved for
suppliers could produce very low values of
Platfonn efficiency can be used as an indicator of platform demise.
is
products had been
make follow-on products
product change and effectiveness. This combined analysis are provided
may
functionality has
should also be noted that technological
It
component
Interpretation of the metric should therefore be
successive follow-on products
new
to 1.0, the less efficient the base platform design
breakthroughs achieved by the firm or
E.
we
decrease) with each successive platform version
that perhaps only that
E is
is
it
taking place,
is
or key resources,
A
decline in
human
or
l\i^c 10
otherwise. For example, the advantage gained by reusing a proven pool of resources can be lost
through organizational change or by shifting the charter for developing a particular product family
Key managers
to another facility in a different geographical location.
been
lost.
All these factors must be considered as
Market factors may greatly influence
the
the efficiency with
form of new products for market applications
case of a firm venturing into a
new
management
or engineers
interprets the
also have
E metric.
which existing platforms can be applied
that the firm has not
in
addressed before. Take the
business area by trying to create products that leverage existing
The requirements of the new market may be so
technical competencies.
may
different that the firm
encounters great difficulty satisfying the customers' needs, even with access to robust product
platforms from
its
core business. The efficiency metric
may
in this
new product
success of management's strategic decision to create a
Platform efficiency
may
case provide insight into the
family.
also be considered from the standpoint of development time cycles.
One
can posit that a product platform, while taking relatively longer to create than follow-on products,
is
robust
if
it
the start and
allows the firm to experience rapid generation of those follow-on products. Using
end of
R&D for product platforms and follow-on products, an elapsed time cycle
measure of platform efficiency can be expressed
rCwUTime
(11) £"'""""' r...ha.e
=
'p.
I p.v.
and replicated
i:
as:
f
ha'it
to the aggregated levels of all products within a platform version,
and across
platforms versions within distinct product families.
It
stands to reason that
on product
efforts,
if
indeed platform development efforts take longer to complete than follow-
R&D aimed at platform renewal should be pursued concurrent within product
developments on existing platforms. This
will insure a continuous stream of products that
embody
competitive technology.
3.2.4
Platform Effectiveness
Computing
R&D returns as accumulated profits divided by development costs on a product by
product basis was proposed by Foster, et al.(1985) to assess the productivity for products over the
full
R&D and commercial
reliability
and
life
accessibility.
comparing the product
cycle.
As
stated above,
we used
product sales instead for reasons of
We propose that the effectiveness of a platform
sales to product
development
may be
assessed by
costs, either at the individual product level, or
P(if;e II
for groups of products within distinct platform versions. This
measuie of effectiveness
may be
considered one of understanding the commercial leverage provided by technologies applied by the
firm in
in the
its
products. Platform effectiveness
is
synonymous with platform leverage,
referred to as
L
equations below.
The platform leverage measure,
What have been
L, seeks to address the following questions:
the returns realized
on the firm's
products to the costs of developing them
R&D investments comparing sales from
?
How has a particular product family trended over time through
versions?
Have
its
successive platform
returns on investments improved or declined?
Are certain development groups more
effective in creating leverage from their
respective platform development efforts than other groups within the firm?
Platform effectiveness
L...
(12)
at the
follow-on product level
is
given by:
^
=
and, aggregated at the platform version level by:
Lr.=^
(13)
which
the
is
the
sum
of the sales of products within a platform version divided by the
sum
of
R&D costs for creating those products.
Platform leverage
and then used
(14)
to
may
be further aggregated across
compare
L,.
the platform effectiveness
all
platform versions within a product family,
between
= ^^^^7?
I S
C'„.
.
The L metric provides information on product performance,
the entire product
distinct product families as follows:
life
cycle inclusive of
individually and in aggregate, across
R&D and commercial activities.
As
the costs of follow-on
Fiii^c
12
product developments decline relative to that of their underlying platforms, the values for platform
effectiveness for the product stream should increase.
There are many factors that can lead
the platform architecture itself
the denominator,
above,
ie.
may have
market
that are competitive in the
to declining
platform leverage within a product family.
outlived
in features
and
its utility
as a basis for creating specific products
This would lead to declining sales. As for
cost.
R&D costs may be rising due to problems in platform efficiency as defined
the economical generation of follow-on products. If this
posit an inverse relationship
platform efficiency.
is
the case, then one could
between the two respective measures of platform effectiveness and
A decline in platform efficiency may well
The causes
platform architecture.
for that decline
indicate the need to create a
may be many. For example, changes
organization
may impede
the ability of the technical staff to perform well.
technologies
may emerge
that competitors use in their products
understand and integrate into
its
own
Or new
new
in the
external
and which the firm
fails to
products.
Applying and Interpreting Meta-Metrics
4.
4.
First,
1
The Sample
The data used
company's
to test the metrics
sales in 1993
were gathered from a measurement systems manufacturer. The
exceeded $1
measurement technologies
billion.
The
firm's core capability has
to industrial applications.
Management
been the application of
considers these products as
"front-end" devices that gather data in real-time and assess them, identifying problems and sending
these data to "back-end" information systems.
Studying the historical record of these product families required many discussions with
engineering and marketing managers working
family.
The
first
goal
was
in the
business unit responsible for each product
to construct product family
prior section to distinguish
between
(a) the initial
to existing platforms, (c) the creation of
wholly
new
new
maps, using the typology described
in the
platform development efforts, (b) extensions
platforms or product architectures to replace
those in existence, and (d) the specific products associated with each platform generation.
Drawings of product architectures
at the
subsystem and interface
platform enhancements and the development of
firm's internal product
new
naming conventions proved
level
were used to help identify
platforms within each product family. The
particularly u.seful for initial grouping of
l\n;c
/.<
product streams within successive platform generations.' The product family maps were presented
to the
engineenng teams working on the respective product
consensus on
their validity.
In total, the data set
used
lines for
in this
feedback and to create
paper contains information for five
product families, which collectively, provided a rich database of platforms, platform extensions,
and follow-on products
to test
our methods.
Figure 3 shows the product family
and control stations
that gather
map
for
Family
line,
information display
and manipulate data from front-end data collection devices
purposes of event handling and data trending. The
denoted by a thick
A comprised of centralized
representing the start of
life
cycles of Family A's
for the
two platforms
are
R&D for each architecture until the end of active
marketing of products based on the platforms. These two platforms contained two very different
product architectures, as described by engineers
at the
subsystem and interface
The
level.
first
platform was based on analog signal processing, and the second, on digital signal processing.
Products based on the two platforms also "looked" very different to the outside observer. The
first
platform had two versions; the second platform, only one. The specific follow-on products based
on each platform version
date of the earliest
lines
line is
into a particular product.
number
The
starting
R&D project associated with each product is represented as the thinnest of the
on the map. This
projects
are described in the text enclosed in the right-angled lines.
in the
an aggregate representation of numerous individual projects folded
For several hundred products
in the total
R&D
sample, individual
thousands. Allocation of these individual projects to specific platform and
follow-on product development was performed by engineers
who had worked on
the development
efforts.
Insert Figure 3 here
While
R&D managers within the firm identified product families and their components, we applied
the technology and market applications
newness method of Meyer and Roberts (1986)
the relatedness of the product families.
Appendix 3 shows
the application of the
A. The reader can note that relatedness of the products within
in
method
to
Family
terms of embodied
that family in
core technologies and market applications. The assessments of newness
to validate
technology generally
correlate with the engineering costs associated with each successive product
shown
for
Family A.
The determination of platform and follow-on product spending within each product family was
based on the following assumption: the costs associated with individual projects who.se
'
Tomatsky and Klein (1982) have argued
that
knowledgeable observers
ot a particular industry will
results
make
consistent
interpretation of those individual products that belong to specific product families.
Pa^c 14
were embodied
deemed
first
to
in the first
product spinning off from a particular version of a given platform were
be platform development efforts for that version. In other words, the spending for the
product of a platform version and the spending for that platform version were treated as one
the same.
Note
R&D efforts longer term in nature.
had
that often the firm
Such
pursued concurrently with work on a current platform version but not completed
new
rolled into a subsequent platform version or an entirely
concurrent platform development that
we deem
efforts
in
would be
until later
and
platform. This illustrates the idea of
essential to product renewal
and sustain success
within the product family.
Maps
similar to Figure 3 were created for
all
the product families incorporated in the study. For
each product within a family we sought to construct from corporate record the following data:
the start of
R&D (for the projects associated with the product), (b) the end of R&D, (c) the start of
marketing, (d) the end of active marketing of a product,
manufacturing engineering costs,
dollar equivalents. All financial
the
(a)
Consumer
(g) the
(e) the
development
costs,
(0 the
product introduction costs, and (h) the sales
amounts were recomputed
to their value in
in
1993 dollars.
U.S.
We used
Price Index to inflate dollars from prior years to their 1993 dollar equivalents.
These yearly adjustment
factors are provided in
Appendix
5.
Because
historical data for
engineering costs and sales revenue were not maintained on a yearly basis for individual products,
we
first
aggregated these amounts to a product and then applied the CPI index to these
the index value for the last year of the
would want
to maintain these data
R&D cycle and commercial cycle respectively.
on a yearly basis so
that the
totals,
using
Ideally,
one
CPI index might be applied more
precisely to the numbers.
The combination of the product family maps and
for
their associated data
comprised the complete basis
computing the measures of platform efficiency and effectiveness (leverage), and
timing factors
in
development. While these data might seem reasonable for any firm to maintain,
On
they proved difficult to gather.
cost allocations for individual
the shelves of
for observing
one hand, the company maintained very detailed information on
R&D projects, either in paper form as project data sheets sitting on
R&D managers, or for more recent years,
man-months of direct
in
computer
labor, an internal "cost" for that labor,
files.
These data provided
and materials charges.
We were
unable to gather manufacturing engineering costs and market introduction costs on a consistent
basis,
and therefore excluded these from the development cost
sample. For several product families
in the
multiplied these amounts by a cost per
man
for the product families in the
sample, only man-year allocations were available.
year amounts for the given years
in
development
were supplied by management.
Pafie 15
We
that
Project data had to be aggregated during interviews where individual projects were assigned to the
products by
maintained
R&D managers who had been associated with the projects.
in the
accounting department in various forms, and for the majority of product families,
were already allocated
a product
was
to individual products
actively marketed.
It is
and merely had
to
be totaled for those years
in the
practice in the firm (particularly for the development of wholly
management between
In the next three sections of the paper
which
firm to provide
information. This reflects the fact that while concurrent engineering
versions), concurrent
in
of further importance to note that the integration of
engineering data with sales data had not been systematically performed
management
Sales data were
we
new
platforms or
functional areas has proven
will
now
more
is
a regular
new
platform
difficult to achieve.
apply and interpret these metrics
at three levels
of analysis: within a single generation of a product family, comparatively between generations of a
product family, and comparatively between different product families.
4.2 Metrics Applied at the Follow-on Product Level
Two
the
product families. Family
A and Family
B, demonstrate the application and interpretation of
measures of platform efficiency and effectiveness
Family
A
,
described above,
is
at the specific
follow-on product
comprised of central information systems products.
successive platform architectures, with the
first
It
level.
had two
platform having two basic versions (the second
being a major enhancement of the technology embodied
were based on these platforms. The data gathered
in the first).
for these products
A total of twelve products
and the computed measures of
platform efficiency are provided in Appendix 4 (which contains the computed metrics for
product families used
in the paper).
The
first
all
products released from a product platform are
considered the denominator of the efficiency equation. All data were normalized to their 1993
dollar equivalents.
Insert Figure
Figure 4 shows both platform efficiency metric and
4 here
its
running average for the products within
each respective platform version. The running average of platform efficiency smoothes wide
fluctuations that
trend.
may
be observed for individual products, helping to show more clearly the overall
Product CIS lA-Fl (Central Information System, Platform
on Product
for 18 cents
was developed on approximately
1)
on the platform
management
initiated a
new component
dollar.
While
this
15 cents
1,
Platform Version A, Follow-
on the platform
dollar,
and CIS
1
A-F2.
follow-on product was being developed,
concurrent effort to substantially improve the underlying architecture with
technologies and software development tools.
It
understood that
its
cunent
hii;c 16
architecture could not support the efficient development of the stream of products planned over the
next five years.
While the
first
The
results
of this decision, and the
came
follow-on product of CIS IB
subsequent products were developed rapidly
IB platform
Management
dollar.
at
initiated yet
in
new platform
a cost of 16 cents on the platform dollar,
very incremental cost of under 5 cents on the CIS
another platform enhancement effort even while
products were being created from the CIS IB architecture. In
and the
availability of
Upon
architecture.
the
new hardware and
the related platform
this case, the
needs of customers
software components mandated an entirely
new
completion, derivative products were created for approximately 10 cents on
its
CIS 2A platform
extension, were dramatic.
dollar.
Across our sample, platform efficiency values of 10 cents or below on
development dollar represent very good values
terms of leveraging platform
in
architectures in specific follow-on products. This level of efficiency can only be maintained over
time with a policy of continuous platform renewal. Otherwise, as an architecture ages,
it
presents
increasingly greater obstacles to engineers to achieve customer requirements.
We turn to another product family
metrics
may used
to indicate the
measurement "boxes"
to illustrate the platform effectiveness
need for platform renewal.
Family
B
is
measure, and
shown
in
Figure
commercial cycle
5.
that gather industrial process information in real-time, identifying
The period from
for products derived
the start of
from
it
both
comprised of
events and sending data to information systems such as those of Family A. The
is
how
map
emergency
for Family
B
R&D of the product platform to the end of the
exceeds ten years. The base architecture contained a
microprocessor, communications, and display electronics, a data acquisition and filtering
subsystem, and a data management subsystem. This product platform took 3.5 years to develop,
and from
it,
developed
five follow-on products
in
approximately
30%
were created. The
first
three follow-on products were
of the elapsed time allocated to platform development, and the
two, spiking up to almost half that time. In
total, the
platform start of
last
R&D to end of commercial
cycle spanned more than a ten year period. Figure 5 also contains the engineering cost and sales
figures for Family B's products.
that they reflect
end of
Time
wise, these data existed only in aggregate form,
meaning
R&D and commercial cycle totals respectively.
Insert Figure 5 here
Using the equations defined above, values
for
for platform efficiency
Family B. These data and the computed metrics
are plotted in Figure 6.
To
facilitate
measures, the efficiency scale
efficiency,
is
is
for
and effectiveness were computed
Family B are provided
in
Appendix 4 and
comparison between the efficiency and effectiveness
plotted in reverse value order
plotted at the top of the axis and
(ie.
a value of
weaker values of efficiency
.05,
at the
showing strong
bottom).
him- 17
The platform
efficiency measure
shows high
levels of efficiency for the first three follow-on
products based on the platform, where each was developed for on average
platform dollar. Then, for the fourth follow-on product, the metric spikes
The
cents on the platform dollar.
fifth
1 1
cents on the original
down
to a value of
42
product experienced a similar lack of platform efficiency,
costing about 36 cents on the platform dollar.
Insert Figure
The platform
commercial leverage of products
indicates the
to
product development costs.
basis (rather than
efficiency).
B
effectiveness measures for Family
By
definition,
6 here
cire
also highly revealing. Platform effectiveness
(or sets of products) as a ratio of total product sales
we compute
comparing any given product
to
its
this
measure on a product by product
base platform as in assessing platform
For Family B, commercial leverage grew strongly from the
Initial) to that
of the third follow-on product
(MS 1-3).
In particular,
commercial success, returning $259 for every dollar spent on
follow-on products
(MS 1-4 and MS
its
initial
MS 1-3
product
was an outstanding
development. The fourth and
-5) experienced a decline in sales relative to
I
(MS 1-
fifth
development
costs.
Taken
together, these measures of platform efficiency
management
words of one
and effectiveness were interpreted by
to indicate that underlying product platform for
R&D manager,
follow-on product in
its
Family
B needed
renewal. In the
the product platform began "running out of gas" after the third
ability to a) facilitate rapid
and cost effective development of derivative
products, and b) deliver the features required by customers at a price point that
would lead
continued strong sales. This deduction would not have been obvious by looking
budgets or product sales, nor by simply looking
rule out several factors that
at the
might have otherwise caused changes
time the market was relatively stable. There were no significant
market growth was
fairly
associated with Family
B
constant
at
about
5%
map
product family
per year.
The
in costs
annual
at
(Figure
and
new market
5).
sales.
to
R&D
We can
During
also
this
entrants and the rate of
.sales force, pricing,
and practices
also remained fairly stable.
Organizational issues strongly affected the evolution of Family B. Earlier generations of this
particular
measurement device prior
laboratory. Senior
Family B had been developed
management moved
in
the charter for developing the
an American
new product
R&D
architecture
R&D laboratory, a group that had demon.strated
in creating low-cost, modular products in related product areas. The R&D charter
represented by Family
competence
to
B
to a
European
/'r/c'c
IS
for
Family
B was
(MS 1-4)
groups.
with
its
then shifted back to the American group starting with the fourth follow-on product
as part of a larger initiative to rationalize
R&D foci among the firm's various R&D
The American group defined follow-on products and jointly implemented
European
partner.
these definitions
As expected, some degree of organizational turbulence ensued and
The
helped lead to higher costs and longer development times.
charter wars experienced within
embark on a platform renewal
the product family contributed to the firm's failure to
effort in a
timely manner and has placed a stellar product line under substantial market risk in the future.
Today, management has embarked on a crash
effort
would have been
better started earlier during the
of efficiency and effectiveness during product
declining efficiency of
development plan
is
effort to create a
MS 1-4.
new product
architecture. This
peak of the platform's success
in
its
terms
MS 1-3 shown in Figure 6 and certainly after the
Seeking to avoid
this
problem, the group's
new platform
being extended to include initiation of the "next generation" several years
hence. Additionally, senior
R&D to place American and European
management has reorganized
laboratories within a single business unit to ameliorate the destructive charter wars of the past.
4.3
Meta Metrics Applied
The measures of platform
at the
Platform Version Level
efficiency and commercial effectiveness
aggregate level for the product family,
product family.
To
illustrate this,
we
ie.
in
turn to
may
also be
computed
at
an
aggregate for successive platform versions within a
Family C, comprised of highly complex imaging
products used for measurement applications. Current products
in the
30,000 different components (making platform version tracking
necessary for this type of analysis). The family's
map
is
at
shown
in
family contain more than
subsystem and interface
Figure
7.
It
level
spans more than
15 years.
Family
C has
had two underlying product architectures, IM
1
and IM2. IMl had three successive
platform versions, IMl A, IMIB, and IMIC. The two platform extensions of IMl introduced
imaging modalities
that
new
provided users with substantially improved diagnostic capabilities. The
second product platform,
IM2
(replacing
IMl) featured an
entirely
new
architecture that
was
a
combination of more powerful image gathering and processing subsystems and new software
subsystems for the user interface and data management.
"horsepower"
IM2D.
to users.
IM2B was
new imaging
IM2
itself
IM2D
effectively offered twice the
had four platform versions, noted as IM2A, IM2B, IM2C, and
an extension of the platform to a
modality.
It
both incorporated a
new market
application.
new modality and applied
IM2C
it
to a
incorporated a
new market
application.
Pane /y
Insert Figure 7 here
These various platform
efforts
perspective, however, the firm
six
were each three to five years
was introducing new products every
months. The ability to generate
platform renewal, which
itself
this
measure plotted
the pursuit of
on a platform version basis
in reverse
recently, every
stream of products was based on a strategy of concurrent
was based on
even longer term research into new
for
C
Family
These measures plotted
platform efficiency and effectiveness.
the customer's
and more
year,
measurement techniques and associated technologies. Appendix 4 contains
sales data aggregated
From
in the malcing.
the engineering
and
and the resulting metrics for
in
Figure 8, with the efficiency
value order.
Insert Figure 8 here
The metrics help
reveal
two very
different types experiences for the
family. Family C's product family
initial
platform,
in a concurrent
of that platform's two successive versions,
were generated
platform version dollar. The quality of the product
efficiency in introducing
new
products,
came
to a propitious
product platform (IMIC). Just as the firm was putting
first
product
IMIB and
manner. Strong levels of platform efficiency resulted. For
three versions of the first platform, follow-on products
their respective
in this
(Figure 7) shows the careful, patient development of the
IMIA. The development
IM IC, was approached
all
map
two platforms
line,
at less
and the accumulating cost
peak during the
its
than 10 cents on
third version of the
best foot forward in terms of
engineering, two major competitors, themselves large, diversified businesses, exited this market
due
to highly publicized scandals that elicited
government intervention. The firm exploited the
opportunity. Family C's products during this time yielded over 80 dollars in sales for each dollar
of engineering cost to develop them (as seen under
The second platform of Family
experienced a different
Management decided
fate.
to
C
IMIC)
in
Figure
(noted in Figures 7 and 8 as
The team sought
to
8.
IM2A, IM2B, IM2C, and IM2D)
double the imaging power of the system.
produce the image processing subsystem boards with a newly installed
process technology (surface mount manufacturing). Given the major change in both product
design and manufacturing process, the platform renewal was started too
earlier,
that
one of the company's lead engineers had
was
directly competitive.
management seemed
new
to
He had
left to start his
initiated a platform
be resting on the laurels of the
startup's efforts, a crash effort initiated to create a
own
late.
Further, several years
firm and created a product line
renewal "on the outside." Meanwhile,
first
new
platform.
When
it
received word of the
platform. Then, the
new competitor
Pai;c
20
announced
IM2A
its
new products
were brought
to
to industry acclaim.
Management had
to respond: products
based on
market quickly thereafter even though the new technology had not yet
sufficiently ripened.
The
initial
this
time offered
product under
little
IM2A experienced problems
new
in the field.
functionality to customers (being
The team was consumed with
new
generated from
the metric for
introduced a
IM
IM2B were
IM2B
platform, a
The
new measurement
introduced
The
a
produce follow-on products during
more acceptable
efficiency, however, that
of Family C's
The
first
(IM2D) extended
was
stabilized
by
cost of 30 cents on the
targeted.
dollar.
Then,
that
in
IM2C,
the product line to a
this point,
IM2D
computed
the
team
IM2C
Figure
in
new measurement
and the derivative products were
platform dollar.
It
was a
level of
to declining leverage in
IM2C, and
fifteen year period, overall
efficiency for these
further decline in
in the creations
two platforms could be
within the platform led
15%
this
per year.
IMl and IM2, our measures of
compared.
In this case,
development of IM2 was nearly twice as costly as IMl (which, as shown
part to the late start
LM2B
should be noted that during
excess of
in
of platforms
clearly
application targeted by
It
IM2
platform architectures. Sales
new products
IM2D.
market growth remained strong
equal costs had been incurred
for the successive versions
new market
reversed the downslide. However, the difficulty of creating
however, the
in
Figure
8,
was due
in
and hasty development of IM2). But even without correction the explosion of
costs for derivative products from
IM2
is
obvious.
An advantage
of product family maps
is
that
both raw data and efficiency measures between successive generations within a
product family can be clearly seen and compared.
Our
the
The products
time (referring to
move smoothly between
of the product line plummeted during IM2A. The
in
IM2B,
platform.
effectiveness measure of commercial leverage
such differences
In
8).
three times poorer than that experienced throughout the products
still
platform shows the results of this failure to
If
Figure
efficiency metric deteriorated to almost 2.5 dollars relative
The platform technology had
at
(in
technique. Significant technical obstacles in the underlying
fourth version of the platform
application.
was
application
30 cents on the platform
architecture had to be surmounted.
8).
new market
IM2A
very successful commercially. However, the efficiency decreased with
rising to over
to the platform dollar to
output peripherals, for example).
fixing problems in the underlying architecture. This yielded the
extremely high platform efficiency measure for products under
next version of the second
The follow-on products during
interpretation of these metrics
and the events
that
caused them was reinforced when
Pcijic
21
management provided
staffing information. Staffing data
had worked on underlying platform projects versus those
over time. Note the sharp
activity
thie
number of engineers who
for follow-on products
on a year by year
9 shows these allocations as percentages, together with the increase
basis. Figure
replace the
comprised
shift in this allocation
when management sensed
in total staffing
the urgent
need
to
product platform, IMl, with IM2. The percentage allocation to platform-related
first
swung from 10% of the engineering
staff to
70%
in
Once IM2
a single year!
stabilized,
resources were shifted largely to follow-on product development.
9 here
Insert Figure
Management
learned from this experience.
development
effort
(which
we may
call
The team
IMS)
initiated
to replace
EM2
an entirely
new platform
with better imaging and user interface
We visited the company during the week when IM3 had passed all the internal quality
control tests. To our delight, the R&D manager was convening a meeting of his best engineers.
capabilities.
Their purpose: to begin the design of the next product platform (IM4). The
allocations
between platform work and
that for follow-on products has
shift in
become
engineering
less dramatic,
indicating a policy of continuous platform renewal.
4.4 Comparisons between Product Families
The platform metrics can
performance for a
we
set of
management
also be used by senior
product families and
facilitate learning
two
will use the platform efficiency experiences of
Family
D and E
.
to comparatively assess
from
that endeavor.
R&D
To show
different product families, referred to as
Both product families are comprised of measurement devices, each targeted a
different but related market application. Family
versions of that platform and
1
1
D had a single product platform, with three
follow-on products. Family
with several or more platform versions, and also
1
1
E had two
follow-on products.
product platforms, each
We gathered data,
maps, and computed platform efficiencies as described above (see Appendix 4 for specific
The two product
groups.
10.
The
families were developed during equivalent time periods by
different
is
shown
two versions of the Family D's platform were highly productive
in
terms of
in
economical generation of follow-on products. The
this regard.
two
terms of platform efficiency
The performance of these two groups
first
this,
Discussions with managers involved
suspicion that the platform had reached
capabilities required by customers.
its
A new
in
limit in
third platform version,
development
at that
data).
R&D
in
Figure
however, suffered
in
time confirmed our
terms of efficiently delivering the
architecture for Family
made
new
D (as opposed to the
platform
Pai^c 22
Dl C
extension labeled
the renewal of
in
Family
in
Figure 10) would have best been started
D was initiated in the late
in the early 1980's.
new
1980's, resulting in a
Instead,
platform (labeled
D2 A
Figure 10) from which products were released starting in 1990. While the efficiency of this
platform for creating derivative products (.27) was better than that of
(.42),
that value
was
below the strong platforms
still
in
its
immediate predecessor
terms of technical leverage within the
company.
Insert Figure 10 here
In contrast, the platform efficiency values for Family
The values
for the successive platforms versions are
range, including the today's present platform (E2
B
E
indicate consistently strong performance.
all in
in
the 10 cents
Figure
10).
on the platform dollar
Note further the overlap
between the development of the second platform (E2 A) and the creation of follow-on products
from
its
precursor (El C). Family
we deem
organization must
memory
facilitates internal
remember
rhythm
in
platform enhancement and renewal that
benchmarking of R&D
past experience. In the context of
activities.
In order to leam, an
new product development,
skills,
Measurement of these
and customer perceptions and participation
factors
and analysis of
their results in areas
in the creation
identified product families with high levels of platform efficiency
effectiveness relative to other product families,
of
such as platform
efficiency and effectiveness provides information for experimentation and learning.
management has
this
includes the evolution of product families and their architectures, the
composition of teams and
products.
indicates a
highly desirable.
This comparative analysis
corporate
E
Once
and
management may then explore underlying
approaches and techniques for product planning, design, and integration.
5.
Discussion
In our field work,
\V/it';i
we encountered
a
number of frequently asked
should we renew our platforms?
to take?
What
complete?
types of engineering
How can we
How much
questions:
will platform efforts cost,
and commercial
benefits
and how long can we
can we expect to gain from these
expect them
efforts
once
improve our approaches and strategies for product development?
Page
2.*
The work presented
new product
to
in this
paper does not exphcitly
tell
management
when
precisely
to create a
platform. However, our concepts and metrics provide managers with a rich context
determine when product platforms should be obsoleted with new platform designs and what to
expect from these
new
new
designs once
products are developed from them. In short, the work
R&D dimension of product
comprises a framework for learning, specifically targeting the
development. (Maidique and Zirger, 1985) The axiom "hind-sight
is
20-20"
is
well remembered.
Without descriptions of factors influencing events, such as those provided for the product families
in this paper,
The
R&D performance may be prone to misinterpretation.
observed measures of
cost-effectiveness of implementing these measures within a
start is to
ask management
if it
company
will vary.
The place
to
recorded the evolution of the firm's product families and can
provide information on platform and product engineering costs and as well as product sales.
Researchers
may be
surprised at
how few
expect that most firms have not taken the
firms have a grasp on these fundamental data. While
initiative to
product development costs and sales are maintained
R&D measurement system will not be excessive in
The
combine engineering and
at least in
company's
traditional skill
had been
to
make
referred to earlier as
As we were concluding our
study
discontinuity.
in this
The
"high end", technically pioneering products for
customers that had the money to buy new features
market forces had recently reversed
we
was confronted with a major market
it
if
light of the value returned.
concurrent management within product developing firms.
the end of 1993,
sales information,
paper form, the cost of creating the
collection and synthesis of these data are essential to support what
company towards
we
this condition:
at
ever higher prices. Certain regulatory and
customers
now
desired low-end, value rich
products. At the time of this writing, the culture, people, and processes within the firm had yet to
facilitate
a change
in
distinct "islands" of
engineering approaches and philosophy. Marketing and engineering remain
work
that
had
to
become
one.
Achieving high levels of platform effectiveness, or leverage,
growing strongly.
If
is
easier to achieve
the market
the target market proceeds to flatten, a successive platform version that
be better designed and enable greater functionality in follow-on products,
leverage.
if
While we have included market growth
in
we were
is
may
exhibit a decline in
our interpretation of the metrics
discussion above, a more formal incorporation of this information
leverage. Unfortunately,
may
is
in the
desirable for the measure of
unable to gather reasonably complete market growth rates
at
the
product family level within the sample firm. Our discussions with management indicated,
however,
that
market growth
rates, suppliers, the structure
and
.scope of distribution channels,
even the individuals key positions of management (business, marketing, and
R&D)
were
Pcii;c
all
24
and
remarkably stable during the time periods studied. Other factors such as manufacturing capacity
and the sourcing of components showed continuous growth and no changes of a disruptive nature.
There
is little
doubt, however, that
management should seek
to
augment measures of an
nature with external information that might have significant bearing on the
internal
phenomena under
investigation.
Another measurement issue concerns a more
ideal data set gathered for product
holistic definition of
product development costs. The
development costs would include manufacturing engineering and
market introduction costs with those for product engineering. One might also wish
to track the
costs of installing products, training users, and repairing initial product defects as part of assessing
We were unable to gather these data. We believe that the
platform efficiency and effectiveness.
lack of integration between information systems in engineering, manufacturing, and marketing
which we encountered
is
not atypical in industry.
How
Product quality also needs to be continuously assessed.
target
customer needs? What competitive benchmarks
the firm
have
to
guide
its
in
well does a stream products meet
terms of overall value and features does
platform development efforts ? Platform efficiency
related to the design quality of platforms.
The platform
may be viewed
as
effectiveness measure should also reflect
the fact that higher quality platforms and derivative products fare better in the marketplace than
those of lesser quality. Nonetheless,
we would
measures product functionality versus
failures),
at the
not wish to use these measures in isolation from
price, service record (such as critical
and for many firms, the cost of integration of
customer's
products with those of other companies
site.
The development of an
products based on
it.
entirely
This
is
greater technical uncertainty
innovations
its
hardware or software
may have
new
platform
is
riskier than the
development of follow-on
because platforms involve a greater innovation effort
—
than derivative products.
The anomaly
is
new
and hence
that while incremental
less technical risk to the firm than architectural innovations,
such architectural innovations (often from
~
entrants) that have historical
made
it is
precisely
firms'
competencies and products obsolete over the long term (Utterback, 1993). The failure to develop
new platform
have found
architectures
on a continuous basis subjects the firm
that engineering
managers generally understand
this.
to substantial
market
risk.
We
Business managers often do not.
leading to inappropriate expectations regarding project time and cost
(ie.
new platforms
are
different than platform enhancements, and both are different than specific product developments).
In turn, these expectations
can result
in
misallocation of resources.
Ptif-c
25
We remember the words of one senior engineer:
How can we get management to understand the differences between platform efforts and those for specific
products? Our new product planning processes focus on single products and the emphasis
faster
and at
lesser cost.
We can
't
to facilitate
to create
them
achieve this without newer and better platfonns. Too often, our efforts to
renew aging platforms are pursued "undercover" and we
reward systems
is
start
them too
and encourage appropriate behavior with
late.
We need our planning processes and
respect to platform development
and
renewal.
The
last
statement
is
particularly intriguing.
foundation for reconsidering their
The approach presented here provides
new product planning
processes and reward systems.
observed the stage-gate new product development processes of many firms to be
product and single period
in application,
making
little if
firms with a
We
have
strictly single
any explicit differentiation between more
fundamental platform efforts and the development of derivative products.
embraces the product family and the renewal of architecture
is
A planning process that
a possible multi product, multi
period basis for further research and application.
Pa!^r
2f)
References
Baker, N. and
Management
J.
Freeland, "Recent Advances in
Science, (21:10), 1975, pp.
1
Balachandra, R., Early Warning Signals for
Batson, R.G., "Characteristics of
Management,
(34:3),
R&D Projects,
MA,
Lexington Books, Lexington,
1991
R&D Management which Influence Information Needs," IEEE Engineering
August 1988,
Beardsley, G. and E. Mansfield,
R&D Benefit Measurement and Project Selection Methods",
164-1 175
pp. 178-183.
"A Note on
the
Accuracy of Industrial Forecasts of the
Products and Processes," Journal of Business
(1),
Brown, M.G. and R.A. Svenson, "Measuring
R&D
Profitability
of New
1978, pp. 127-135.
Productivity," Research Technology
Management, July-August,
1975, pp. 11-15.
Cooper, R.G., "Stage-Gate Systems:
1990, pp. 44-54.
A New Tool
for
Managing New Products, "Business Horizons, May June
Cordero, R., "The Measurement of Innovation Performance in the Firm:
1990, pp. 185-192.
Deming, W.E., Out of Crisis, Massachusetts
Cambridge, MA, 1982.
Institute
An
Overview," Research Policy,
(19),
of Technology, Center of Advanced Engineering Study,
Dodge, M., "Rate Your Company's Research Productivity,
"
Harvard Business Review, November-December 1963,
pp. 109-122.
Edwards, S.A., and
M.W. McCarrey, "Measuring
Management, January-February 1973,
Ellis, L.,
"Viewing
R&D Projects
the Performance of Researchers," Research
Financially," Research Technology
Management, March-April 1984,
"Improving the Return on
R&D
-
I",
Foster R., L. Linden, R. Whitely, and A. Kantrow, "Improving the Return on
R&D
-
11",
Foster R., L. Linden, R. Whitely, and A. Kantrow,
Management, January-Febmary 1985,
Management, March- April 1985
Foster, R., Innovation:
,
Technology
pp. 34-41.
pp. 29-34.
Research Technology
pp. 12-17.
Research Technology
pp. 13-22.
The Attacker's Advantage, Summit Books,
New
York, N.Y., 1986.
Government R&D Projects: The Measurable Effects of Performing Required
Management and Engineering Techniques," IEEE Transactions on Engineering Management, (2) February 1975, pp.
Gerloff, E., "Performance Control in
6-13.
Gupta, A, and D. Wilemon "Accelerating the Development of Technology-based
Management Review, (32:2), 1990, pp. 23-33.
,
Hauser,
J.
New
Products", California
and D. Clausing, "The House of Quality, "Hanard Business Review, May-June 1990, pp. 63-73.
Henderson. R. and K. Clark, "Architectural Innovation: The Reconfiguration of Existing Product Technologies and
the Failure ot Established Firms." Administrative Science Quarterly, (35) 1990, pp. 9-30.
lansiti.
M., "Real World
R&D: Jumping
the Product Generation Gap,"
Han-ard Business Review, May-June 1990,
pp. 138-147.
unpublished survey of 248 research directors
Industrial
Research
Research
Institute lo (he authors.
Institute,
May
17,
1993 provided by the Industrial
Page 27
Kaplan, R., Measures for Manufacturing Excellence, Harvard Business School Press, Boston,
Koenig, M.
MA
1990.
"Bibliometric Indicators Versus Expert Opinion in Assessing Research Performance," Journal of the
(34:2), 1983, pp. 136-145.
,
American Society for Information Science,
Krogh, L.C.
"How 3M
Evaluates
Its
R&D Programs," Research Technology Management, November-December
1988, pp. 10-13.
Lehnerd, A., "Revitalizing the Manufacture and Design of Mature Global Products," Technology and Global
Companies and Nations in the World Economy, ed. B.R. Guile and H. Brooks, Washington D.C.; National
Industry:
Academy of Engineering
Levitt,
Press, 1987, pp. 49-64.
R.M., "Process Measures
to
R&D Scheduling Accuracy," Hewlett Packard Journal,
Improve
(39:2), 1988,
pp. 61-71.
Lewis,
C, "Monitoring
R&D
Projects Costs Against Pre-Specified Targets,"
R&D Management, (23:1),
1993, pp.
43-51.
Liberatore, M., and G. Titus, "The Practice of
Science, (29:8), August 1983, pp. 962-974.
Maidique, M.A. and
B.J. Zirger,
"The
New
Management Science
in
R&D Project Management,"
Management
Product Learning Cycle," Research Policy, (14) 1985m pp. 299-314.
Mansfield, E. and R. Brandenburg, "The Allocation, Characteristics, and Outcome of the Firm's Research and
Development Portfolio: A Case Study," The Journal of Business. (39:4), 1966, pp. 447-464.
J. Rapoport, J. Schnee, S. Wagner, and M. Hamburger, Research and Innovation
Corporation, Norton, New York 1971.
Mansfield, E.
,
W.A. and J. Pederson, "Measuring
Management, July-August 1989, pp. 19-21.
the Performance of
Meinhart,
R&D Professionals," Research
Modern
in the
Technology
Meyer, M. and J. Utterback, "The Product Family and the Dynamics of Core Capabilities", Sloan Management
Review, Spring 1993, pp. 29-48.
Meyer, M. and E. Roberts, "Focusing Product Technology
(29:4).
Summer
for Corporate
Growth, "Sloan Management Review,
1988, pp. 7-16.
Meyer, M. and
the
J. Utterback, Moving Ideas to Market and Corporate Renewal, International Center for Research
Management of Technology, Working Paper #69-92. Massachusetts Institute of Technology, July 1992.
Moore,
and Baker, N.R., "Computational Analysis of Scoring Models
J.R.
Management Science,
Moser, M.
,
(16:1)
December
"Measuring Performance
in
for
R&D
in
Project Selection,"
1969.
R&D
Settings,"
Research Management, September
-
October 1985, pp. 31-
33.
Packer, M.B., "Analyzing Productivity in
R&D Organization," Research
Technology Management, January-
February, 1983, pp. 13-20.
Patterson.
W.
"Evaluating
R&D
Performance
at
Alcoa Laboratories." Research Management. March
-
April 1983,
pp. 23 -27.
and T.L. Case. "Implementing
August 1991, pp. 37-42.
Pickett, J.R.
Porter. J.G.
,
"Post Audits
--
An Aid
to
E.xpert
Systems
in
R&D." Research Technology Management.
Research Planning." Research Management. January 1978, pp. 28
July-
30.
Paiic 2S
Quinn,
J.,
The Intelligent Enterprise, Free Press,
New
York,
NY,
1992.
WP
Roberts, E., "Strategic Management of Technology: Global Benchmarking (Initial Report)", Working Paper
#101-94, Sloan School of Management, Massachusetts Institute of Technology, Cambridge, MA.
Rockwell,
and M.
J.
Particelli,
"New
Product Strategy:
How
the Pros
Do
It,"
May
Industrial Marketing,
1982, pp.
49-60.
Ruch, W.A.
,
"Measuring Knowledge Worker Productivity," paper presented
at the
Conference on Productivity
Research: American Productivity Center, April 20, 1980.
Sanderson, S. and K. Uzumeri, "Managing Product Families: The Case of the Sony Walkman", Rensselaer
Polytechnic Institute, Troy, NY, October 1993.
Sanderson, S. and K. Uzumeri, "Cost Models for Evaluating Virtual Design Strategies in Multicycle Product
Families", Center for Science and Technology Policy, Rensselaer Polytechnic Institute, 1991 and forthcoming
Journal of Engineering and Technology Management.
Schon, D., Technology and Change: The
Schainblatt, A.
May
,
New
"How Companies Measure
Heraclitus, Delacourte Press,
New
York,
NY, USA,
in
1967.
the Productivity of Engineers and Scientists," Research
Management,
1982, pp. 10-18.
Smith, P. and D. Reinertsen, Developing Products in Half the Time,
Souder, W., "Analytical Effectiveness of Mathematical Models for
(19:8) April 1973, pp. 907-922.
Stalk, G.,
Von
Nostrand,
New
York,
NY,
1992.
R&D Project Selection," Management Science,
"Time - the Next Source of Competitive Advantage", Harvard Business Review, July-August 1988, pp.
41-51.
Steele, L.,
"What We've Learned:
March- April 1988,
Selecting
R&D Programs and Objectives," Research Technology Management,
pp. 1-36.
Tomatsky, L.G. and K.J. Klein, "Innovation Characteristics^ and Innovation Adoption-Implementation,". IEEE
Transactions on Engineering Management, (29:1), 1982, pp. 28-45
Uhrich, K. and K. Tung, "Fundamentals of Product Modularity," Working Paper #333-91 -MSA, Sloan School of
Management, Massachusetts
Urban, G. and
J.
Institute of
Technology, Cambridge,
MA,
September 1991.
Hauser, Design and Marketing of New Products, Prentice-Hall, Englewood Cliffs, N.J., 1980.
Utterback,
J,
Mastering the Dynamics of Innovation, Harvard Business School Press, Boston,
Utterback,
J.
,
M. Meyer,
T. Tuff, and L. Richardson,
"When Speeding Concepts
to
MA,
Market Can Be
1993.
A
Mistake,"
Interfaces, (22:4), July- August, 1992, pp. 24-37.
von Remoortere, F. and R. Cotterman, "Project Tracking System Serves as Research Management Tool," Research
Technology Management, March- April 1993, pp. 32-37.
Wheelwright.
S.
and K. Clark, Revolutionizing
New
Product Development, Free Press,
New
York, N.Y., 1992.
Page 29
Figure 1
Product Family Evolution
Platform Renewal and New Product Generation
Time
Development of Original Platform Architecture
Mult'^ples
Figure 2
Platform Change
Initial
Platform Architecture:
Common
Subsystems and Interfaces for Multiple Products
Derivative Products
Platform Extensions: A new generation where number and types of subsystems
and interfaces remain contant, but where subsystems and interfaces are enhanced.
Derivative Products
Platform Renewal: A New Architecture, where subsystems and interfaces from prior
generations may be carried forward and combined with new subsystems and interfaces
in the
new
design.
CSTl
<
a.
E
3
o
u
c
_o
c
u
m
• •
g
3
0£
00
E
a
oo
4—*
a.
oo
C
_o
C3
GO
C
_o
c
oo
C3
E
k.
<£
c
3C
e
oo
Platform Efficiency
(Follow-on / Platform Costs)
o 00
o p
o —
fe
o
NJ
CIS lA,
Initial
{?>
O O O O
NJ
—
—
U —
o
0\
00
——
.
^
5 f2 S
3 K
n
CIS lA, Fl
CIS lA, F2 T
CIS IB,
Initial
<*
3
CIS IB.Fl
3
^
>
fl
a
c
n
CISIB, F2
t3 i;^
o
o
H
CO
O
ON
1
u
a
iM
3
Cm
in
o
a
CU
!2
^
o
o
u
a.
en
c/T
3
CD
>%
s
3
b
Ou
O
S
-a
o
u
ON
U
S
o
On
O
u
u
o
o
o
o
00
NO
ON
00
s
OO
oo
c
00
E
a
u>
3
c«
«J
00
IT)
00
r
Platform Effectiveness
8
2
(/I
3_
2
2
C/3
O
a
s
«
2
oo
1>J
2
O
00
O
o
a
Cm
o
a
7S
E
CO
U4
u
in
00
00
u
u
1
.o
o E
"I
a" o
u
b a.
c
u
Z
-a
c
a
i)
o
Cu
a.
(A
m
Q
OS
c
o
ON
3
<
2
OS
Q
T3
O
OS
as
00
oo
00
oo
so
oo
O
O
E
o
8
1
4
1
1
X
Platform Effectiveness
O
8
o
8
o
8
00
O
8
O
O
8
%
to
o
8
Percentage of Total
R&D Staff
— tou>*.Lno-Joovo
oooooooooo
1976
%
Staff
Development
Allocat
Platform Efficiency
GO
n
to
3
PC
o
In
O
00
'^
o
o
TO
o
to
o
p
o
o
"3
3
O
—
c
:t
£
«j
—
c
o
^
—
i^
r
^
2
Appendix 2
Classirication of
Platform
Architecture
Key Events
in the
Evolution of a Product Family
Appendix 3
Technology and Market Newness
Assessed for Family A
Technology Newness
New, Unrclalcd
Q)rc
Tixhne)loi!\
WluTf PI
IS
.
till'
Orii;l/ial
ilw hu\c i>n>du<
I
In
Central
term
v
liijoriiuilioii S\ \rrifi
of con' ra
and murker appllcarlon.
Now. Related
1^')
Core Techn(>loi:\
V\2
I'KI
Major LJihaiiceniciit
""
Core Technoloiix
1'4
Exislini;
Minor lmpro\einent
1'5
IMl
I'S
EmsUiii: Core TechnolouN --
No
Chani:e
Chanj^c
in
One
Change
in
Two
Change
in
Alf
Market Applications Newness
Customer Groups. Functional Uses. Distnbulion Channels
PrcxJuct
li/tolo{;\
I
he
1
2.
3
4
Ic\ els ol tcchiioloyicil clijuigc
(McNcr aiitl Roberts,
I9K6):
N linor improvements to core techiiologj (ies) that exist in earlier products
Major enhancements to existing core technologyCies)
The addition of a new core technolog> which is combined or otherwise integrated with existing core
teclinologies in earher products to create the new product
.
Ne w umvlated core
,
tt"chnoiogy
- Nhirkcl Applications Newness; For each product, observe relative to iUI prior products, ;ui\ chiiniie m a) customer
group, b) functional use of the product by the same or new customer group, or c) distnbutioii chaimel
1
No change
ill
(a), (b)
2
Change
in
.V
(.~Ii;ui!.'e
in ;inv
4
(
one of
oncurrcnl
or (c)
(a), (b)
l«o of (a),
cliaiiiie ui (a),
or (c)
(b) or (c)
(b) or (c)
en
«
"S
a
3
a
E
-a
=
c ^
C«
•c
E
c
0.
Appendix 5
CPI Dollar Adjustments
(Multipliers to
Year
;
CPI
1983
'
1993 Dollar Equivalents)
MIT LIBRARIES
lljllll
|ll|l
I'l
ill
III
II|1M|I||III||||
I
3 9080 00932 7732
^197
m8/:
Date Due
APR.
^-^•^
1
19«
1a
!£}(};
1
SEP 3
1999