2. Decision trees, real options and scoring techniques

Size: px
Start display at page:

Download "2. Decision trees, real options and scoring techniques"

Transcription

1 abstract There are fundamental challenges valug technology, particular dealg with uncertaty, complexity aggregation issues. In this paper we identify desirable properties valuation tools then review existg tools agast these criteria. We then establish a need for a tool to explore communicate the value technologies at the earliest stage development, propose a tool, the value roadmap, that fills this gap. We report on itial learng from pilotg this tool at a multational 1. Introduction corporation. Valug technology is a critical busess activity many companies. For example. selectg the best portfolio projects is ten key to the long-term future the company6). It is usually this that creates the new busesses that generate future revenues. And it is usually this that fds solutions for busesses whose systems are operatg at the limits performance. With the rise technology tradg licensg, the valuation technology has taken on new importance. The ability to correctly value technology is useful to many: employees structurg their projects to maximise value; managers selectg among projects; higher management wishg to underst the benefits spendg on technology development; employees responsible for acquirg or sellg technology. This paper reviews a range tools techniques available to do this technology valuation for dividual projects, proposes a new one, value roadmaps, too fill an identified gap with early stage valuation techniques. Before embarkg on this, we consider this troduction the fundamental challenges that valuation tools face, the desirable characteristics a good tool. We then go on to outle the structure the rest the paper. First, we consider the challenges. The challenges valug technology are significant. The startg pot is that technology has no herent value: it has value when combed to a complete solution delivered to a customer who is willg to pay a certa amount for it their particular application. Major challenges attributg a value are: uncertaty, complexity, aggregation effects. We start by considerg uncertaty. All predictions future value are uncerta, but those volvg early stage technologies are particularly so. There is more technical uncertaty as to what is physically possible, when dog thgs for the first time there is also far more scope for unseen problems to emerge. Even if the technical problems are resolved there rema market uncertaties. These are ten considerable sce the conjectured market for an early stage technology will lie some way to future. All these uncertaties make it difficult to produce useful estimates the value an early stage technology, particularly sce the actual fancial prit derived from a technology will be the difference between an uncerta revenue an uncerta development marketg cost. These uncertaties lead naturally onto the problems complexity. The more uncertaties there are, the greater the number possible futures that need to be considered. If, for example, a technology proves effective for one application, there may be other applications where it could still be valuable. If an superable problem occurs with a technological solution, there may be parts the solution that can be salvaged usefully exploited. A further complication to consider is that the route to obtag value may volve other partners who, for example, control distribution channels or other complementary assets. Bounded rationality means that it is impossible to take to account all the potential teractions that occur as a technology is developed exploited. In short, valug early stage technologies is complex.

2 Figure 1 Simple DCF valuation 5) Figure 2 Decision tree valuation5) This complexity leads turn onto issues aggregation. As mentioned above, a technology needs to be encapsulated to. a solution for a customer before it generates revenue. This solution ten combes other technologies. Thus aggregatg technologies has this case creased their value. How should we value technologies lked this way? How much the combed value should be attributed to an dividual technology? It may conversely be that two parallel technology projects the loss focus. are less valuable than a sgle one, due to In summary, valug technology faces fundamental challenges, particularly with respect to uncertaty, complexity aggregation. Different valuation tools tackle the challenges different ways. What are the desirable characteristics such tools? Obviously a key desirable is accuracy valuation, with the accuracy the model matched to the accuracy the available put data. Beyond this for practical reasons the tool needs to be easy-to-use. Clearly a balance needs to be found between the desires for accuracy for simplicity. Another desirable characteristic is that the tool be tuitive generate understg - managers engeers want to be able to discuss the project light the valuation, to underst where the value comes from particularly the value. to be able to see ways to improve Ideally the tool should be widely applicable to a variety projects be scalable. Fally the tool needs to be credible accepted as useful. There is little pot a tool that is easy-to-use, tuitive accurate, willg to implement if no-one believes the answer or if no-one is the process. This paper has four further sections. The next sec- tion reviews decision trees real option techniques that address the well known weaknesses discounted cash-flow techniques the presence uncertaty ( telligent formed management). The section then looks at scorg techniques before identifyg the need for a tool to complement scorg techniques. Section 3 describes the value roadmap, a tool aimed to fill this need, reports on prelimary learng from applyg it. Section 4 concludes by identifyg areas for further work. 2. Decision trees, real options scorg techniques Discounted cash flow (DCF) techniques score well on the desirable tool characteristics identified the troduction. They are easy-to-use, tuitive, widely applicable, credible accepted. However their accuracy can be poor if there are high levels uncertaty, if the project can be actively managed to reduce the impact bad outcomes or boost the impact good ones. This pot is well made a very simple example Faulkner (1996), shown Figures 1 2. In figure 1, a stard DCF approach is taken on a hypothetical prter project. The net present value this project is negative. In figure 2 however, by distguishg the different outcomes assigng probabilities to them, decidg to launch the prod- the expected discounted value is positive. This decision tree method valuation produces a more accurate reflection the value the project, assumg the estimates probabilities the correspondg market values are reliable. This is simply due to it beg a more accurate model the staged vest-

3 ment process - a reasonable manager will not launch a product that is likely to make a loss (unless there is some other benefit to be obtaed). Transferrg this simple example to the real world immediately leads to the question where probability figures such as g0.3 h came from. It may be that there is a database previous similar projects thirty percent them had excellent outcomes. This is the approach adopted for generatg probability figures for a pharmaceutical decision tree reported by Loch Bode-Greuel (2001). Another approach is to view probabilities as representg subjective judgments experts based on experience as to how much should be bet on an outcome. Neither these approaches gives great confidence thus can underme the credibility acceptance decision trees. To counter this, sensitivity analysis can be performed on the parameters, seeg how the value changes as the outcome probabilities change. The logical extension this yields Monte Carlo valuation methods. The advantage the decision tree approach is that it generates a more accurate value than DCF the case where there is uncertaty formed management flexibility. It achieves this at the expense beg slightly less easy-to-use slightly less credible, due to appearance probability estimates. However it is still tuitive understable. It can be used as the focus a discussion managers can consider if there are other configurations the project that have more value, whether it is possible to do particular de-riskg pilot studies to determe the likely value the project earlier, hence before major vestments have been made. A key observation is that it is ten more useful to underst the range likely outcomes have plans to hle these outcomes, than it is to simply have a sgle figure representg the value averaged over these different outcomes. This decision tree approach has similarities with the so-called real options approach. The fundamental idea that uncertaty is good if the downside can be mitigated is common to both. For example it may be worth dog a pilot study to a technology to establish if it will work or not. If the answer is yes then a large vestment can be made to reap an even larger benefit. If the answer is no, then only the small cost the pilot study has been curred. This is contrast to the DCF approach where uncertaty is typically penalized by raisg the discount rate. The term real options valuation is ten reserved for approaches which derive from the Black-Scholes- Merton1) model for valug options contracts on the fancial markets. An option contract gives the right but not an obligation to e.g. buy a certa amount an asset at a particular price on a future date e.g. 3 months time. The model assumes the asset price moves accordg to a rom walk process. It is then possible to theoretically construct a portfolio some the asset some options which is isolated from the ran- can then be calculated, option14). from this the value the The real options valuation approach makes the analogy between an option contract a project. In the world fancial markets, by payg a small amount for an option, you can buy an asset if the price is favourable abon the option otherwise. In the world development, by payg a small amount for some, you can launch a technology if the result is favourable, abon the otherwise. However the details this analogy are open to question12). In particular it is unclear whether the risk-free portfolio argument carries across, to what extent the value the output a project follows a rom walk. It tuitively seems that a decision tree with discrete events correspondg different stages is a better model9). to the end Hybrid models have been constructed which use decision trees for modelg the early stages development then lk these to models the market value that follow particular stochastic processes7,10),11). However the promise real options valuation based approaches seems low when judged agast the criteria suggested the troduction. Direct application the Black-Scholes-Merton model is usually flawed2), producg recommendations such as that delayg the product launch will necessarily crease the value. A more realistic model will be more complicated. Even if this more realistic model is accurate easy-to-use, it is unlikely to be tuitive help managers ers fd ways to crease the value. Sce the mathematics underpng the models is high level, the underpng assumptions are not self evidently reasonable, it is hard for these models to become credible accepted. In conclusion, decision trees Monte Carlo extensions these seem to be the best tools for evaluations

4 where the uncertaties make DCF appropriate, but where it is still possible to identify likely development branch pots ascribe credible probabilities. Where the uncertaties are such that it is difficult to identify likely development branch pots or ascribe credible probabilities, new techniques are needed. The most commonly used approach is scorg projects agast a number qualitative factors such as how broadly applicable the technology is, or how well it fits with company strategy, then rankg the projects accordg to their total scores. As well as selectg the highest scorg projects, attention is paid to achievg a balanced portfolio, typically by a visual plot. Cooper et al. (1997; 2001)3),4) review a range these tools make the tellg observation that companies who rely exclusively on fancial measures to rank projects are less successful at new product development than those who also clude qualitative factors. One weakness scorg methods is that it may be hard to justify why a particular score on a qualitative factor was given. For example, it is hard for senior managers to query a strategic fit score without some understg the project. Also the ability managers engeers to underst improve the value a project, which was possible with decision trees, has been lost. To address this gap we developed the concept value roadmaps, extendg previous work on technology roadmaps 13) 3. Value roadmaps 3.1 Value roadmap structure Value roadmaps are a way to explore improve the value technology projects at a very early stage. As well as supportg communication with the project Figure 3 Value roadmap architecture team, the roadmaps can be post-processed to emphasize key messages can then be used as a tool for communication with senior management. The structure a value roadmap (VRM) is depicted figure 3. Typically a roadmap is developed on a wall chart durg the course a half-day workshop. It comprises the followg four layers: environmental, technological political) ternal busess factors that fluence the development products technology the area terest, cludg strategic milestones goals. gs: products, services, busess/facilities, technology/ip, cost/risk reduction, strategic position). All these value streams relate directly to the generation cash revenue, except for estrategic position f, which cludes all non-fancial factors that provide a foundation for future revenue generation. challenges risks, together with complementary assets actions needed to exploit the potential value the technology or capability) vestment. A key feature the VRM is the time axis, which lks the short-, medium- long-term perspectives for all the layers sce vestment now is tended to generate revenue the future. The time horizon for the VRM will typically extend considerably further to the future than the project plan, providg a forward The value proposition that is explored mapped the VRM will typically depend on the strategic context or scenario that governs the discussion defes the broad direction with which novation is desired. It is important that the strategic context is clearly articulated, cludg assumptions, constrats desired end result the vestment. 3.2 Value roadmap process The followg process stages are used to generate the VRM: 1. Defe strategic framework/vision/scenario (assumptions, boundaries, constrats). 2. Map market busess trends drivers (social, economic, environmental, technological political), milestones goals. 3. Map project milestones vestment (cur-

5 rent future/potential), terms the technical capabilities that will be achieved at key milestones. 4. Map barriers enablers associated with technical capabilities, terms the challenges risks associated with realizg the commercial value from the technology, together with the associated complementary assets actions that must also be place. Consider both technical non-technical factors. 5. Steps 1-4 provide the context (strategic framework, market pull technology push) with which the potential value that may result from the vestment can be explored. The goal is to identify specific sources potential future revenue, articulated as clearly as possible. 6. Review project plan VRM, cludg key lkages between elements. 7. Generally, for non-trivial projects it is expected than the terestg technical challenges or capabilities produced. It also caused them to rethk the order the capabilities developed the. The major difficulty encountered was placg formation on the chart without the chart becomg too densely packed with formation. Experiments were performed post-processg the fal roadmap, havg multiple versions at different levels detail, also the possibility usg IT support that enabled the terrogator the VRM to dig down to items particular terest. An alternative was to generate narrative roadmaps which the er used the headgs as prompts to enable them to communicate the value. However clearly the advantage the one unifyg visual representation the project is lost by dog so. 4. Conclusions further work The value roadmap approach appears to be a helpful value roadmap will be dense, complex fragmented, with gaps data varyg quality. Further effort will be required to tidy up the roadmap, addition to the technology valuation toolkit, although it by no means solves the key central challenge fancial valuation long-term. However, it does provide although the end result will still probably be a complex roadmap-the VRM is designed to reflect the complex ebig picture f. This erich picture f VRM can be considered as a edatabase f, contag a great amount relevant formation at a fairly high level detail, is likely to be too dense to clearly communicate key messages about the project its value. For this purpose, summary or communication roadmaps need to be developed. It is necessary to consider carefully what the key messages are, who the audience is, then use the pact events new formation on the plan as a whole. 8. Mata the VRM associated documentation on an ongog basis, preferably as part the busess process (project management, new product development, technology strategy). 3.3 Value roadmap application The value roadmap technique was successfully piloted at a multational company. The feedback from users was positive particularly that it forced ers to thk communicatg the value a project rather a framework with which the full context technology vestment exploitation can be explored communicated. The VRM approach is a particular type the more can be learnt from practice that area. While the concept behd roadmaps is simple (a multi-layered timebased graphical framework), the reality developg matag roadmaps is much more challengg, reflectg the complexities strategic planng more generally. Essential gredients a successful roadmappg itiative clude: an effective process (lked to other key busess processes), support (facilitation potentially stware), ownership. Our next steps developg the VRI approach are to: 1. Undertake further pilot studies for a range project types, capturg key learng pots. 2. Develop solid historical cases to demonstrate application VRM, also to illustrate the non-lear path to commercialization that is typical many successful activities, which might well have not been funded at the time (without hdsight) if stard accountg approaches such as DCF were used isolation.

6 layer VRM, terms strategic milestones technical teams, also set that for because a can consistency this trends, be provided across formation is drivers to projects ten not 14) readily available. 4. Develop VRM process frastructure guide (workshop based on further support facilitation trials Technology roadmappg developg a practical approach for lkg resources to strategic goals. Proceedgs the Institute Mechanical Engeers 217. Wilmott, P., S. Howison J. Dewynne (1995). The mathematics fancial derivatives: a student troduction. Cambridge, Cambridge University Press. - stware), [Author's Mr. Francis (Received July 1, HUNT 2004) Francis Hunt is a er tre for Technology References 1) 2) Black, F. other omy 81, corporate The pricg Journal G.T. 12-6,, S. Edgett management for University. options Political Econ- decision Real makg. new practices study. Cooper, G., S.J. (1997).: Portfolio E. R&D the development: 31-4, E.J. 1. (1996).: valuation. product ences develop- Technology thkg f (1999).: Six reasons to do stitute versity to 39-3, long-term 42-4, Valuation brid model. ence on 8) Iansiti, 9) 10) Loch, great J. West 75-3, C.H. options projects. Neely, J.E. valuation sources practical hy- School selection. Harvard value for Busi- product Technology, Mr. real options projects. Policy - for TechnolUniversity 1997, a is currently engaged to vestigate strategic technology issues manufacturg organi- He has terests roadmappg valuation. re- 231/248. Hybrid development has Rob Phaal joed the Centre ogy at Cambridge technology Evaluatg pharmaceutical 31-2, Neufville PHAAL programme management tegration- products. de risky Journal Rob Portl. Bode-Greuel Philips Phaal Confer- Technology great with general manfellow at the In- for Manufacturg at Cambridge Uni Visitg Pressor at Cranfield sations. K. as a International Technology. to explorg Portl 23 years. 69/&. search Wong (1997).: Review J.C. technology: growth tional Proceedgs M. Turng ness Probert, project He spent strategic planng He is a Visitg 8/11. Mr. D. plc., agement. Research-Technology Applyg eoptions Research- F.H., (2003). Rick Mitchell was until recently Group Technology Director Domo Prtg Sci- 361/380. terests Hunt, are technology MITCHELL 16/28. Research-Technology 7) Rick an Kleschmidt new leaders. Mr. Portfo- 50/56. D. terests stware op- results Edgett from 40-5, T.W. Hicks, Organization Kleschmidt product management Lessons 6) His technology valuation management. at the Cenat Cambridge 772/777. dustry Faulkner, Moskowitz strategic Cooper, (1973).: liabilities. E.H. analysis ment; 5) Scholes Science lio 4) M. 637/659. Bowman, tions 3) Priles] examples. PROBERT David Probert is a senior lecturer the Department Engeerg is a foundg Interna- David 1-1, member Head the Centre for Technol- 29/46. 11) Perdue, K., Berkey (1999).: tions W.J. pricg McAllister, Valuation decision P.V. R analysis D Kg projects models. B.G. usg Interfaces op- ogy with the Institute for Manufacturg at the University Cambridge. His terests clude make-or-buy 29-6, strategy technology valuation. 57/74. 12) 13) Perlitz, M., tions valuation; tion? Probert, 計測 と制御 T. the new D., 第43巻 Peske C.J.P. 第10号 Schrank frontier 29-3, (1999).: project Real op- evalua- 255/269. Farrukh 2004年10月 号 Phaal (2003).: 735