Modeling the Innovation-Pipeline
Andreas Jost
DaimlerChrysler Research & Technology
IT for Engineering and Processes - Digital Engineering Competence Center
Wilhelm-Runge-Strasse 11 - 89013 Ulm - Germany
phone: + 49 731 505 2364 - fax: + 49 731 505 4400
mail: andreas.jost@daimlerchrysler.com
Tobias Lorenz
System Dynamics Group Bergen, Universitaet Stuttgart
Nordbahnhofstr. 179, 70191 Stuttgart - Germany
mail: space56@freenet.de
Gerald Mischke
DaimlerChrysler Research & Technology, Project Controlling
Mercedesstr. 137 - 70327 Stuttgart - Germany
phone: + 49 711 1720742 - fax: + 49 711 1734946
mail: gerald.mischke@daimlerchrysler.com
ABSTRACT
To generate innovative products regularely is a key success factor in established industries.
Major companies frequently outpace each other with innovations and product
offensives. But what are the effects of such initiatives? And how can companies
organize their innovation pipelines in order to successfully manage such ventures?
The process in which innovations are developed and integrated into marketable products is
highly complex and can be organized in various ways. An important distinction introduced
by this paper is to separate between product development processes and processes for
innovation generation. In established industries the first ones regularly initiate product
development projects and strive to meet certain launch periods. The latter are problem-
solution oriented and driven by the search for new, innovative concepts. They are
characterized by risk and a high degree of uncertainty regarding success und completion
time.
Both processes connect and interact in various ways. Nevertheless there are two general
and contrasting alternatives. They can either be organized as processes which are tightly
coupled and integrated by associating innovative ideas early with new product projects, or
rather independently, integrating innovative components into product concepts in later
phases of the product development process. In this paper we refer to these alternatives as
Integrated Innovation Pipeline! and Shelf-System Innovation Pipeline. Both show distinct
characteristics and requirements.
" An innovation pipeline in this context is interpreted as a (timely) structured flow of innovations and product projects
1/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
Being aware, that in reality innovation pipelines often represent a mix between these
polarizing alternatives, it is essential to understand the basic dynamics of each alternative,
in order to explore policies in a mixed environment.
This paper introduces a work-in-progress-model of such innovation pipelines oriented at
the typical structures in the automotive industry. In comparison to classical views (mostly
linear downstream approaches) the aim was to broaden the perspective by including typical
delays and non-linearities of such innovation pipelines. This was achieved by developing a
system dynamics model that pins down the structural elements and relations. The intent of
the developed model is to explore the dynamics and characteristics of alternatives in
structure, organization and policy within typical automotive innovation pipelines. The
presented work also adds a new perspective and approach to the study of multi-project
product development within the field of System Dynamics. It is the result of a five month
process of modeling and exploration in close interaction with internal and external experts
in the field of innovation and product development.
INTRODUCTION
Product development processes are without any doubt innovative processes. However in
order to capture the characteristics of automotive innovation pipelines this paper
distinguishes between the innovation process and the product development process..
Further on both concepts are combined in order to develop a comprehensive model of their
interaction: The so called Innovation-Pipeline.
THE PROCESS OF INNOVATION GENERATION
Classical views of the innovation process” typically refer to a sequence of activities
represented in a downstream process. Commonly, these are described as an arrangement of
around four phases (e.g. Idea generation, Idea formulation, Problem solving, and
Utilization). Please note Figure 1. An innovation process is typically characterized by a
Technical
feasibility
recognition
Search, research
and development
activity
Fusion into
design concepts
(idea)
Solution
implementation
Potential
demand
recognition
Evaluation
Recognition Idea Problem
Idea generation formulation solving
Utilization
Fig 1: A classical view of the innovation process (adapted and enhanced from Myers/ Marquis (1969))
> Compare Myers/ Marquis (1969), Gebert (1979), Stern (2003) and various others.
2/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
high number of ideas (projects) in the early phases, combined with mechanisms to develop,
evaluate and select the most promising ideas in the course of the following phases.
Numerous ideas are often generated in parallel. Further on in the process, the non-
applicable ones are sorted out and valuable ideas are carried further through the more
resource-intensive phases of development and evaluation. Outcomes of the innovation
process are innovative solutions/concepts, more or less ready for application. The
utilization stage refers to the introduction of a novelty as a product or process. Figure 2
provides an alternative illustration indicating the funnel characteristic of such processes.
Pre- Innovative
Ideas studies Studies Projects Concents
Fig 2: Process of Innovation Generation
In many cases, innovative concepts are not to be introduced separately but need to be
integrated into existing products (for example, new car safety systems needs to be
integrated into a new vehicle for utilization). At this point the link to the product
development process becomes apparent.
THE PRODUCT DEVELOPMENT PROCESS
In the automotive industry, new products are developed in a project-type form, following a
product development process. This process can be described as a downstream phase model
composed of several stages. In the following example (see Figure 3) five phases are
distinguished: Concept, Pre-development, Development, Production and Sales. Product
projects follow through such processes in cycles. Typically, as soon as a new model is
launched into the market, the subsequent development project for its next generation
vehicle is already in preparation.
3/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
Pre- Serial
Concept Development // Development
Production Sales
Fig 3: Product development process
As mentioned, product development processes are innovative processes. However, they
differ in various ways from the classical view of typical innovation processes as described
above.
In the automotive industry product development processes are planned and managed
towards an indicated launch period. Therefore, they depend to some degree on operational
security and the limitation of risks and experimentation. In other words: If a new and
innovative concept is not available in a certain quality on time, product development is
forced to use proven concepts.
Contrarily, innovation processes (interpreted in the classical view) are driven from the
other end. Initiated by the identification of a potential demand (e.g. a technical problem)
and the recognition of potential in technology, the focus lies on the development of
innovative concepts. Typically such processes show a high degree of risk and
experimentation. If and when an innovative solution is finally found is highly uncertain.
As described above, at one point the innovation process and the product development
process have to meet. Product development projects frequently need to integrate available
innovations generated by innovation processes into the product concepts within the product
development process. Depending on the definition applied, new automotive vehicles
sometimes integrate several dozens of innovative concepts, though the border between
innovative concepts and improved traditional concepts is rather vague.
There are various ways innovation processes and product development processes interact.
Nevertheless there are two general and contrasting alternatives. They can either be
organized as processes which are tightly coupled and integrated by associating innovative
ideas early with new product projects, or rather independently, integrating innovative
components into product concepts in later phases of the product development process. In
this paper we refer to these alternatives as Integrated Innovation Pipeline and Shelf-System
Innovation Pipeline. Both are described in the following. Being aware that in reality,
innovation pipelines often represent a mix between the two different alternatives, the
approach described aims at the understanding of the basic dynamics of each alternative as a
basis for the understanding of a mixed environment.
4/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
MODELING APPROACH
The model is based on a dual-pipeline approach which represents the typical structure of a
product development process based on a feedback system incorporating investments and
resources. It explicitly defines a structure for further analysis and policy testing.
The first pipeline represents the classical innovation process characterized by its high
degree of uncertainty and the sorting out of conceptual ideas through iterative development
and evaluation activities. This part of the pipeline represents highly innovative activities
with high risk, which solutions (yet) cannot be expected for a certain launch date. In Figure
4 this pipeline is represented by the industrial research phase. The second pipeline
represents the product development process, which managed towards an intended launch
period of new product lines. In this process product development projects are initiated
regularly and carried through subsequent maturation stages. In terms of innovations,
projects need is to pick up available innovations and integrate them into product concepts.
The product pipeline includes all activities of conception and integration of classical and
innovative concepts into a product.
INTEGRATED INNOVATION PIPELINE
In the Integrated Innovation Pipeline view, innovative ideas are picked up early in the
product development process and integrated and developed tightly coupled along the
conception of the new product. Figure 4 illustrates this idea. The innovation process as
described above is referred to as the industrial research phase, which is then followed by
the product development phases of concept, pre-development, serial development and
production. Each phase requires corresponding resources.
Industrial
Research
Innovative
Concepts
Product Pre- Serial predaeaion
Concept Development Development,
Fig 4: Structural model - Integrated Innovation Pipeline
SHELF-SYSTEM INNOVATION PIPELINE
The Shelf-System Innovation Pipeline represents an alternative to the structure described
above. It is associated with the idea to extend the first pipeline in order to pick up generated
innovative ideas and develop them (independently from certain product projects) towards a
3/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
higher degree of maturity e.g. into innovative modules or components. These would then be
made available in the so called Shelf-System. The Shelf-System represents a stock of mature
concepts and tested modules which are available to be integrated on demand into a product
project at significantly later stages. The entire construct (referred to as Shelf-System
Innovation Pipeline) is illustrated in Figure 5. The introduction of parallel and independent
stages in the innovation pipeline leads to corresponding resource requirements.
Industrial Module based
Development activities
Research :
per Innovation
Innovative Innovative
Concepts 2 Modules
Product Pre- Serial Production
Concept Development, Development,
Fig 5: Extended structural model - Shelf-System Innovation Pipeline
MOTIVATION FOR SYSTEM DYNAMICS
This study investigates problems characterized by elements (Feedback, Non-linearities and
Delays) which have always been an explicit strength of the System Dynamics
methodology. In the following these effects are explained in more detail.
FEEDBACK
Major feedback is generated by the fact that the resources, which are attained via the
market introduction of innovative products, are again invested into the innovation pipeline
to contribute to the development of further innovative products. A macro feedback
influencing the various aspects of the Jnnovation Pipeline is a reinforcing one.
NON-LINEARITIES
Several non-linearities have been identified while structuring the problem:
= The effect of the innovational content of a product on its attractiveness.
= The resource backflow generated by innovative products in the market.
= The impact of varying resource availability per phase on project duration and
pipeline throughput.
Furthermore, the main allocation of the resources is typically executed on a yearly basis,
which can also be regarded as nonlinear. The non-linearities governing the impact of the
resources and the impact of the innovation content are caused by the concept of diminishing
6/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
marginal returns. The marginal result of the invested resources into the R&D sections
decreases. Similar applies to the marginal success of the innovational content of a given
product on the market.
DELAYS
Delays are obvious in innovation pipelines, as phase durations regularly add up to several
years from the first concept to the production of a new product. Further delays are brought
in by market cycles and the generation of resource backflow. Therefore, innovation
pipelines are confronted with significant time delays that need to be considered in the
discussion of alternative structures and management policies. One of the main themes of
this paper is that these delays are often misunderstood, underestimated or even outside the
range of consideration.
LITERATURE REVIEW
The review of previous work related to the focus of this paper identified several
publications applying System Dynamics to questions of R&D and product development.
Milestones have been the works of Roberts (1964), Cooper (1980), Abdel-Hamid (1988),
Ford (1995) and Milling (1996). More recent papers have been published by Ford/Sterman
(1998), Repenning (2000), Lyneis/Cooper/Els (2001), Hilmola/Helo/Ojala (2003), Milling
(2002) and Black/Repenning (2001).
The focus of the majority of the previous work has been on the development of single-
project-models as opposed to multi-project-models. Repenning, for example, explicitly
states, that “while interest in managing R&D function (as opposed to specific projects) is
on the rise, there are few formal models focused on understanding multi-project
development environments.”
The approach presented in this study suggests a model motivated by current, practical
questions. In contrast to previous models that focus on resource allocation (comp.
Repenning (2001)), the focus is on organizational structure and the incorporation of
feedback from attained resources on possible investments into the innovation pipeline.
The presented model focuses on a structural approach oriented at organisational structures
found in the automotive industry. It differs from previous approaches which focus on a
project and activity based perspective (comp. Ford (1995)) and attempt to incorporate “all
the significant structures developed for other systems and which apply to projects into a
single model” into a multi-project perspective. By the integration of the identified
feedback the presented model may be considered as related to the approach presented in
Milling (2002). Main driving factors in the model are available and invested “resources”
which represent human as well as financial resources. The diffusion part of the presented
model is simply computed via table functions, but may be extended to a sector of its own
* Repenning (2000), p. 174
* Ford (1995), p. 27
W270
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
(e.g. related to classical approaches developed by Bass (1969) or more recently by Maier
(1998)).
The aim of the presented model is to provide a tool for strategic analysis and policy testing
in organisation and structure of innovation pipelines. Some of the presented ideas may be
related to the concepts developed by Powell/Schwaninger/Trimble (2001) as well as to the
work of Zahn/Dillerup/Schmid (1998) regarding the choice of production systems.
MODELING APPROACH
In the following the modeling approach is introduced by the illustration of the dominating
feedback loop and the definition of the model boundaries. After that, conceptual model is
explained as a basis for a better understanding of the explicit model structure.
DOMINATING FEEDBACK LOOP
A straight forward access to the central hypothesis of the described model is provided
through the description of the main feedback loop. Investments in research and
development activities lead to an increased throughput of the innovation pipeline resulting
in an increased number of innovations in products. This, in return, increases the
attractiveness of the products, resulting in more units sold. Units sold again contribute to
the generated resources, also increasing possible investments in R&D (comp. figure 6).
Generated
resources
Investment of
ft ¢ R&D a]
Throuput of Innovation in
Maing? roducts
Piney
Fig 6: Main feedback loop
MODEL BOUNDARIES
The model boundaries have been drawn narrowly to keep a clear focus on the problem and
its characteristics (See Figure 7). Exogenous parameters and model inputs are the average
market potential and the average number of innovations per product. Both variables could
be explained by demographic changes or changes in the consumer behavior, but are simply
assumed to be static in this model. Also exogenous and static are the average resources
8/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
required per project and the attained resources per unit introduced into the market cycle. In
practice R&D ratios are, according to interviews, rather stable over time. In the model,
therefore, they are assumed to be a constant fraction of the generated resource backflow.
Endogenously modeled are the effects of the two major feedbacks conceptualized in this
problem context: the influence of the resources on the throughput of the innovation pipeline
and the effect of innovations per product on the resource generation when introduced into a
market cycle.
/ Influence of >
titi
tel lect sel Omitted
Macroeconomic
influences
‘Average market
&
potential per unit saint
Average number of
Demographic innovations per unit
development
Average costs
Ber proiect Endogenous
Non-industrial | Attained
research resources Effect of resources on
per unit project duration
Effect of number of
innovations on market
success
R&D ratio
a
Fig 7: Model boundaries
The influence of the competition has been omitted although it would have given the
problem additional and differing dynamics. Nevertheless, this advantage would be bought
with the loss of the specific focus. Furthermore, the dynamics of research competition have
already been studied and might be implemented. Non-industrial research, as well as
macroeconomic and demographic development, is not implemented in this study, but might
be included through time series at a later stage.
CENTRAL ASSUMPTIONS
A major assumption in structuring the problem and developing a model is based on the
identification of key drivers of successful innovation processes such as the abstract notion
of resources, which include work as well as capital. The specific effects of and on
individuals in development processes as researched by Ford (1995) are not accounted for,
because the focus lies on the comparison of two different organizations of the R&D section,
whereas human interaction is assumed to be similar. Additionally, the market success is
simplified by a function dependent on the number of innovations in a product.
9/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
The policies governing the distribution of the attained resources in this model are assumed
to be partly revenue-based and partly requirement-based. For example, the total amount of
resources, which are spent in the R&D sections, are defined as a fixed fraction of revenue
(the so-called R&D ratio, here assumed to be 5 %) while the distribution of the resources
within the R&D sections is realized in regard to the requirements. This means that the total
required resources are calculated first, then (for each section) the fraction of its
requirements to the total requirements is computed. Finally, the resulting fraction times the
actual number of available resources is distributed to that section:
Required resources .
x Available resources
Total required resources
CONCEPTUAL MODEL
In accordance to the identified organizational elements as described above, the model is
organized into four sections. The three main parts are concerned with the development of
innovative concepts, product development and resources attainment. An explicit section is
devoted to the two structural and organisational options. Figure 8 explicates the various
aspects of each of the four sections.
Innovation Process Product Development
Generation and evaluation of innovative Development of new products following
ideas.
Development into innovative concepts to.
be integrated into products.
the phases of product development
processes and the integration of
innovations.
Production and
Resource Allocation
Market cycles of products and the
distribution of attained resources to the
innovation pipeline.
Innovation Distribution
Comparison of two different
scenarios determined by the
modus innovations are integrated
into product development.
Fig 8: Conceptual model
10/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
EXPLICIT MODEL STRUCTURE
The aforementioned sections are linked and interrelated in various ways. Figure 9 illustrates
the main path. In the following all four sections are described in detail.
Innovation
Process
Production Innovation
and Distribution
Resource Allocation
QT
ve
Product
Development
Fig 9: Interrelation of the model sections
INNOVATION PROCESS
The innovation process is represented as an aging-chain. It is designed to regulate the
relation between the number of resources and the time needed to complete a given project.
The number and completion time of the innovation projects in every stage is influenced by
the available resources. The structure, simplified for better understanding, displays the
organizational alternatives characterized by different outflows: Innovative concepts can
either be taken out of the research stage or developed into innovative modules and made
available in the shelf-system. The time needed is treated as an average time span around
which the actual development durations are distributed. Innovations within each phase are
assumed to be uniform and interchangeable.
PRODUCT DEVELOPMENT
The product development structure is also formed by an aging chain. In contrast to the
innovation process every project is represented by an individual subscript. Maturation time
is realized as a pipeline delay. This approach was taken in order to accommodate the
characteristic of fewer risks and operational planning towards a certain launch period.
The balancing feedback, as introduced before, is also active in this section. For the model,
it was assumed that every eight time units a new product development project is launched.
Nevertheless, starting dates as well as duration depend on the availability of sufficient
resources.
11/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
PRODUCTION AND RESOURCE ALLOCATION
In this section of the model, market cycles are simulated. Any product being introduced
into the market by the production stage initiates a market cycle which generates new
resources. This is a highly non-linear process. The sales distribution of product units is
assumed to follow the course of a typical model cycle which is characterized by two peaks:
the first is caused by the new product reaching full production capacity and market demand;
the second (local) peak is caused by the so called model-upgrade. This curve has been
discussed with the interviewed experts and incorporated through a table function depending
on the time elapsed since production start (See Figure 10).
Potential cars sold per month
4,000 LA
0 2448 72 96 «4120 «144~=C168~=SC«192~=S:«SCSC*« AO
Time (Month)
Potential cars sold per monthflined3] : marketcycle ———————__—_—_ Product/Month
Fig 10: Typical market cycle of a product line
The output of this variable is the percentage of the total sales over the entire market cycle
per time unit. It is then multiplied with the market potential and the effect of the
innovational content within a given product line. This is assumed to be S-shaped, i.e. an
innovation sensitive part is framed by a phase where a product with relatively few
innovations still sells quite reasonable and a phase where (above a certain number) further
innovations do not increase market success significantly anymore. Through this approach,
generated resources of a given product line correspond to its respective innovational
content. In this section it proves particularly useful, that each product project is described
by an individual subscript. Each subscript can be associated with a specific number of
innovations. The individual market success of a product line is then summed up in total
product units per time unit.
Another nonlinear process is the redistribution of the attained resources. Similar to
industrial practice it is calculated on a yearly basis. In this study five percent of the total
units in a year multiplied by the conversion factor of units into resources are reallocated to
the innovation pipeline for the next twelve time units.
12/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
INNOVATION DISTRIBUTION
As stated above, in the Integrated Innovation Pipeline a product project picks up
innovations early whereas in the Shelf-System alternative these innovations are developed
further and kept available to be integrated into product concepts at a later stage. Both
alternatives are realized via IF THEN ELSE functions. In order to capture the innovational
content of products, the number of available innovations is considered as an inflow of a
stock. In the market cycle this might be interpreted as the innovational content of a given
product line.
In order to calculate the standard innovation distribution, the influence of uncertainty is also
considered. This takes into account that innovations are still in early stages of development.
Throughout further development of the final product, there still is a potential loss of
component projects due to incompatibility or problems discovered within the integration
procedures. The functions for the contributing inflows of the innovation distribution are:
Integrated Innovation Pipeline:
IF THEN ELSE(Concept Start[line]=1, "Components:Development"*Percentage of used
innovations, 0)
Shelf-System Innovation Pipeline:
IF THEN ELSE(PreDeveloping[line]=1, "Components: ShelfSystem",0)
See Figures 11-13 on the following pages for a structural overview of the individual
sections. Figure 14 indicates the alternative ways innovations are picked up by product
projects, integrated and developed referred to as the Integrated Innovation Pipeline and the
Shelf-System Innovation Pipeline.
13/27
Modeling the Innovation Pipeline
A. Jost, T. Lorenz, G. Mischke (2005)
(uappry oz syuejsu0d pur sapquizea Mopeys ‘paryTdurs st yo}ays ay} Surpueyszspun 19}}9q 104) ssav01d uoFLAouuy :[] 31
jee)
Bujsiuio1, yon
yey vonealg Pepl
yeouognpold
voiresey gg SZ
sisting © \
sswongiedeu, "10; 240nenOUUL
ope ulioasu oneeta Rani
sseoons _
ne aa rn
uagersouiy
seoinosay Jo waya
yey voneRUL Uo
Sipe sevoons Uo, quswidojenagaig seoinosay Jo ays papaau seoinosey
seounosal jo PPR za te
uy papaau seainosey _/
aiaisees ion
qusuidojaneg
ssquauodwiog
yuewdojnepieussenpold
40) suogeacuyy
wasisyeus
———_| ssuieuadies,
sey uone!
¥ Youeasey 10) saaindsey
qoudojenag +
21d 10} seoinosey
wowed
3p] 104 se0inosey
aveoduos fee
10 ss0inoseyarey
14/27
Modeling the Innovation Pipeline
A. Jost, T. Lorenz, G. Mischke (2005)
(uappry a1e syue}su0d pue sapquizea Mopeys ‘paytydurs st yd}ays 9y) SurpueyszapuN 19339q 10,4) JUSUIdOPIAap JonpOIg :Z7] BI
pnpoid 10}
seoinosey GY
yde0u09 104 ss0inessy
qusudoraaa gies
+ quauidojanaqeid
sop ssounosay
quauidojanaqgieuesynpord
105 papaau sacinosey
uo sseinosey so aH
quaudojenagievesynpold
195
quewdoyansqpnpeg
10 papeau sacinozay, Sundsouosionpotd
r 430 0H ae a‘ . Jo1 papaou seoinosey
. Bundsoueayonporg +
so) au,
quoudojeaagienespnpod wouldoresqpnpald
4p But at
Buidojanaqieni Burdojeneqard + Burdojsaag + yeys yuo
uownporg =| usudojanaqienas| qusuidolsncqelg beat ydeou0l lag
venpoid vanpoid ypnpold venpord ©
15/27
Modeling the Innovation Pipeline
A. Jost, T. Lorenz, G. Mischke (2005)
(uappry a1v syue)sUOD pue saTquizea MopeYs ‘part{durs st yd}oys ay) SUIPUL}s1IpUN 19}}9q 10.7) UOLBIOT[ 9dANo0sai puv UOT INpOA :¢] Sly
easuayo yonpoud
40) seounoseyesxg
seokmay ag | sek
0) enpuoneiead + uo ul dey Ul
Pasn saainosay
ae
@8y Uo yuads saainosoy
I
GRY UI pasn seainosey ary
S$ 101 aigzH1200
seainosey
yea,
seek au0 04s
jonpoid sad
suonenouul |eWHON
GRY 0} pareoo}1e.
uonse1jequnosoy
x
‘seounosey MeN
s9|es
yonpoid mou
suonenouus Jo 94)3
ales
ee
seaunosey peinqusid 120k
= S| jad supanpolg
Pos payeinwnooy
sueo plog
ie
/syun-pnpold pios,
paeinwnooy |e10)
pjos sie9
ms
yluow sed syur-onpoid Me
uwwow sed
los sie0 leNua}od Pos
40) suoneAOUU
ssa180lg
uononpoig
eis Uononpolg
Ss
A oynanssieg
lenuajod yayew jewon
16/27
Modeling the Innovation Pipeline
A. Jost, T. Lorenz, G. Mischke (2005)
Fe
uononpeld Pid]
Budopaagiey
voronpoly qwewdojaragieues
sionpalg \ rH Og
uognqunsiq
wajsks-s84s,
aurpodid
WOTLAOUUT Wa}sAs-Jpays “SA BuTJadId UOKVAOUUT pajessazUy pT BI
buydojmagarg> urdojaneg * veIs Wdouey
quawdo)magaig yda0ueg $C)
SRA Og
spnpalg
uidojanapag rf
uonnquisiq
uonenouu
Puepueys
up were ion!
—s
7 ;
So _ —
ganseess <—% inaweeoo - seme, =O
swxong aad aus ‘ps suoKENoU, waned ite
17/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
DISCUSSION AND BEHAVIOR
In order to explore the behavior of the developed model various scenarios have been
investigated and simulated. Many of them are related to current industrial questions. In the
following section the scenarios steady state, response to changes in market conditions and
response to product offensives are introduced and selected model insights are presented.
STEADY STATE
The steady state of both alternative model structures is calibrated to comparable runs and a
steady flow of projects. This includes an equal number of total resources available, regular
project launches (new launches every eight time units) and a constant number of
innovations per product. Through precise calibration of key variables, a relatively stable
condition could be reached (see Figure 15). The term base refers to the integrated
organization of the innovation pipeline; the abbreviation she/f refers to the shelf-system
alternative. The constant oscillations are caused by the number of products on the
market (each in a different stage within the market cycle) and the summarized
distribution of product units over time.
The steady state was calibrated to enable a comparison between the two alternative model
structures in a series of scenarios. In order to assure a constant project flow in the time
period under investigation, an even project distribution and a constant market pull were
assumed and sufficient resources were generated.
Product-Units per month
50,000
47,500
POR AUAVAVAUAVAYAVAVAVAVAUVAVAVAVAVAVAVAVAVAVAVAVAVAVAVANAVAT AVA
42,500
40,000
0 24 48 72 96 120 144 168 192 216 240
Time (Month)
"Product-Units per month" : base. ——_—__—__—_ Product/Month
"Product-Units per month" : shelf. ——————_ Product/Month
Fig 15: Steady state
The graphs of both runs in Figure 15 are difficult to separate, as the parameters are
calibrated to provide an equal run for both alternative model structures. This calibration sets
the basis for further comparison with selected scenarios.
18/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
A major problem in constructing and calibrating the model was caused by the high
sensitivity of both structural alternatives to initial values and changes in decision
tules/policies. Although the model is partly initialized with formulas, the remaining
parameters are extremely sensitive while generating a somewhat fragile steady state. In this
aspect the model confirms the repeated statements of experts that innovation pipelines
of this character are highly sensitive to changes in values and policies. One of the
major problems is that a steady state, as described above, is not achieved in industrial
innovation pipelines, causing significant problems by the fluctuation of resources and
throughput.
RESPONSE TO CHANGES IN MARKET CONDITIONS
The first scenario which was investigated is the response of the alternative pipeline
structures to a sudden change in market conditions. The underlying idea was to test the
reaction to a change in customer preferences caused by an external effect. This was
simulated through the implementation of two step-functions causing an abrupt
increase/decrease in the number of innovations required to meet a certain market potential.
The results of these experiments indicate a noticeable faster response of the shelf-system
version compared to the integrated (base) system (see Figure 16).
Product-Units per month
200,000
150,000
100,000
50,000
0
0 24 48 72 96 120 144 168 192 216 240
Time (Month)
"Product-Units per month" : shelf_up ——————————_ Product/Month
"Product-Units per month" : shelf_ down ————_ Product/Month
"Product-Units per month" : base_up_ —————————_ Product/Month
"Product-Units per month" : base_down ——_—_ Product/Month
Fig 16: Response to sudden change in market reaction to innovational content
The figure above shows four graphs, two for each pipeline organization. In the “up” cases,
a step function increases the number of required innovations to reach a certain market
success. In the “down” cases, the respective innovations required are decreased. Therefore,
in the latter case, similar market success is easier to attain compared to the steady state run.
This leads to a higher resource backflow and, therefore, to more resources, which are
available for R&D and so finally to a higher innovational content in products, leading to
even higher market success. The opposite argument takes place for an increase in required
19/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
innovations. The faster response of the shelf-system structure (although incorporating
positive as well as negative response) can be considered as a gain in flexibility providing
potential advantages (compare Zahn/Dillerup/Schmid (1998) and the concept of flexibility
as a strategic resource). The departure from the steady state follows an s-shape which is due
to the nonlinear relationship between the number of innovations in a product and the
respective market success. Above or below a certain number, additional or fewer
innovations do not change the total number of products sold anymore and a new stable state
is reached.
First model tests indicate some advantages of the shelf-system structure compared to
the structure of the integrated innovation pipeline caused by its flexibility and the
faster response time. Figure 15 suggests that advantages of the shelf-system exceed the
disadvantages in case of a downturn. This is, nevertheless, to be attributed to the
nonlinear relationship between the number of innovations and the products sold.
Further tests and comparison with empirical groundwork may show whether this can
also be accounted for in reality. Nevertheless, interviews with experts indicate that the
model is considered to behave quite realistically.
RESPONSE TO A PRODUCT OFFENSIVE
The second scenario investigates the model’s response to the initiation of a so-called
product offensive. A product offensive refers to additional initiation of product
development projects in order to extend a given product portfolio. This case was simulated
by an increase in projects launched in the product pipeline. In addition to starting a product
project every eight months, an additional one is launched at time period 12 (between the
first and the second regular one). Figure 17 shows the response for the product units per
month.
Product-Units per month
60,000
Initiation of an extra
product development project
50,000
40,000
30,000
20,000
0 24 48 72 96 120 144 168 192 216 240
Time (Month)
"Product-Units per month" : shelf_prodoff —————————_ Product/Month
"Product-Units per month" : base_prodoff —————————_ Product/Month
"Product-Units per month" : base -————————————_ Product/Month
"Product-Units per month" : shelf ————————————— Product/Month
Fig 17: Product offensive with no extra resources or gains in resource efficiency and a high sensitivity
to innovational content
20/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
The attributes “shelf” and “base” mark the steady state runs, “shelf_prodoff’ and
“base_prodoff’ the scenario runs of the product offensive. The displayed parameters
“Product-Units per month” show a clear downward trend which is caused by the
implemented requirement-based resource allocation policy and the initial high sensitivity to
the innovational content in the final products. With an increase above the normal number of
product projects, the resource requirements in the product pipeline rise (see Figure 18),
whereas the resources requirements in the innovation pipeline remain equal. With no extra
resources or gains in resource efficiency, this causes the resources distributed to this
part of the innovation pipeline to rise, leaving fewer resources for the other parts of
the pipeline sectors. This includes the sectors for innovation generation, finally (and with
time delay) leading to a decrease in available innovations. The decrease in the number of
innovations is not noticeable in the innovational content of a product line until the new
product projects reach production and are launched into the market. Due to the sensitivity
of the market to the innovational content, a decrease in units sold per month is observable
leading to a further downward trend. The additional superimposed fluctuations in Figure 16
confirm a pipeline behavior described in Braun (1994). Through product offensives,
innovation pipelines tend to develop oscillatory behavior caused by parallel (or close by)
initiation of extra product cycles and varying resource requirements. In addition, resource
shortage and strains within the organizational stages associated with the innovation pipeline
are believed to cause additional problems, such as the effect of fire-fighting (compare
Repenning (2001)) or the 90%-syndrome (compare Ford/Sterman (2003)). As a first
approach to capture such organizational problems also compare Jost/Sauer (2004) and the
concept of Business Syndromes.
Total Resources needed in Product
6,000
Initiation of an extra
5,000 | product develdpment project
4,000
3,000
2,000
0 24 48 72 96 120 144 168 192 216 240
Time (Month)
Total Resources needed in Product : shelf_prodoff —————— Resources
Total Resources needed in Product : base_prodoff —————— Resources
Total Resources needed in Product : base --————————_ Resources
Total Resources needed in Product : shelf —__________ Resources
Fig 18: Changes in resource requirements in the product pipeline caused by the initiation of an extra
product development project
21/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
In a further analysis step the required additional resources, which would be sufficient
to fund the product offensive, are derived based on Figure 17. Therefore, a table
function is introduced, which adds extra resources for the time the additional product
project runs through the innovation pipeline. This was simply realized by comparison of
available and required resources and the development of an according table function.
Results show, that product-units per month continue to show an oscillatory downward trend
(see Figure 19), which is accorded to the sensitivity to the initial conditions and the high
sensitivity of the market demand to the innovational content. The innovation pipeline in its
steady state still represents a quite sensitive and unstable equilibrium.
Product-Units per month
0 24 48 +72 #96 120 144 168 192 216 240
Time (Month)
"Product-Units per month" : shelf_prodoff ———————_ Product/Month
"Product-Units per month" : base_prodoff ——————_ Product/Month
"Product-Units per month" : base. ——————————_ Product/Month
“Product-Units per month" : shelf. —____________—_ Product/Month
Fig 19: Response to a product offensive with high innovation sensitivity and sufficient resources
provided
However, the results indicate, that the initial sensitivity to the innovational content may be
too high. In a further simulation step this sensitivity is reduced by changing the nonlinear
relationship governing the market response through offsets in the market cycles depending
on the innovational content of a given product line. In the simulation run with a market
response which is less sensitive to changes in the number of innovations (implemented as a
linear relation with a relatively small slope), the effect of a product offensive shows an
increase in products-units per time followed by an oscillation with a frequency of
approx. 48 time units (see Figure 20). This ongoing fluctuation (caused by the
initiation of an extra project in spite of providing sufficient extra resources) is a major
factor to be considered when planning product offensives. It is considered to be a major
challenge in industrial practice and supports findings of Braun (1994) and Le Corre/
Mischke (2004). Causes are sought in the typical market cycles of products combined with
changes in resource distribution and throughput of both pipelines. To sensitize, that these
effects might occur through the initiation of product offensives and to develop
strategies how to reduce and mange such fluctuations are of major interest for
22/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
industrial practice. The developed model provides an important basis for further
analysis and opens a new range of possibilities to support this process.
Product-Units per month
60,000 T
of an/extra
juct development project
55,000
50,000
45,000
40,000
0 24 48 72 96 120 144 168 192 216 240
Time (Month)
"Product-Units per month" : base. ——_—__—__—_ Product/Month
"Product-Units per month" : shelf. ——_——____ Product/Month
"Product-Units per month" : shelf_prodof! ————_ Product/Month
"Product-Units per month" : base_prodoff —————_ Product/Month
Fig 20: Response to a product offensive with sufficient resources provided and less sensitivity to the
number of innovations
CONCLUSIONS, MODEL INSIGHTS AND FURTHER RESEARCH
The developed model represents a combination of two different pipeline concepts. The first
incorporates the characteristics of a funnel, which filters numerous inputs (ideas) through
deeper investigation and development activates. A substantial percentage of the creative,
but not feasible ideas are sorted out on the way. The second pipeline depends on a high
degree of operational security and a smaller percentage of non-usable output is tolerated.
Whereas the first pipeline needs to cope with very high degree of uncertainty, the latter is
based on operational security and managed towards indicated deadlines. In both pipelines
the required resources per project and stage rise enormously throughout the subsequent
stages. Where this is (to some extent) balanced by the decreasing number of projects in the
innovation pipeline, the overall required resources clearly mirror this behavior in the
product pipeline. These effects are regularly underestimated in practice. Furthermore (as
the first stages of the pipeline are less resource intensive) these effects become crucial with
a considerable time delay which again is regularly underestimated in practice. Once the
resource intensive stages are reached it becomes extremely difficult to cancel and/or delay
product projects in practice. The developed model contributes to the development of
effective polices in the question of how to plan and manage product offensives. It takes
into account the effects of product initiatives, the resulting resource requirements in
the innovation pipeline and the consequences in the innovational content of a product.
23/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
Another factor that needs to be considered is the inertia of resource shifts and changes.
Whereas in the current model changes in resource allocation are assumed to be immediate
and smooth, in reality the shift and allocation of resources (particularly human resources
such as experienced product designers and engineers) involves (next to costs) considerable
time delays. The same applies to the reduction of resources which is (particularly in case of
human resources) an extremely sensitive undertaking. Fluctuations in resource
requirements should be minimized whenever innovation pipelines have to be modified
in throughput, structure or organization. The developed model can provide a
significant support in the investigation of resource effects and the development of
innovative polices.
The comparison of the two alternative pipeline structures in the steady state run indicates
different model dynamics through variations in resource allocation and risk distribution.
The Shelf-System structure is generally observed to provide advantages in response
time, flexibility and risk reduction within the innovation pipeline. This advantage is
achieved by shifting resources to an additional innovation-based maturation stage. In the
model, these effects are taken in account through the allocation of resources to the
additional stage and the alternation of filter mechanisms. However, advantages gained
through flexibility are believed to be counteracted by a decrease in product
homogeneity. This is due to modularisation and standardisation efforts which are needed to
assure the integration of innovative modules at later stages in the product pipeline. Yet,
these effects have not been implemented in the model and are subject for further research.
To broaden the empirical foundation is also believed to lead to additional leverage points
for sustainable policy development.
It can be summarized that the developed model was able to explicate major elements of the
complexity of automotive innovation pipelines. It is believed that the result can be
transfered to other industries with similar character. Innovation pipelines in established
industries need to continuously generate successful product innovations. These product
innovations represent a combination (“system product”) of proven (traditional) concepts
and new, innovative concepts. By taking an integrated, dual pipeline approach, the
developed model was able to incorporate these effects. Furthermore, through the integration
of the dominating feedbacks it was possible to take into account system inherent limits to
resources and resource distribution. In combination with the delays of the individual
pipeline stages the model explicates the system inherent “inertness” within such innovation
pipelines. Also the model confirms and reproduces effects suggested by Braun (1994),
Repenning (2001) and Le Corre/Mischke (2004) successfully. Nevertheless, the authors
recommend confirming the developed outcomes through further analysis and empirical
data. Also, additional effects which have not been implemented yet need to be considered.
Particular issues worthy of further research are:
= Limiting effects to the dominating loops through market reaction and saturation.
The governing positive feedback loop might be weakened by a counteracting
negative feedback loop.
24/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
= The effect of alternative options in resource provision for short term initiatives has
not been considered in detail. This option might be of interest in order to increase
pipeline throughput in product innovation through a worse-before-better strategy
and smoothen out the reinforcing behavior.
= Competition within the market has not been considered within the presented model.
The increased development efforts of one company could induce reactions of
competitors, causing the character (and number) of innovations a customer expects
in a given product to rise. This effect would weaken the dominating loop.
= Also certain decreases in product homogeneity due to module based innovations
might be observed in the final “system-product”. These effects, in addition to the
efforts needed for modularization and standardization, are worthy of further
investigation.
25/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
REFERENCES
Abdel-Hamid. 1988. The economics of software quality assurance: A simulation-based case
study. MIS Quarterly 12(3): 395-411.
Bass. 1969. A New Product Growth Model for Consumer Durables. Management Science
15: 215-227.
Black, Repenning. 2001. Why firefighting is never enough: preserving high-quality product
development. SD Review 17(1): 33-62.
Braun. 1994. The Innovation War. Prentice Hall: PTR New York
Cooper. 1980. Naval ship production: a claim settled and a framework built. Interfaces
10(6): 20-36.
Ford. 1995. The dynamics of project management. MIT: 27.
Ford, Sterman. 1998. Dynamic modelling of product development processes. SD Review
14(1); 31-68.
Ford D, Sterman J. Overcoming the 90% Syndrome: Iteration Management in Concurrent
Development Projects. Concurrent Engineering Research and Applications 111 (3):
177-186.
Hilmola, Helo, Ojala. 2003. The value of product development lead time in software
startup. SD Review 19(1): 75-82.
Jost, Sauer. 2004. Business Syndromes. Proceedings GWS-Conference 2004: Liineburg.
Le Corre, Mischke. 2004. The Innovation Game: A New Approach to Innovation
Management and R&D. Springer: New York.
Lyneis, Cooper, Els. 2001. Strategic management of complex projects: a case study using
system dynamics. SD Review 17(3): 237-260.
Maier: 1998. New product diffusion models in innovation management- a system dynamics
perspective. SD Review 14(4): 285-305.
Milling. 1996. Modelling Innovation Processes for Decision Support and Management
Simulation. SD Review 12(3): 211-234.
Milling. 2002. Understanding and managing innovation processes. SD Review 18(1): 73-86.
26/27
A. Jost, T. Lorenz, G. Mischke (2005) Modeling the Innovation Pipeline
Myers, Marquis. 1969. Successful industrial innovations: A Study of Factors Underlying
Innovation in Selected Firms. National Science Foundation NSF: Washington DC. 69-
17.
Powell, Schwaninger, Trimble. 2001. Measurement and control of business processes. SD
Review 17(1): 63-91.
Repenning. 2000. A dynamic model of resource allocation in multi-project research and
development systems. SD Review 16(3): 173-212.
Roberts. 1964. The Dynamics of Research and Development. New York.
Stern, Jaberg, 2003. Erfolgreiches Innovationsmanagement: Erfolgsfaktoren -
Grundmuster - Fallbeispiele. Wiesbaden. Gabler.
Zahn, Dillerup, Schmid. 1998. Strategic evaluation of flexible assembly systems on the
basis of hard and soft decision criteria. SD Review 14(4): 263-284.
27/27
CC BY-NC-SA 4.0