All of this leads to a somewhat practical means of determining the proper discount rates for NPV analysis of projects.
Choose a security of a company that performs many projects similar to the project under consideration.
Determine the stocks
This can be determined by looking at on-line financial information sites such as Yahoo or Google finance.
Calculate the rate of return required from the CAPM.
The complexity of doing this correctly points to the need for participation of both technology and finance departments in assessment of project value.
4.2.7 Sensitivity and Scenario Analysis
Discounted cash flow (DCF) or NPV analysis is critically dependent upon estimates for costs and for future cash flows. Obviously, these estimates are error prone. One way to understand and quantify the possible errors in costs and cash flow estimates and ultimately in NPV analysis is via sensitivity analysis. The estimates are functions of sets of independent variables. For instance, cost might be a function of lines of code. Base estimates for each independent variable can be made. From these, base-case cost or cash flow estimates can be produced. Then, one variable at a time can be changed from its base value to pessimistic and optimistic values. Cost and cash flows are recalculated as is NPV. From this analysis, it is possible to determine key variables that impact NPV and to further study them to reduce possible errors.
One criticism of sensitivity analysis is that only one variable is modified at a time. Another type of analysis, scenario analysis, attempts to correct this by modifying several variables at a time. For instance, in a software project, choosing a particular programming language may impact lines of code, programmer productivity and defect rates. Modifying each independently, as down with sensitivity analysis, may not reflect reality.
Monte Carlo simulations can also be used to study dependencies on several variables at once.
4.3 Real Options
4.3.1 Introduction
The valuation techniques discussed in Section 4.1 have a common theme. There is an implicit assumption that project is always carried out as initially planned and that management does not actively modify the project once it is in progress. Cost estimates and cash flows forecasts were prepared and were discounted using a discount rate appropriate for the project risk and NPV, IRR and payback period were calculated. As a project proceeds valuable information is learned. There are opportunities to exploit good fortune and success and opportunities to change direction or even stop the project in the case of bad fortune and failures. Additional knowledge as the project proceeds leads to additional choices and additional opportunities to create wealth or to avoid loss. “Opportunities to modify projects as the future unfolds are known as real options” [42]. Additional information gathered as the project unfolds and the ability to modify the course of the project have real value. Intuitively, all things being equal, a project that is easy to modify based on better knowledge has more value than a project that is harder to modify once in progress.
There are several types of real options or opportunities to modify projects as they progress. The first is the option to expand. An example of this is a pilot project or initial version of a product with a minimal feature set. Suppose that a new web based store is proposed that will allow users to order coffee that will be delivered to them. This new business might be piloted in a limited market and then expanded nationally based on the observed success.
Create initial pilot in limited market Observe Results in Test Market Cancel Effort Expand Nationally
Figure 9 Decision Tree for Option to Expand
The high values of certain of certain internet companies such as Amazon and EBay are related to the option to expand. The high stock prices of these firms are related to the perception that they have options to expand infrastructure and technology into new markets. Based solely on a discounted cash flow analysis (NPV) of their current business, their stock prices would be significantly lower.
A second type of real option is the option to abandon. Although not the case in the software industry where old software and computers have little or no value, in other industries after the choice is made to discontinue production there is often significant salvage value to equipment and facilities. Depending on choices made during a project, the salvage value may vary. For instance, one means of production may utilize machinery with significant salvage value while another may utilize machinery with less salvage value.
The following example is modified from [43]. Assume that a brokerage firm has a choice between developing a specialized object oriented database or using a standard vendor provided relational database, such as Oracle, to support a trading system for a completely new market. The object oriented database is highly specialized, has lower operational costs and no salvage value. The relational database is more standard, has higher operational costs and some salvage value. Suppose that 10 million in licensing costs could be saved by using the relational database on another project if the new trading system fails. Suppose that the project development costs are equal for both technology options. Further suppose that the expected payoffs (incoming revenue discounted to the current time) from each technology are based on the demand for this product and are as follows:
Payout (Millions) O.O. Database Oracle
High Demand 20.5 18.0
Sluggish Demand 8.5 8.0
The payout represents the NPV of all future cash flows at the time the system is put into use. Clearly, if we were certain that demand would be high, the custom O.O. database is preferred. But suppose that demand is sluggish. With the custom database, the best choice is to keep using the trading system to receive a payout of 8.5 million. The specialized object oriented database cannot be reused. With the standard Oracle database, we could continue to use the system and receive 8.0 million in payout, but a better choice would be to shut it and reuse the database for another purpose saving 10 million. The option to abandon increases the value of the project when a standard reusable technology is chosen.
A third type of real option is referred to as a timing option. Suppose that a project is calculated to have a positive NPV, but there are uncertainties. For instance, these uncertainties could be about the overall economy or about projections on future demand. In this case, there may be value in delaying. The overall economy could improve or additional details about future demand might become available. Erdogmus [44] described a phased migration option as a type of timing option. A basic system can be deployed and enhanced in a phased manor. Each enhancement is done only as demanded by market conditions. Real options can quantify this approach.
The fourth type of real option is referred to as a production option. Part of the value of modularity in software is due to this. Suppose a new software system is being developed and there is a choice between making the system modular and expandable at higher cost or less modular and more restrictive at lower cost. A simple NPV style analysis might indicate that the less modular approach is preferred. Real options techniques provide a way to quantify the value in a flexible, modular design. Having the ability to combine, and selectively replace or improve software modules increases the value of a system [45].
4.3.2 Decision Trees
Decision trees can be used to calculate the value of a project when options exist and can also be used to quantify the value of an option. The following example is taken from [46] with modifications.
A company is considering entering a new web-based business. It can initially build a rich full featured system or a limited system with minimum functionality. The cost of the full system is 550K, but it will attract more customers. The cost of the minimal system is
250K and is less appealing to customers. The minimal system can be expanded after one year of use for 150K. Due to shortcuts made during its initial construction, even with an upgrade the minimal system will not be as appealing to customers as the full system.
There is uncertainty in the level of demand and thus in future cash flows. Assume that there is a 60% chance of high demand in year one and a 40% chance of low demand. This was determined by marketing and sales groups within the company and will be assumed accurate. Assume that given high demand in year one, there is an 80% chance of high demand and a 20% chance of low demand in year two. Assume that given low demand in year one, there is a 40% chance of high demand and a 60% chance of low demand in year two. Assume a discount rate of 10% is appropriate. Assume that the future cash flows shown have been produced from accurate marketing and sales forecasts. This is depicted below in Figure 10.
Full Syst em -550 Pa rtia l S yste m -2 50 High Dema nd (0.6) 150 Low Demand (0.4) 30 High Dema nd (0.6) 100 Expans ion -150 No a dditiona l investm ent Low De mand (0 .4) 50 High Demand (0.8) 960 Low Demand (0.2) 220 High Demand (0.4) 930 High Demand (0.8) 800 High Demand (0.8) 410 High Demand (0.4) 220 Low Demand (0.6) 140 Low Demand (0.2) 100 Low Demand (0.2) 180 Low Demand (0.6) 100 YEAR 1 YEAR 2
Figure 10 Decision Tree Example
As can be seen from the figure, if the full system is constructed, there is a 60% chance of high demand in year 1 with a cash flow of 150K and a 40% change of low demand with a cash flow of 30K. Given high demand in year 1, there is a 80% chance of high demand in year 2 with a cash flow of 960K and a 20% chance of low demand with a cash flow of 220K.
Using discounted cash flows and expected values, it is possible to calculate a NPV for each system choice.
Full System: