HOME | ARTICLES | RISK STRATEGY ANALYSIS| LEGAL NOTICES
© 1998 - 2007 Sensors Research Consulting, Inc. All rights reserved.
The risks of failure of individual research and development
projects for developing innovative technological products are
always significant and can be very high when truly new
technology is involved. A recently reported survey indicates that
it now takes an average of 6.6 ideas to produce one successful
new product, while on average 50-60% of new product development
projects fail. Industries which rely heavily on internal
R&D for new products need to be able to take those risks without
compromising the profitability of the company. Using risk strategy
analysis, RSA, corporate R&D project portfolios can be designed
and analysed in a way which directly accounts for those risks
in terms of an overall risk. RSA also provides a means for developing
customised risk based project selection criteria which can then
support that portfolio design. Risk strategy analysis is particularly
applicable to situations where higher degrees of innovation and
higher degrees of growth are top ranked strategic objectives.
Industries which rely heavily on internal development for new products, particularly where those products are technology or science based, face a difficult challenge when planning R&D project portfolios for new products. Due to the uncertain nature of innovation, each R&D project undertaken with the objective of eventually developing a new product has a degree of risk of failure which is significant and potentially very high. Projects aimed at providing strategically important brand new products to replace obsolete products typically have a long lead time and a higher degree of individual uncertainty in eventual success than other projects aimed at producing enhancements to or the maintenance of existing products, product lines and technologies. For many innovation based industries, the success of research and development activities which lead to new products is key to long term profitability and the risks involved need to be controlled. These industries need tools and methodologies to assist them with managing potentially long lead times for product introduction and the significant risks of failure for individual projects. This paper considers those risks in terms of a risk strategy and presents an analytical technique for managing the risk of failure of projects and designing and maintaining R&D project portfolios, which is particularly applicable to industries which develop innovative technological products. The approach is referred to here as risk strategy analysis, or RSA.
Many people would accept that the greater the risks
taken, the greater the potential rewards. This is after all the
key expectation governing choices of financial investment. When
this axiom is applied to developing new and innovative products,
breakthrough products typically have higher levels of profit and
greater initial risks compared with minor product improvements.
Regardless of the level of innovation and the degree of efficiency
with which projects are conducted new product development projects
remain inherently risky. In reviewing the findings of the 1995
PDMA survey on new product development (NPD) practices, which
incorporated data from 383 respondents, Griffin [11] reported
that on average it now takes about 6.6 ideas to generate one successful
new product, implying an average risk of failure for each idea
of 5 in 6.6, or 76%. According to that report, this figure drops
to 50-60% for ideas which enter development, and as a further
subset, drops again to 1 failure out of every 3 projects for "best
practice firms", [ibid., figure 13]. The latter risk of failure
figure for new product development projects entering development
may thus represent a reasonable lower limit estimate of currently
unavoidable risk for the average project undertaken and it is
still a significant number. Moreover, the 1 in 3 failures, 'best
practice' figure may be considered low compared with that of projects
with highly innovative objectives such as may be aimed at developing
breakthrough technological products. These projects can have risks
of 90% or more, yet the survey results indicate average risks
which are much less. One simple reason for this is that these
types of project do not tend to dominate project portfolios because
projects which enter a product development process typically do
so via a risk-value selection procedure which biases against high
risks. In some companies the selection procedures can be very
complex and take into account many different strategic considerations.
Cooper, Edgett and Kleinschmidt, in their recent book on project
portfolio management [7], provide a very comprehensive account
of the many considerations involved in attempting to design a
balanced portfolio and in so doing also review risk-value project
selection practices at a number of firms. Many of the procedures
described are complex and some involve scoring schemes which also
account for numerically estimated project risks of failure. The
overall objective is to maximise value while minimising risk.
Not surprisingly, risk-value product development project selection
procedures which seek to minimise risk for a certain specified
'value' (which may have multiple characteristics as well as financial
return) bias against higher risks, which in the case of highly
innovative technological projects manifests in a high technical
risk as well as the commercial risk as well as the
uncertainty in the value of the risk. Another option is
of course to maximise value for a certain level of risk. Clearly,
in those industries where innovation and growth are major strategic
goals, project selection must somehow first select on the latter
basis without unduly compromising other important objectives.
This implies a need for special consideration in project selection
for the role of risk and the different levels of risk in innovation
and the corresponding need for a risk strategy in the host R&D
project portfolio which can accommodate those risks.
Unfortunately, formal product development procedures
can also bias against the inclusion of more highly innovative
projects in a portfolio. For instance, according to Crawford [8,
p.189] there is a trend associated with the fast or first
to market approach to new product development, as for example
Rosenau [15], which is a reduction in the degree of innovativeness
of new products due to internal competition with the attractive
short lead times associated with low risk, less innovative projects.
This particular trend implies that speed to market and high degrees
of innovation in new products are incompatible objectives. Additionally,
the effectiveness of formal new product development processes
in general when applied to highly innovative projects is
unclear.
Risk of failure is risk of failure for any reason.
In technological, innovative industries, new products arise from
a combination of innovative technological as well as marketing
concepts. In these cases there are scientific or engineering problems
to solve resulting in a technological risk as well as the commercial
risk of failure in the development projects which create them.
Risk is defined here as the risk of failure to achieve success
as described by the technological specifications which enable
the product and the profit generating objectives which are established
at the outset. Accordingly:
Risk of failure = 1 - Probability of success
(1)
In this context, risk manifests itself as a growing mortality of ideas and projects as each project proceeds. Figure 1 depicts the main stages of internal research and development for new technological products and illustrates how risk manifests as a growing mortality of ideas and projects from concept generation through to R&D, commercialization and success.
Referring to this figure the numbers of survivors
at the different stage boundaries may be expressed in terms of
probabilities of technical and commercial success. The actual
definition of commercial success is not necessarily simple or
obvious. For example the research conducted by Griffin and Page
[10] indicates that success measures vary according to project
category and business strategy. Nevertheless, one result of that
research indicated that companies with an innovative strategy
should focus on the ability of a product development program to
provide company growth. In this sense the detailed definition
of risk also reflects business strategy.
Major reasons for project failure are mistakes, unpredicted
changes and the unforeseen impossibility of the objective. There
are four functional alternatives to risk management:
The NPD process improvement approach to risk reduction basically attempts to reduce avoidable mistakes due to basic inefficiencies, uncoordinated efforts or poorly timed changes in the objectives. NPD process approaches focus on attempting to improve the structure and organisation of the new product development process, largely through teamwork arrangements. Currently, two of the more successful processes described and reported are the stage-gate process due to Cooper [5, 6], and quality function deployment, (QFD), [9]. These are integrated, as opposed to linear (over-the-wall) processes [1], and utilise cross-functional teams. Risk is never defined or directly addressed in the NPD process approaches, except to define a structure within which project activities, which will include risk assessment in some form or another, can be performed. On the other hand, highly innovative technological products are based on original thinking and creativity rather than obvious next step improvements and are especially risky due to fundamental uncertainty in the possibility of attaining the objective. These types of project are strong candidates for attempted risk reduction by employing any one of a series of problem solving or creativity techniques, [3, 12], and product design methodologies, [18]. Such projects nevertheless have a high degree of unavoidable risk and are popular candidates for risk reduction by not doing them at all or hiding them from consideration. The risk hiding approach entails 'hiding' high risk projects from corporate visibility, or at least close scrutiny, by doing them in a basic R&D facility or even a skunk works, which has special budgets and special protection. One problem with this approach is the continuing vulnerability to internal cost reduction. This is because basic or higher risk R&D projects tend not to have a clear or steady link to production. Hence, if R&D budgets are later isolated and expected to provide evidence of returns according to the same criteria used for other departments, it may be 'discovered' that R&D is unproductive and expensive. The last option for managing risk is risk acceptance. In order to take this approach managers need either a willingness to assess, take personal responsibility for and defend the risk (gamble), an appropriate, formally derived risk based project selection criteria which reflects corporate strategy or statistical tools and statistical data to feed into them with which they can plan and at least share the responsibility. In either case the overall risk is likely to be reduced by diversification or spreading of the risk. The gambling approach is used extensively in high risk research, where the elected gambler (R&D manager) assesses projects and is willing to undertake sufficient projects until success is achieved. Concerning the second alternative, as noted by Cooper [7, p.32], there is a concern that formal financially based project selection practices tend to bias against high risk, highly innovative projects, while other risk-value based project selection criteria do not apply to decisions concerning resource allocation according to project types such as high risk breakthrough, low risk product maintenance, etc. The latter, statistical, approach is the dominant approach used to manage financial investment portfolios, where risk is defined as the variance of the return on an investment. Unfortunately, although aspects of and certain modifications of financial portfolio theory are considered applicable to a consideration of product portfolios [4, 14], there are some problems applying it in this context. For instance, the process which leads to technological innovation entails more than commercial development to the extent that projects can fail before they ever have a chance to reach the market. As a result there are certain minimum numbers of projects (levels of investment) required versus different levels of risk when attempting projects in order for there to be any reasonable return. In this respect investment opportunities are not infinitely divisible as normally assumed in financial portfolio theory. Moreover the theory cannot help a manager (investor) to determine what those numbers (investments) should be. Because of this, it is not a generally applicable option for R&D project portfolio management for innovation based manufacturing industries. Currently, there is no statistical approach generally available for analysing R&D project portfolio risks which include high degrees of risk of failure (high degrees of innovation). As a result, high risk R&D is a particularly stressful endeavor.
Risk strategy analysis focuses on risk as a key factor in the development of new technological products. A risk strategy, in this context, describes a specific extent and type of involvement in research in terms of the different levels of technical and commercial risk associated with the different types and relative numbers of projects making up the portfolio referred to here as the project mix. Accordingly, a risk strategy is analyzed in terms of statistical risk of failure as opposed to probable outcome. As a result of this approach the minimum likely overall success of a strategy for research by innovation can be determined at a preassigned overall level of risk of failure to achieve that minimum success. Given a number of basic project descriptors, RSA can then be used to design as well as analyze risk strategies Risk strategy analysis is thus a potential tool, which allows one to design a portfolio project mix with particular strategic business objectives and in which the overall risk of failure is controlled or can be estimated. Similarly, it provides a risk based screening technique for individual projects needed for the design which can be used in series with other screens. The technique is computationally, relatively simple. The terminology involves common language concepts such as worst case scenarios.
Research and development projects support a range
of needs, and designing and maintaining an appropriate portfolio
comprising a mixture of projects with different basic objectives,
risks and costs requires consideration of the individual and overall
risks involved versus the short and long term product development
and support requirements and profit needs. For example, an established
company which developed new products via internal research some
years ago, and consequently has patent rights to specific intellectual
property upon which several current high profit key products are
based, knows that these rights will expire in so many years time.
Upon expiration of these rights, profits from those products is
confidently expected to decline drastically as competing products
from other industries emerge to take a share of the market, and
as a result new products will be required at that time to take
their place, (in the US, patent holders are currently granted
exclusive rights to patent claims for a period of 20 years from
the date of filing of the application. Prior to mid 1995 the period
was 17 years from the date of issue.) The question then arises
as to how much high risk research needs to be conducted now in
order to have a reasonable chance of developing replacement high
profit products and thus avoid drastic loss of profit in the long
term, while still maintaining other lower risk projects in support
of the existing products, thereby also satisfying short term profit
needs. In general, an appropriate balance of projects and risk,
has to be identified. Where R&D portfolios include high risk
projects, such as those aimed at providing breakthrough products,
it is particularly important to adopt a strategy whereby the overall
risk of the portfolio of projects is managed to acceptable levels.
Conversely, where R&D portfolios are too conservative overall,
adjustments to the risk strategy can potentially increase the
overall profitability and enhance a company's strategic potential
for innovation.
The analysis applies to science and engineering research
and development project portfolios which are part of a new product
development process involving technological as well as market
innovation. Furthermore it applies specifically to those new products,
such as machines, devices and systems, which are created predominantly
via technical innovation, which is defined here as that
process which entails the invention and/or application of specific
new or modified scientific or engineering concepts. This is as
opposed to discovery-based research activities which, for
example, dominate in the development of new pharmaceuticals. Discovery
involves a search and a trial-and-error test or screening process,
whereas innovation involves a concept, proof of feasibility, a
product design and an evaluation. Hence in this terminology, research
by innovation is different from research by discovery, which is
conducted differently and obeys different statistics. Nevertheless,
the majority of innovation based manufacturing industries generate
products obtained predominantly via innovative R&D processes.
One example, with which the author is particularly familiar, is
the development of a new sensor. In this example, the specification
could be for a sensing device to monitor a parameter such as temperature,
pressure, flowrate of a fluid or chemical composition to an accuracy
and at a unit cost, not obtainable with current commercial devices.
In order to accomplish this project a new or modified measurement
principle may need to be identified or conceived. The R&D
project must then prove the feasibility of this concept via analysis,
experimentation, the construction of a functional model and technical
performance testing. Subsequently an engineering prototype must
be designed to explore and meet format, cost and manufacturing
specifications. Hence, in this project, as in innovation based
research and development projects in general, there will be a
series of project milestones which will be reached only after
the accumulated solution of a series of technical problems, some
of which will be unique and some of which may be unforeseen. In
this respect the project can be considered a continuously creative
or innovative process, where the accumulated uncertainty in the
existence of practical solutions to the many component problems
encountered gives rise to the overall project technical risk.
The fundamental characteristic of RSA is the upfront recognition that risk, degree of innovation and potential profits are all connected and that although the actual levels of risk associated with different development activities in different industries and firms will vary, projects can nevertheless be generally categorised according to descriptive risk types which reflect the innovative objectives. The actual risk levels associated with those categories can then be numerically identified according to different situations and this step can then introduce a new degree of risk predictability to a portfolio. An R&D portfolio can thus be designed in terms of a risk strategy, which is then also directly related to a company's objectives in terms of desired degree of product innovativeness, desired growth and overall risk tolerance.
A number of different product and project classification
schemes have been described in the literature. One of the more
frequently referenced schemes is due to Booz, Allen and Hamilton
[2] and describes new products in terms of newness to the world
and newness to the company. This scheme of classification has
very broad application. For this analysis, however, one is only
concerned with the key categorizations which parallel risk and
relate to innovation for a specific industry, within which certain
boxes in 3 X 3 'newness to the firm' versus 'newness to the world'
matrix may not be filled or may be duplicated. From the viewpoint
of the project team scientist or engineer charged with conceiving
and conducting an R&D project and trying to assess potential
project risks, the key dimension for categorizing a project is
the degree of technical difficulty inherent in the technical objectives.
For the market analyst the degree of uncertainty in the success
of commercialization is likely to increase with more innovative
product concepts. Although the apriori evaluation of risk
cannot be exact, most experienced professionals will agree on
relative risk and the greatest risks are generally associated
with the most innovative or creative objectives. Generally, within
any innovation based industry, one can identify three broad objectives.
The three objectives, as listed below, also typically correspond
to decreasing levels of research risk and financial reward.
The level of risk associated with any single attempt to solve an issue has to be supportable by the potential rewards. For this reason, only the very largest industries would consider engaging significantly in unrestricted very high risk, sometimes called 'blue sky', research where the objectives and rewards are undefined and hence which constitute an appreciable additional risk of commercial failure. This latter category of research, which is omitted in the above list, often has a different objective and as such is conducted more intensively by institutions where long term knowledge growth is it's own reward. For industrial situations, I have therefore classified research project risk levels into three nominal categories, namely high, medium and low. These correspond to the listed objectives. Also, in each category I give example values of the probabilities of technical and commercial success where the example value of the commercial risk, (nominal risk of failure to commercially implement successfully), has also been equated to the technical risk. Clearly, situations will arise where technical and commercial risks are different from each other and may be specifically distributed or clustered into categories differently. This would not invalidate the approach. For example, a high risk research project may be undertaken to develop a new technology destined solely for use within the parent company where it will be used to enable production of something else. In this case the technical risks are high whereas the risks inherent in the commercial implementation may be lower because of reduced market uncertainties and instead fall into the medium or low risk category. The probability values selected here as example or nominal values are for the purpose of illustration and are representative of my own judgment and experience in industrial sensors research. The values selected are very different for the different categories. The basis for the chosen values is outlined in the following. For the purposes of this analysis a project is defined as any concept which enters into the R&D stage with the hope of commercialisation. It thus does not include the preliminary stages of concept generation, concept screening and business analysis, which are considered here as preproject activities. These preliminary stages tend to have a relatively low or hidden cost. In particular, basic concept generation is an especially creative task and regardless of formal creativity processes is likely to also occur in unstructured situations. Within a company, the officially recognised portfolio of new product ideas and concepts is thus unlikely to be a true representation of the total number of concepts generated since many will be discounted early and off the record. Hence, risk includes the risks associated with solving and proving scientific and engineering solutions to a product concept and the risks of successfully implementing that solution commercially.
High risk research is defined here as the type which is based on a new concept which, if it can be made to work, will result in a new technology or technical breakthrough, one or more patents and a very high level of potential commercial profit. Typically a top scientist or engineer conducting high risk research with adequate resources will always generate good quality results. Over the course of time most of the high risk projects a good scientist or engineer is given will also have a high degree of purely technical success. However, for the sake of generating example nominal probability values, let us consider a possibly familiar scenario where three out of four times these technical solutions will miss the mark or go only part way and hit a roadblock and the research will be abandoned or shelved. The company wanted A and the project delivered B instead. It's a solution without a problem, even though in foresight everything was done correctly. This is nevertheless very good technical performance and indicates a top level of competence. On this basis, the example nominal probability of a technically successful solution for a well conducted high risk research project is 1 in 4. Given a technically successful solution, the next set of problems involve developing, manufacturing and selling it. However, this particular project is a new technology with brand new capabilities. To introduce it successfully there has to be a carefully orchestrated series of changes not only in how to conduct business but also in the marketplace, even though all the market studies indicate good acceptance. Here again there is substantial risk. Accordingly, the example nominal probability of successfully commercializing a technically successful, high risk project is set at 1 in 4. This makes the overall example probability that a high risk research project will be commercially successful 1 in 16.
Medium risk projects are those which do not require inventing a whole new technology. They may comprise substantial improvements to an existing, maturing technology and product, such as a Mark II or III device with real performance enhancements or added features, and may also have the benefit of a lot of previous technical and commercial experience in that general area. Patents may still be generated but may be contingent on a patent history involving the core technology. The commercial life will be less than for a brand new technology because of competition or the nature of the maturing business, but still significant. Here, as an example the assigned technical risk is 1 in 2, (a 0.5 probability of technical success), and the commercial risk of a successful implementation, which may be a product replacement or addition to an existing line, is estimated to be the same, making the overall probability of commercial success 1 in 4.
Low risk research projects are predominantly developmental in nature with perhaps some basic research component to them if necessary. They invariably succeed in meeting the project objectives although once in a while there is a problem and they are abandoned. They are likely to be conducted in close support of a current product or business. A particular customer may need a special type of an existing product for a limited time or application or a particular product will sell better if it has one simple refinement, or small improvements or fixes may be required. Either way good scientists or engineers are still involved because the project tasks are still highly technical. However, no significant technical breakthroughs are required or expected and the whole project may be expected to conclude in a matter of a few months. Here, the example technical and commercial probabilities of success are 0.9, making the overall probability of commercial success 0.81.
Table 1 describes the input parameters required for this analysis along with the example values which are used later to illustrate the analysis.
Table 1: The R&D project category input parameters giving the variable names and example values referred to in the text to illustrate the analysis.
Assembling and refining this data is the first step
towards portfolio risk strategy design and analysis and strategies
for this step are discussed in the next main section, whereas
this section provides definitions. Referring to the table one
can see that each project category is described by five parameters,
namely; probability of technical success, probability of commercial
success (probability of a commercially successful implementation
of the technical success), time to conduct the project, estimated
commercial lifetime and estimated profit potential per year. Additionally,
one needs to specify the average cost per year of conducting a
research project. If desired, one can specify additional project
categories in order to describe projects undertaken in different
divisions or strategic business units of a large corporation,
where for instance the estimated parameter values for high, medium
and low risk projects could be significantly different one from
the other, or to describe special situations where for instance
technical and commercial risk categories are mixed. However, in
the former case one would preferably perform separate analyses
for each company division.
The project parameters thus described are averages. In the case of expected profit potential this reflects an assumption that profits remains flat throughout the commercial life. This is not usually the case and, depending on the product, it may be possible to predict profit profiles with some degree of confidence. Although an estimated profile per project category could be incorporated in the profit projection part of the analysis this has not been done in the example analyses here.
The example values assigned to the probabilities
of technical, PTSx, and commercial, PCSx, success
for high, medium and low risk research projects are 0.25, 0.5
and 0.9, respectively. The precise values selected are based on
the author's judgment and experience in a specific industry and
research facility. Hence appropriate values for different industries
and facilities may be different. Cooper et al [7, pp 33-36] recently
reviewed a number of techniques used in various industries to
obtain estimates of these numbers for individual projects. The
foremost consideration here is that the probability numbers for
different project risk categories are very different from each
other and the initial task in portfolio design is to then distinguish
projects on the basis of risk category. Without the benefit of
experience it is very difficult to assess research risk, especially
when the risks are high. Moreover, it is logically impossible
to know the actual risk entailed in any single project
whether it is judged to be high risk or not. If it succeeds the
risk was zero. If it fails the risk was 1 or 100%. The actual
risk taken at this point is that one cannot know the actual outcome
ahead of time. Any single low risk project can fail. Any single
high risk project can succeed. This is why individual project
risks can only be subjectively assessed. It is technically
possible however to assign a numerical risk value to a class of
projects and thereby assign risk to other projects which fit into
that class. To do this one needs to have knowledge of the performance
of a number of projects in the same class. Hence if one has specialized
information about very similar research in a similar setting by
which over the course of a large number of projects at that risk
level (say 50 or so for high risk, making sure to include all
projects undertaken even ones that failed early), it turned out
that 1 in x number was a purely technical success and 1
in y multiplied by x was also a commercial success,
then one should assign the values of 1/x and 1/y
for the technical and commercial probabilities of success to that
risk category. Also, as discussed earlier, there can be specific
cases where the categories of technical risk and commercial risk
are mixed giving rise to other classes of project such as high
risk technical, and medium or low risk commercial.
Assessments of the probability of success of individual projects are based on judgment. If one does not have the benefit of historical data or sufficient data upon which to base an objective evaluation of probabilities of success for project categories then one also has to employ judgment in selecting those probability values by either deciding that the nominal values described here are appropriate or by formally surveying experienced individuals with questions such as "What is the likelihood of technical success for a typical project aimed at providing a breakthrough product for this industry ?" According to the results of operations research studies in this area, for example [13], appropriate measures for probability judgment questions need not be finer than a seven point scale. In most cases for high risk categories an appropriate scale is a choice from 1 in 3, 1in 4, and so on up to 1in 9. For the medium and low risk categories the linear scales 0.4, 0.5, 0.6, 0.7 and 0.8, 0.85, 0.9, 0.95, respectively, are appropriate.
This is similar to the cycle time, which is standard terminology for the time it takes to bring a new product to market The definition for L_x is that it is the period of time over which most of the project development and commercialization resources are tied up in a project averaged over all projects in a category whether they succeed or not. Typically, R&D projects do not run for a set period of time with full resources. Assuming projects are managed and not just allowed to drift, then some will be abandoned due to failure to achieve a milestone or pass a stage. Also, a project rarely starts from day one with a full team. R&D project costs and manpower tend to grow as confidence grows and project tasks become more detailed and then decline as the project nears completion. A successful project will tend to require some degree of R&D team involvement well past the official conclusion of the R&D phase in order to support technology transfer through the engineering, design and manufacturing stages and even into product launch. Hence, a particular, successful high risk project might entail 7 years of significant support, while projects which are later unsuccessful may run for two, three or even four years. L_x is thus the average or estimated period of time over which R&D and commercialization resources are applied for all projects in category x whether they succeed or not.
The commercial lifetime is defined here in terms of profit generation performance. It is the estimated average length of time for that project category between corresponding product introduction and the time at which profits from that class of product are expected to decline substantially or the time at which that product needs to be replaced by a new product, whether by planned obsolescence or otherwise, [16]. For example, in the case of a highly innovative, technological product which resulted from a successful internal, high risk R&D project, the high profit commercial life may end when the key patent expires, thus allowing competitors to introduce similar products onto the market. The commercial life in this case might then be close to 20 years. Typical values for this parameter for high risk projects may indeed be in the region of 10 to 20 years depending on the maturity of the industry.
This parameter describes the financial objective.
A commercial success is defined here as one which achieves or
exceeds profit expectations. To determine this number, one needs
to estimate the average annual profit expectations for typical
projects within a project risk category over the course of the
average expected lifetime on the basis of industry specific experience
and current market information and forecasts. It is calculated
as the average expected revenue minus all the costs excluding
the NPD costs, divided by the expected commercial lifetime:
(2)
Estimation errors for Profit_x derived from randomly distributed errors in the input data are partly accounted for in the probability of commercial success. However the commercial risk parameter does not account for an overall bias or systematic misreading of all markets for all projects. If a company is involved in multiple projects for different markets then taking an average figure for all high risk projects, medium risk projects and low risk projects, will decrease the effect of errors or bias in any single market study. Marketing information is most unreliable where new markets are concerned. These may more frequently be associated with the most highly innovative product concepts and high risk projects and this is one reason, as reflected in the example values, for the example assumption here that the commercial risk for a high risk project is high and the same as the technical risk.
The example value assigned to the average cost per year of conducting an R&D project is 1 MM per project, per year, ( MM is used throughout this paper as shorthand for 1 million US dollars). This might be equivalent to a very well resourced sensors project with a dedicated team of about 5 professionals and a team leader plus overheads. The assumption here is that projects are adequately resourced. Attempting a research project with inadequate or compromised resources can be, and is frequently, done to some extent, but it is a false economy if it increases the risk of failure or excessively impacts the time it takes to complete a project. The latter relationship is discussed in more detail in the section concerning management implications.
Risk is defined as the likelihood of failure. If, for example, the probability of success of a typical project is 1 in 10, then the risk of failure of any single project is 90%. If two such projects are conducted, then the risk of failing to have at least one success is the probability of two failures, which is 81% and the risk of failing to achieve no more than one success is 90%. For ten projects the risk of failing to achieve no more than one success is 0.99 = 39% and for 22 projects it is 10.9%. In a portfolio of multiple projects, the objective is to create new products at a rate and with a degree of reliability which reflects the business objective. The overall risk of not achieving those objectives is a function of the portfolio design and business strategy and can be controlled or estimated by portfolio design and analysis. In the analysis, overall risk is specifically defined in relation to a particular portfolio performance outcome as the risk of failing to achieve no more than that performance. Where the cost of failure is high, the overall risk level should preferably be low. It is an input parameter in the design process. For the design examples the example 'acceptable' overall risk for high priority objectives is set to 10%.
This concerns maximum lead times versus risk and
relates to generation or maintenance of product streams. Tc_x
is defined as the time immediately after which there is c level
of confidence or probability that a research strategy will have
given rise to at least one commercially successful project in
a particular category x. Clearly, it is a different number
for different levels of confidence or overall risk. Equation 3
states how to calculate this time.
(3)
where x denotes project category, L_x
is the average time it takes in years to conduct a project in
category x, c is the level of confidence that there
will be at least one success, N_x is the number of concurrent
projects in category x, PTSx is the estimated probability
of technical success for a project in category x and PCSx
is the probability of a commercially successful implementation
of a technically successful project in category x.
Table 2 gives some example values for the T90 (c=90%) and T50 times for high and medium risk projects versus different numbers of concurrent projects where the values were calculated using equation 2 and the example project parameter values described in table 1.
Table 2: Example Tc times for high and medium risk projects calculated using equation 2 and the example project parameter values in table 1, where T90 is the time after which there is 90% confidence that a research strategy will have given rise to at least one commercially successful project and T50 is the time after which there is 50% confidence. (Where appropriate the following adjustments were made. * denotes values corresponding to the time at which there is c confidence of at least one success and is equal to Tc + L_x/N_x ; denotes values which have been rounded up to allow for conclusion of a project.)
To help explain the significance of these numbers
let us refer to the table and consider two hypothetical situations.
In the first situation consider a project portfolio which includes
10 concurrent high risk projects at a total yearly cost of 10
MM, where the probabilities of technical and commercial success
are each estimated to be 0.25 (see table 1). According to the
table values, there is then a 90% probability that there will
be at least one success after about 18 years. The rounded up value
of the commercial lifetime divided by the T90 time yields
a 90% confident forecast of the long term minimum mean peak number
of concurrent profit generating products. Hence, if in this company
the overall risk tolerance for investment in R&D is 10% and
the objective is to avoid future dramatic profit loss where current
profits include only one high reward product, with a commercial
lifetime of 20 years, and which is currently yielding 50 MM profit
pa, then one would plan for this number to be at least 2 and accordingly
need to conduct no less than 10 concurrent high risk projects.
As one can see by examining column three of table 2, if instead
the research portfolio comprised only 3 high risk projects then
there would be a 50% risk of failing to at least maintain the
product stream. In the second situation, let us consider a company
which currently has 10 products on the market, each with a useful
commercial lifetime of about 5 years and which developed from
medium risk projects having estimated parameter values as described
in table 1. These products thus generate about 30 MM total profit
pa. In this case the required average rate of new product introduction
to maintain the business is about 1 every 6 months. Referring
to table 2, one can see that in order to generate at least
the required number of successful new products with 90% confidence
the portfolio must include 18 concurrent medium risk new product
development projects, at an annual cost of 18 MM, (compared with
7 concurrent projects at the 50% risk level and a cost of 7 MM).
In a new business, the Tc times for the different
project categories afford an estimate of the maximum likely startup
time at a 1-c risk level.
To review then, the Tc time for a category is the time after which there is a 1- c risk of failure of all projects in that category which are conducted during that time. This means that there is a c probability of at least one success immediately after that time, where at least one includes all cases of 1, 2, etc., successes up to the case of all projects succeeding. Risk of failure is thus singularly defined whereas success is not. If one or more projects succeed, which ones they will be and when will the successes occur are not predictable. When a portfolio of concurrent projects is continuously maintained, overall success rates will eventually trend towards the average probability of success values. However, this may take many years to establish, especially for the higher risk project categories and is not therefore a practical basis for evaluating portfolio performance in terms of risk.
PPV allows one to estimate the value of typical projects
from different project categories and also can be used as a preset
value (reflecting business strategy) to determine a project profit
potential versus risk screening criterion for project selection.
Equation 4 defines the parameter PPV_x, which is the probable
project value per project year for those projects in category
x.
(4)
where Profit_x is the estimated average profit
generating potential per year for a product from a category x
project and CL_x is the expected commercial lifetime
in years. This is different from net present value (NPV),
which is sometimes used to evaluate projects, (NPV is essentially
the total expected revenue from a project minus the total costs
and accounting for an increase in the cost of capital over time.
It might be adjusted for project risk by multiplying the expected
revenue by some factor less than 1). The advantage here of PPV
is that it easily allows one to evaluate the relative value of
an R&D project category and indeed of an individual R&D
project by comparing it with the yearly cost of conducting
that project, whereas NPV provides an evaluation of a potential
product and hides the cost of conducting the project in
total costs. Accordingly, in order for category x projects
to be profitable, PPV_x must exceed the yearly cost of
conducting a research project. Additionally, the higher risk categories
should also have appropriately higher PPV values. If they are
lower than the next lower category then, on a strictly profit
basis, there is no point in doing them since lead times are less
for the lower risk projects. Using the example values for the
input parameters from table 1, one in fact finds that the probable
project values calculated using equation 4 are 12.5 MM pa for
the high risk category, 3.8 MM pa for medium risk and 2.3 MM pa
for low risk. These example numbers are respectively greater than
each other and the 1 MM yearly cost of conducting a project.
An extension to PPV is the risk adjusted value of
a project in category x, PV_xrisk. Here
risk adjusted refers to an adjustment to the length of time required
to produce profits according to the reduced success rate associated
with the preassigned risk level for failing to succeed at a higher
rate, and does not have the customary financial meaning relating
to adjustments for the cost of capital or the variance of returns.
It provides an estimate for project value at the minimum success
rates associated with risk of failure.
(5)
In order for it to make sense for a company to conduct higher risk projects then earlier, there must be sufficient numbers of them to ensure the needed success rate with the minimum acceptable or assumed risk of failure. In order to design and maintain an R&D portfolio with certain new product success objectives it is thus necessary to understand and specify the different risk categories of new product development project which will comprise the portfolio, develop appropriate corresponding screening criteria for admitting individual projects into the portfolio categories and then proceed to plan the relative levels of R&D project investment in each category.
The first step in the design process is the putting
together and refinement of the specifications for the different
project categories using the project category descriptors described
previously as the analysis input parameters. These specifications
then become the input data for the design of the project portfolio.
During this specification selection process functionally appropriate
project screening criteria are also developed which need to be
used as an extra filter for new project concepts into the portfolio.
The parameters associated with each category and the associated
screening criteria reflect the industry and the business
strategy implicit in a project portfolio which includes projects
selected from within those categories. Since industries and company
strategies vary, so will the procedures for selecting specifications
and screening criteria. The following thus emphasizes the role
of risk in implementing this step.
Data for the project parameters needs first to be ascertained by survey or subjective assessments for typical projects and then organised according to project category. The nominal estimates thus define the typical characteristics of projects associated with developing the associated new products for that industry. This data will then include estimates for the nominal probabilities of success (PTSx and PCSx). Selecting the actual nominal probability values to be used in the portfolio design, as with all the other input parameters, must then reflect an appropriately large subset of that data. For instance, in a particular industry a survey of high risk projects could conclude that they comprise projects with an overall probability of success less than some value such as 0.1, and which thus encompasses a small proportion of projects with extremely low probabilities of success, such as 1 in 500. Hence, an appropriately large subset for the specifications would exclude the blue sky type of project. This is a reflection of business strategy. Similarly, it may well turn out that the typical industrial project undertaken has an overall probability of success more or less evenly spread over the range 0.02 to 0.1 or with one or more peaks in the distribution related to different types of product within that category. However, for a particular company the available investment for that class of project may be insufficient to support projects with a probability of success of only 0.02, but sufficient to support projects with a probability of success of 0.05 or greater. Again an appropriate subset may be to exclude all projects with probabilities of success below, for example, 0.04. However, this may conceivably exclude a peak associated with a certain type of product, which again relates to business strategy. Regardless of how this is approached, the portfolio design input probability parameters must reflect an appropriate and realistic nominal value and an appreciation of the business implications associated with the corresponding project selection screening criteria. Projects which subsequently feed into the portfolio design which is based on the assigned nominal values will need to be screened according to individual assessments of probabilities of success in order to preserve the overall validity of the design and its performance. An appropriate screening probability is some other value lower than the nominal category value above which candidate projects pass but which nevertheless still admits a sufficient flow of candidate concepts into the NPD process. In particular, that 'some other lower value' must appropriately correspond to the capacity to conduct sufficient numbers of projects. A second screen is then eventually calculated to ensure that each project admitted to the portfolio has sufficient profit potential to ensure the needed overall financial returns taking into account the risk strategy of the design. The following describes the basic factors involved in analyzing, selecting and refining the specifications for project categories and screening criteria with particular regard to overall investment levels (total number of projects) and profit expectations once the raw data for the input parameters has been assembled.
An appreciation of investment breakpoints is a key
factor in portfolio risk strategy design. Portfolio risk strategy
design breakpoints occur when the total numbers of projects conducted
in a risk category are only just sufficient to maintain the product
stream at the required overall level of risk or conversely when
the overall probabilities of success are so low that a fixed investment
portfolio is only just capable of supporting a sufficient number
of projects. For example, on the basis of profit and the need
for a stable industry, there are breakpoints in investment levels,
(numbers of projects) at which all higher risk research is best
either abandoned or undertaken wholeheartedly. In-between strategies
which are not simply transient situations enroute to building
a new or more stable strategy, increase the risk of excessive
payoff times and subsequent occasional periods of loss of profit.
Using the example project parameter values listed in table 1,
the T90 time for 8 high risk projects turns out to be 22.5
years. Hence, if the total number of projects in a portfolio is
expected to be 10 or 11, where at least two or three of these
need to be projects in the medium and low risk classes, one may
have to accept a level of overall risk greater than 10% if high
risk projects are to be included. 9 concurrent high risk projects,
for which the T90 time is 20 years, are required in order
to maintain a 10% risk of failure to maintain the high profits
from this class of project for one product. In this example, this
is the breakpoint for involvement in high risk research. The alternative,
in an example portfolio of 11 projects, is therefore to eliminate
the high risk project category from the design process and design
a strategy involving only medium and low risk projects. For instance,
the T90 time for 10 concurrent medium risk projects is
about 10 months which is well within the 5 year commercial life.
There is a second breakpoint for engaging in medium risk research.
Based on the example project parameter values and assuming a risk
tolerance of about 10%, then this breakpoint occurs at the one
to two project level.
Table 3 gives the breakpoints for one product for high and medium risk projects at the 10% and 20% risk of failure levels versus a range of probabilities of success and for two different commercial lifetimes, namely 10 years and 20 years, which is an appropriate example range for very highly innovative, or breakthrough products and 3.5 and 5 years for medium risk.
Table 3: Breakpoints for high and medium risk projects versus probability of success values, level of overall risk and commercial lifetime where PTSx = PCSx, L_high = 5 years and L_med = 1 year. (The shaded areas correspond to the example values for probability of technical success and probability of commercail success given in table 1 and referred to in the text to illustrate the analysis.)
| |||||
These data were calculated assuming an average time
for project completion of 5 years for high risk and 1 year for
medium risk. This table is also discussed later in connection
with probability assignment errors. However, as an example, referring
to the table and in particular the top shaded row, one can see
that the minimum number of concurrent high risk projects needed
to replace one major product with another breakthrough product
every 20 years is 9, at the 10% risk of failure level and where
the probabilities of technical and commercial success are both
0.25, giving rise to an overall probability of success of 1 in
16 or 0.06. Looking down column 3 of table 3, to the next lower
row one can see that the corresponding minimum number rises to
15 when the overall probability of success is instead 1 in 25,
and, if the commercial lifetime were instead 15 years one can
interpolate the data between columns 3 and 5 and deduce that the
minimum number of projects would instead be close to 20 at the
10% risk level. Hence, as an example of how to select project
risk category probability specifications, if the total number
of projects in a portfolio is 20 where a large portion of that
total can be devoted to high risk and the uncertainty in the commercial
lifetime for high risk projects places it in the range 15 to 20
years, then the nominal overall probability of success for that
category would preferably be in the region of no less than 1 in
16 and an appropriate project screening criteria for a nominal
value of 1 in 16 would be no less than the next lower probability
increment, which in these examples is 1 in 25.
The breakpoint numbers for projects where the probabilities of technical and commercial success are instead 1 in 9, are 47 and 33 respectively. This indicates that only the very largest industries should entertain undertaking such aggressive high risk R&D projects for profit. Moreover, in order for very high risk projects (overall probabilities of success in the 1 in 50 to 1 in 100 range) to be financially attractive the corresponding profit potential must be appropriately high and preferably at least several 100 MM per year. For example, the probable project value (PPV, see equation 3) for a high risk project with a probability of technical success equal to the probability of commercial success equal to 1 in 9, an expected project lifetime of 5 years, an expected commercial lifetime of 10 years and a profit potential of 50 MM per year is only 1.2 MM per year, or 20% more than the cost of running the project. On the other hand if, for instance, the estimated profit potential is 500 MM per year the PPV is instead ten times this amount. These considerations introduce the subjects of project and project category profit potentials.
Table 4 provides example probable project values and risk adjusted project values for high and medium risk projects versus the same range of probabilities of technical and commercial success, and commercial lifetimes as described in table 3.
Table 4:
Probable project value (PPV), project value at x% overall
risk, (PVx% risk), required profit potential to achieve
probable breakeven and to achieve breakeven at 10% overall risk,
for high and medium risk projects versus probability of success
and commercial lifetime where PTSx = PCSx, L_high
= 5 years, L_med = 1 year and Cost_pa = 1 MM pa.
(The shaded parts of the table illustrate example boundaries for
the example high and medium risk project categories which are
used to develop project selection criteria for high risk projects
where for high risk the nominal overall probability of success
is 1 in 16 and the commercial lifetime is expected to be between
10 and 20 years and for medium risk the nominal overall probability
of success is 0.25 and the commercial lifetime is expected to
be between 3.5 and 5 years.)
| |||||||||
| |||||||||
|
| ||||||||
|
|
|
| ||||||
Additionally, it provides required profit potentials
per year in order to achieve probable breakeven and breakeven
at preassigned risk of failure levels. The former figures, which
assume average probable success rates and commercial exploitation
of every success, indicate the minimum required estimated profit
potential per year in order that over the long term there is likely
to be a positive return on the total cost of development, which
here incorporates the estimated cost of conducting all projects
within that category at the estimated overall success rate. The
latter figures instead assume a minimum success rate at the stated
level of overall risk.
Corresponding to the nominal probabilities of success
describing the appropriate subset of typical projects in a category
will be the nominal profit expectations. These need to be analyzed
to determine if they fit with strategic objectives. For instance,
in order to justify the longer lead times of higher risk projects
then on a profit basis one may decide that the estimated average
profit potential used to specify the project category needs to
correspond to an average PPV for high risk equal to or greater
than a X Cost_pa, for medium risk b X Cost_pa,
and for low risk c X Cost_pa, where a>b>c
and c >1. If for instance, it turns out that a
< b, or if a is less than some multiplying
factor, then one may decide not to emphasize or perhaps even include
high risk projects in the portfolio. As an example, referring
again to table 4 and considering the requirement that a
> b or some specific multiplier of project cost,
let us assume that the average profit potential for medium
risk is confidently expected to be about 3 MM pa with commercial
lifetimes ranging between 3.5 and 5 years, while the uncertainty
in the commercial lifetimes for the high risk category is 10 to
20 years. The corresponding range of PPV for medium risk is then
2.6 to 3.8 times the cost of conducting a project. Hence b
3.2 and one may then prefer to have a 6. Looking along
the heavily shaded upper row in table 4 one can see that the corresponding
PPV range for the high risk category is 6.3 to 12.5 MM pa, which
is an average of 9.4 times the project cost thus satisfying this
broad objective. For comparison, one can also calculate the required
profit for high risk corresponding to a = 6 by taking the
average of the required profit per year to achieve probable breakeven
over the range of uncertainty in the commercial lifetime and multiplying
this number by 6. Referring again to table 4 and following the
same shaded row one can see that this value is 6 times the average
of 4 and 8 MM pa, which is 36 MM pa. This number can now be compared
with the example nominal profit potential of 50 MM pa. Appropriate
corresponding minimum profit expectations values are then needed
for individual project screening. Here for guidance in selection
of an appropriate number one can refer to the calculated probable
breakeven values and breakeven for that risk level and also at
the limits of the uncertainties of commercial lifetime and probability
of success. This selection must be made according to the priority
assigned to a project category and the degree to which it is expected
that successes will be exploited. If all successes will be exploited,
reflecting a highly aggressive growth strategy, then an appropriate
screen is a number greater than the breakeven profit per year
corresponding to the average probable success rate. Columns 7
and 8 of table 4 give example values for these numbers for the
illustrative example cases. On the other hand, if for example
high risk R&D is pursued with the objective of maintaining
a stream of a fixed number of breakthrough products and the priority
is such that the portfolio is designed to keep the overall risk
of failing to do this to within about 10%, then any breakthrough
successes which emerge from this portfolio in excess of the required
numbers of successes will not be exploited. In the long term,
the net cost of all the projects undertaken will thus be higher
than that associated with the average probable success rate and
the project screening, profit potential for that project category
should then be no less than the breakeven profit corresponding
to the risk adjusted project value. As an example, this number
which in this example corresponds to the extremes of the example
range of uncertainty in probability of success and commercial
lifetime of 1 in 25 and 10 years, respectively, and as highlighted
in table 4, turns out to be 28.5 MM pa.
Clearly, the actual preferred values for a, b
and c as just described, reflect the business strategy
inherent in the portfolio. For instance, in some industries a
breakthrough product may not by itself be expected to give rise
to stellar profit performance, yet it is necessary once in a while
in order to establish a new product line for a subsequent series
of new medium risk products which will provide the core of company
profits. This category of project might then have as its objective
the invention and establishment of a new core technology. In this
case the preference would be b>c and c>1,
while a for project screening may be less than 1 or even
zero.
Taken together, the example data in tables 3 and 4 illustrate the factors which need to be considered when refining the project category specifications and project screening criteria for overall probability of success and minimum profit potential. Risk based project screening is briefly discussed later in relation to inclusion in an NPD process.
Depending on the business strategy, the relative priority of the different project categories may be significantly different. For instance, a company which is highly dependent on the sales of one or two breakthrough products and which also has a series of less profitable products derived from previous medium risk projects, may want to design a project portfolio which has a much lower overall risk of failing to provide the essential one or two new breakthrough products compared with that associated with successfully providing an exactly specified and higher number of products from lower risk project categories. In this example, the project portfolio would therefore include sufficient numbers of high risk projects to satisfy the desired success rate at a specified low overall level of risk, which may be 10%, along with sufficient numbers of medium risk projects to satisfy the preferred success rate at say 20% overall risk or even at the risk level associated with the long term average probable success rate. In this example, the high risk project priority is then expressed as a willingness to conduct and bear the cost of extra projects in order to improve the likelihood of obtaining the required minimum successes. If the actual success rate is higher and the additional successes might not be exploited then the associated value of projects in that category is lower. As previously discussed, table 4 also gives example project values which are adjusted for risk for the high and medium risk projects in a portfolio where the overall risk of failure to provide a new product within the commercial lifetime is 10% and 20%, respectively. Referring to the table one can see that in this example, the project values for high risk have dropped below PPV further than for medium risk. In fact, when designing a portfolio based on prioritizing project categories it is important to be aware of the possibility that if the risk level is, for example, set very low for the high risk category the project values may in fact drop below those of the medium risk category.
As a result of the fact that the component projects
in an R&D portfolio all have certain risks of failure, which
of the projects will actually succeed is unknown. This is not
a problem if an industry is sufficiently flexible to be able to
accommodate any and all successes, which is the optimum situation
for maximum profit potential. However, different projects typically
have different priorities. Specific new products can be needed
to address highly specific commercial concepts which have special
value, or, the project portfolio is specifically managed into
projects of different relative priorities. The need is therefore
to generate technical solutions for those specific product concepts
according to mandated priorities. The solution to this problem
simply involves conducting sufficient numbers of independent projects,
each having the same objective, so that the overall risk of failure
for a specifically prioritized product concept is reduced. Hence
if the prioritized project concept is high risk, thus placing
it in the project category where the probability of technical
success is estimated to be, for example, 1 in 4, then one can
decrease the risk of total failure from 75% to for instance 32%
by conducting 4 projects, or to 10% by conducting 8 projects.
Generally, the required numbers of projects, N, for a risk
risk of failure to provide no more than one success is given by:
(6)
where INT is the integer function and returns the greatest whole number less than or equal to the bracketed term. If the N projects are conducted concurrently, the possibility of multiple technical successes is then admitted with the advantage that the most efficient solution can then be selected, but with the disadvantage that the overall profit potential of the portfolio is reduced by the additional project costs, since not all successes in the portfolio will be exploited. Hence, on this latter basis and if the nature of the business permits, such need-to-have concepts could otherwise be planned as far in advance as possible so that the R&D projects can be started early and performed in series as much as possible until the one required success is attained.
A portfolio design approach based entirely on maximizing
potential profit implies a business strategy which is prepared
for, and biased in favour of, attempting to commercialize any
and all innovations which emerge from the R&D project portfolio.
If lead time is not an issue then such a design approach would
be based on selecting the maximum probable project value (PPV)
which will always select the single project risk category which
has the highest PPV_x, that typically being the highest
risk category. However, in general there will also be shorter
term profit needs in order to support that endeavor and other
strategic factors such as the need to provide the minimum acceptable
R&D project support to the more highly innovative products.
Short term profit needs favour a predominantly low risk portfolio.
A certain number of lower risk project categories are generally
required for the support of products which arose from previous,
successful higher risk category projects. Accordingly, if a portfolio
includes high risk category projects, then medium and low risk
category projects will also be required in order to support that
industry which was created by previously successful high risk
projects. There is thus no general single optimum since individual
strategic factors bias the choice of portfolio mix. At this point,
the SDF function can be a useful tool if one is designing
a project portfolio from scratch with a priority of profit growth
and one first needs a nominally profit-optimized risk resistant
design for a mixed risk project portfolio. It also provides a
semi arbitrary basis for comparisons of the 'risk resistant profit
potential' of different portfolio mixes. What this means is discussed
in the following. The SDF function is a modified form of
the parameter PPV applied to a portfolio, where instead the term
for probability of overall success, PTS X PCS, (see
equation 4), has been replaced by the average probability of the
portfolio not failing to provide a project success in a
project category, per year assuming an even distribution of project
completions. Overall, this quantity has a single maximum value
versus the full range of portfolio design options. The corresponding
'optimum' design then represents the best 'worst case new source
of profit scenario per year', and depending on the project
parameters may include a mixture of different project categories.
The function, SDF, is defined in equation 7:
where the variable names are as previously described
and where:
..(8)
represents the portfolio design, and where max{previous
term} is shorthand for the maximum value of the whole previous
term in the curly brackets versus different values of N_high,
N_med and hence N_low according to equation 8, and
where total projects is constant. A candidate design obtained
via SDF thus corresponds to:
..(9)
The SDF is called semi-arbitrary because it
provides a single design based on a semi-arbitrary criterion.
However, perhaps more importantly than the ability to select that
design is that it also provides a rational one-number basis for
coarsely comparing the relative merit of actual designs to an
arbitrary standard, or reasonably defined risk resistant, profit
oriented design.
Examples
Let us consider an example where R&D resources are fixed at about 20 MM per year allowing for an R&D project portfolio comprising a total of 20 concurrent projects. To put this in perspective, this level of commitment could be in support of an industry with a turnover of around 1 bn. In order to determine a candidate strategy utilizing the SDF, equation 7 must be evaluated for each of the 231 possible combinations of values for N_high, N_med and N_low where N_high+N-med+N_low=20. The result is an array of numbers which is best examined graphically. Figure 2 is a triangular contour plot which was generated using the example project parameter values given in table 1.

Referring to the figure, one can see that the two
axes are the number of high risk and the number of low risk projects.
At each point on the plot the corresponding number of medium risk
projects is therefore equal to 20 - (N_high+N_low). The
SDF function has a maximum value of one which corresponds
to a design represented by 16 concurrent high risk projects, 3
medium risk projects and 1 low risk project. Also noted on the
plot are two other example designs which are referred to later.
One is an example of a relatively poor strategy for growth comprising
18 medium risk projects, 1 high risk and 1 low risk. For this
design the SDF value is 0.4. This strategy might be viewed
as a fiscally conservative strategy. However, in these examples
it is an overly conservative strategy for this high level of investment
if the objective is long term profit growth, although near term
profit growth is likely to be better than that of the maximum
SDF value design. Furthermore, the single high risk project
in this portfolio is a very poor financial risk because at this
level of commitment, according to the example T90 values
shown in table 2, there is a 10% risk that there will be no profits
from a high risk project for 180 years, and in fact the risk of
failing to provide one success within the twenty year commercial
lifetime is actually 77%. The other strategy noted comprises 1
high risk project, 1 medium risk project and 18 low risk projects.
The corresponding SDF value for this design is below 0.2
and the design is quite simply a waste of precious R&D resources
because it so relatively unprofitable as will be seen later.
The next step in evaluating a candidate design is
to select a risk level and check the Tc time for the highest
risk category represented. The risk level is defined as the overall
level of risk that a portfolio will perform no better than the
projections. A 10% risk level is taken as the example so that
it is now necessary to check that the T90 time for the
highest risk project category is indeed less than the commercial
life. In this example, the candidate strategy generated using
SDF comprises 16 high risk projects, 3 medium risk projects
and 1 low risk project. The T90 time for the 16 high risk
projects is 11.25 years which is less than the estimated 20 year
commercial life indicating that this strategy is acceptable.
The previously discussed breakpoint example for investment in high risk R&D occurred at a total project number of about 11, assuming there would be a need for at least one or two lower risk category projects. If at this level of total projects one wished to exclude all participation in high risk projects and concentrate instead on medium and low risk projects one can again use the SDF to determine an initial candidate strategy. In this case N_high = 0 and one obtains a single row of 12 numbers which can be examined or plotted. Accordingly, figure 3 depicts the SDF plot.

Referring to this figure one can see that the value of this function peaks for a design comprising 10 medium risk projects and one low risk project.
Profit projections are a key tool for evaluating
a portfolio design. Another indicator is the projected numbers
of successes in each category. However, here I will concentrate
on profit projections. In a research strategy involving multiple
projects conducted continuously, the longer term profit p.a. performance
is dependent on past performance. This is because the results
of successful projects extend over the commercial lifetime which
would usually be after the projects are concluded and while new
ones are being conducted. The slate is thus never cleaned. Because
of this, short duration projects are more likely to generate higher
early profits if the Tc times are short compared with other
project classes. Lower risk content risk strategies always give
faster startup times and this effect would tend to bias the acceleration
of that time for those strategies. In the technique described
here, projections are based on a time series of worst case scenarios
according to prescribed overall levels of current risk and compared
with average probable outcome projections. Worst case scenarios
are thus generated via the Tc time. For instance if the
risk level is set to 10%, then the success rate for high risk
projects is fixed at one success after each T90_high time.
Clearly, the risk of experiencing two consecutive Tc times
is less than for one. However, at any time in the future the risk
of experiencing one Tc time is the same. The result of
this approach is that the risk projections represent a continuing
risk throughout the timeline or in other words the risk is always
applied to the then current situation. Hence when as time progresses
the actual performance of a strategy becomes established a new
set of projections would be performed based on the actual initial
conditions at that time. In this dynamic application, the projections
would then be used to detect potential profit loss problems versus
risk levels or help elucidate general trends.
Portfolios with different research risk strategies give rise to not only different potential profit levels and different lead times, but also different behaviours or shapes in potential profit generation over time. Figure 4 depicts calculated profit projections made at year zero assuming no project history prior to year zero, for four different portfolio designs for a research laboratory continuously engaged in 20 projects.

The four different designs are numbered in the figure
such that design #1 comprises 20 high risk projects, (SDF
= 0.9),
#2 comprises 16 high risk projects, 3 medium risk projects, and
1 low risk project (SDF = 1), #3 comprises 1 high risk
project, 18 medium risk projects and 1 low risk project (SDF
0.4) and #4 comprises 1 high risk project, 1 medium risk project
and 18 low risk projects (SDF 0.2). Figure 4A depicts projections
corresponding to a 10% risk level while 4B depicts projections
based on the average probable success rate. Considering 4A, the
first thing to note is that the risk level of each projection
is the same, where the risk level is the continuous risk after
each success of doing no better than the projection, whereas the
profit projections for each design are significantly different.
In fact designs 1 and 2 have vastly better profit expectations
compared with designs 3 and 4. The key reason for this is due
to better design. However, there is also a built in bias in the
projections against designs 3 and 4 due to the fact that the projections
become increasingly pessimistic the further out they are taken
and for shorter duration, lower risk projects there are more frequent
successes. Figure 4B depicts the average probable outcome projections.
These projections are instead based on a steady stream of successes
at exactly the most probable success rate and represent a level
of performance towards which a strategy will in theory trend.
Comparing figures 4A and 4B one can see, for instance, that even
the 10% risk projection for design #2 is superior to the less
risky, average probable outcome projection for design #3, and
in fact overall these projections would lead one to conclude that
designs 1 and 2 are highly preferable for longer term (after 5
to 10 years) profit performance compared with designs 3 and 4,
while design 3 is more likely to outperform designs 1 and 2 during
the first 5 to ten years. Also, in these examples, the 10% risk
projections for designs 1 and 2 feature occasional sharp dips
in profits. These coincide with gaps in various product obsolescence
and introduction phases. These kinds of variation are a likely
characteristic feature of a high risk, high profit, research-based
business. They may be alarming, but provided the cause is appreciated,
then they are definitely no cause to doubt or attempt to change
the research risk strategy. Instead, it would be wise to plan
ahead for the inevitable occasional profit valley. Note also that
in these examples at 10% risk, at no time do these projected dips
in profits ever get as low as the profit projection levels of
the other two lower risk content strategies.
The average probable outcome projections assume one successful project at exactly the rate attributed to the probability of success. This is the most probable outcome. However the actual probability at the outset that these performance histories will actually become established is low. Over the course of time, there are many possible outcomes of similar probabilities. In this respect it is useful to consider these curves in terms of current risk. As an example, the assumed probability of success for a high risk research project is 1 in 16. This means that on average 1 in 16 projects will succeed. Which one of the 16 actually does succeed is unknown. The chances that it will be the 16th project is the risk of 15 failures which for this case is 38%. There is instead a 52% chance that the 11th high risk project will succeed and in the same context a 56% chance that the third medium risk project will succeed. This means that there is an even chance that the actual long term performance of any strategy involving a high risk research content would be one where the average probable outcome performance is approached or trended towards from the high side as opposed to from the low side. This an important statistical effect which applies oppositely in the case of low risk research, (defined as where the overall chances of commercial success are greater than 1 in 2). The tendency for early higher rates of success for high risk research is a bonus as long as it is understood and not expected to repeat. This effect is in fact highlighted in the 50% risk projections depicted in figures 5A and 5B which lie above the average probable outcome projections. Whereas figure 4 gives example projections for four very different portfolio designs, figure 5 instead depicts projections for two similar designs; namely design #2 and design #5.

Design #5 comprises 9 high risk projects, 4 medium
risk projects and 7 low risk projects, (SDF 0.7). These
designs are similar in so far as they are dominated by high risk
research at levels of commitment which should be capable of sustaining
a stream of highly innovative and profitable products at an acceptable
risk of failure level. The projections in figure 5 correspond
to 4 different levels of risk, namely 10%, 15%, 20% and 50%, as
well as average probable outcome. On a projected profit basis,
design #2 is significantly more attractive in the longer term
with the reasonable expectation of eventually returning roughly
50% more revenue than design #5. However, design #5 does have
the significant advantage in that the near term performance has
less risk of losses. Hence, in a startup or recovery situation,
one could consider starting out with a design such as #5 with
a view to gradually increasing the high risk research content
of the portfolio over the course of ten to thirty years.
Table 5 provides a range of numerical profit projection
data for the 5 example portfolio designs described.
Table 5: Comparison of portfolio designs 1-5 projected profits (losses) at the 10% overall risk level and average probable outcome, in MM pa, versus time since design implementation. Calculated using the example input parameters given in table 1. (* average probable outcome.)
|
|
|
|
| ||||||
|
|
| ||||||||