CHAPTER FIVE: Sensitivity Analysis

Transcription

1 CHAPTER FIVE: Sensitivity Analysis 5.1 Case for Using Switch Option to Justify SOA Investment In this chapter, we use the proposed two switching option models (Switch1 & Switch2) to do a sensitivity analysis. y comparing the two option values, entrepreneurs could understand how to align the uncertainty of investment. The application is simple. If an enterprise switches its system to SOA, the opportunity value of SOA couldn t be less than non-soa. We assume the two scenario cases: Case1: The net value of SOA is presently higher than non-soa (Advantageous) Case2: The net value of SOA is presently lower than non-soa (Disadvantageous) Why do we set these two cases by net value? In order to prove that real option is a flexible decision approach, we will show the opportunity value by pricing model and compare it with NPV. Case 1 is in advantageous situation, and NPV of SOA is originally higher than non-soa, so this situation cannot show the advantage of real option evidently. On the contrary, we should set a disadvantageous scenario that NPV of SOA is lower than non-soa, and try to find out the opportunity. Then we analyze and induce strategies - how investor could get the management flexibility and higher opportunity value. Scenario: An entertainment enterprise s movie ticket transaction system has been implemented several years, it can sell tickets on web site and help the enterprise step up the earning on virtual place and reduce the operational cost on physical place. ut the system is outdated right now and out of service capacity. The company wants to switch the system to new architecture. There are two proposals for new switch plan. One is switch to a new specific system, faster, higher capacity but it s a non-soa system (Maybe it s not different to system, besides the capacity and speed are higher). The other is switch to an SOA implementing system which could be reused by other components (i.e concert, amusement park, air line or CRM analysis.), but its investment is high. The managers prefer switching to SOA because it will be better developed. ut they also wonder that pay off in SOA will be less than non-soa. We apply the switching options to justify this question. Assume the system will be used for 5years, and the detail of benefit and investment now are: The discount rate and exercise time are set as 2% and 1 year, so the net value at stage 40

2 2 will be discounted to present by Legacy =37 Ticket sales Cost savings from physical place $ 4.6millions/year $ 2.8millions/year enefit: =( ) 5= 37millions Ioprt=15 Database and website management $ 1 millions/year Network fee $ 0.8 millions/year maintain/training $ 1.2 millions/year Investment: I =( ) 5= 15millions Net value: (37-15)/1.02 = million dollars Table 5 1 enefit/investment scenario in system non SOA =50 Ticket sales $ 4.6millions/year More sales capacity $ 2millions/year Cost savings from physical place $ 2.8millions/year Less information latency $ 0.6millions/year enefit: =( ) 5= 50millions Ipre_=0.8 Consulting fee $ 0.2 millions planning $ 0.2 millions SA/SD $ 0.3 millions Design tools and IDE $ 0.1 millions Ioprt=17 Database and website management $ 1.2 millions/year (includes backup/recovery) Network fee $ 1 millions/year maintain/training $ 1.2 millions/year ISC (switch cost) Infrastructure (HW/D/NW) $ 4.5 millions = 11 Application implementation $ 6.5 millions Investment: Ipre_ = 0.8 I =( ) 5+( )= 28millions Net value: (50-28)/1.02 = million dollars Table 5 2 enefit/investment scenario in non SOA 41

3 This research sets the benefit/investment of SOA into two cases; its net value will be presently higher than non-soa (21.2millions) in first case and lower than non-soa in second case. A parameter N means the services number connecting to the SOA ticket transaction service (reusing number). SOA (Case1: The net value of SOA is presently higher than non-soa) =40 Ticket sales $ 4.6millions/year Cost savings from physical place $ 2.8millions/year Less information latency $ 0.6millions/year =4*N Cost reduction from reusing projects $ 2 millions/each reusing Time to market $ 2 millions/each reusing =( ) 5=40(millions) =4 N= 4 N (millions) enefit: = + SOA Ipre_SOA = 1 Cost of consulting $ 0.2 millions Ioprt = 5+3.5N ISC (switch cost) =12.5 Investment: planning, requirement analysis service design Metadata catelogs Integrated development tools Database and website management (includes backup/recovery) Network fee $ 0.25 millions $ 0.25 millions $ 0.2 millions $ 0.1 millions $ 1 millions/year $ 0.3 millions/year each reusing Maintain/training $ 0.4 millions/year each reusing Infrastructure (HW/D/NW) $ 5.5 millions usiness application Front-end application service interface $ 3 millions $ 2 millions $ 2 millions Ipre_SOA = =1 million dollars I =(1+0.3 N +0.4 N ) 5+( )= N (millions) Net value: -1 + (40+4N N)/1.02 = *N millions Table 5 3 enefit/investment scenario in SOA (case 1) 42

4 SOA (Case2: The net value of SOA is presently lower than non-soa) =40 Ticket sales $ 4.6millions/year Save cost from physical place $ 2.8millions/year Less information latency $ 0.6millions/year =3.5*N Reusing project cost reduction $ 1.7 millions/each reusing Time to market $ 1.8 millions/each reusing =( ) 5=40 (millions) =3.5 N= 3.5 N (millions) enefit: = + SOA Ipre=1.2 Cost of consulting $ 0.25 millions planning, requirement analysis $ 0.3 millions service design $ 0.3 millions Ioprt =5+3.5*N ISC (switch cost) =13.5 Metadata catelogs Integrated development tools Database and website management (includes backup/recovery) Network fee Maintain/training Infrastructure (HW/D/NW) usiness application Front-end application service interface Investment: Ipre_SOA = 1.2 million dollars SOA Net value: (40+3.5*N -( *N))/1.02 = millions Table 5 4 enefit/investment scenario in SOA (case 2) In case 2, we set that net value doesn t imply the effect of N. This helps us that the option value still rise as N is increasing even connecting number doesn t raise net $ 0.25 millions $ 0.1 millions $ 1 millions/year $ 0.3 millions/year each reusing $ 0.4 millions/year each reusing $ 5.8 millions $ 3.3 millions $ 2.2 millions $ 2.2 millions I =(1+0.3 N +0.4 N ) 5+( )= N (millions) value. The option value will be affected purely by but not by the net value. ( N ) If the investor uses NPV to justify the project value by equation 2-1 V SC V NPV = I + new() t () t T ( 1+ r) NPV of Switch 1= 1 + ( ) /1.02 = 1 43 where : V = I oprt In case1: NPV of Switch2 = -1+( )/1.02 = *N In case2: NPV of Switch2 = -1.2+( )/1.02 = pre NPV is not an option methodology. If N>1, NPV will be positive and the project will be undertaken in case 1. y the way, there s no flexibility and opportunity that

5 switch2 will be rejected anyway (negative) in case 2. ut rely on real option approach, we could justify uncertainty additionally and switch2 could have opportunity to surpass switch1 under case Sensitivity Analysis for Scenario Case The sensitivity analysis is conducted under different settings of uncertainty parameters (risk), T (exercise time) and N (reusing number) Option Value under risk ase on Table 4.1~4.4 we set the parameters as follows: Legacy non-soa SOA (Case 1) SOA (Case 2) N= N= I 15 I 28 ISOA ISOA : 0.3 N+1 ( N ) = ln / LC ( μ r) +1 2 T : 1 year (52 weeks ) We mainly discuss the risk of underlying projects what a company wanna switch the system to, so the is set fixed. We change & from 0.1 to 1.1 and examine the impacts on option value. We set axis x as option value of Switch1 and Switch2, axis y as, &. Figure 5 1 Sensitivity Analysis of in Switch1 44

6 \ Figure 5 2 Sensitivity Analysis of in Switch2 (Case1) Figure 5 3 Sensitivity Analysis of in Switch2 (Case2) Figure 5-1 is the sensitivity analysis of Switch1. Figure 5-2 and 5-3 perform case 1 and case 2 of Switch2. The analysis shows that the higher the risk is, the more the connection number is, the higher the option value is. Indeed, the connecting number (N) will increase the risk factor of service-integration ( ). For this reason, the option value of Switch2 would rise as N is increasing. More specific findings are summarized: 1. In case 1: no matter what risk is, Switch2 will be higher than Switch1. 2. In case 2: Switch 2 is lower than switch 1 unless N is high enough (N >= 7). ut if among SOA is much higher than (i.e., = 0.9 and =0.4), there exists probability that Switch2 could be higher than 45

7 Switch1. 3. In case2: Switch2 is lower than Ipre_SOA(1.2 million dollars) while is too low (<0.2). This could result that the budget of initial investment will be too low Option Value under Various Exercise Time ase on Table 5-1~5-4, the parameters are set as follows. All risk factors are given as 0.3, besides is computed by ( N ) Legacy non-soa SOA (Case 1) SOA (Case 2) I 15 I 28 SOA N= N= ( N ) I (N+1) ln 2 = +1 LC ( μ r) ISOA ( N ) ( N ) We set the exercise time from 24 to 88 weeks and conduct the sensitivity analysis. In the following figures, axis x presents option value of Switch1 and Switch2, and axis y presents exercise Time. Figure 5 4 Sensitivity Analysis of T in Switch1 46

8 Figure 5 5 Sensitivity Analysis of T in Switch2 (Case1) Figure 5 6 Sensitivity Analysis of T in Switch2 (Case2) Figure 5-4 is the sensitivity analysis of Switch and 5-6 perform case 1 and case 2 of Switch2. The option value rises as the longer exercise time is deferred and the connection number is more. We summarize the findings: 1. In case 1: no matter how long the exercise time will be, Switch2 will be higher than Switch1. 2. In case 2: Switch2 is difficult to be higher than Switch1. Only if exercise time in Switch2 is much longer than in Switch1 (i.e., T = 50 in switch1 and T = 88 in switch 2), the option value of Switch2 could be higher than Switch1. 3. In case2: Switch2 is lower than Ipre_SOA(1.2 million dollars) while exercise time (T) is too short(t<30). This results that the budget of initial investment is too low. 47

9 5.3 Analysis for Measure Errors Caused by Implementation Delay This research statistic the number of measure error 1 and error 2 after Monte Carlo simulation (the statistic method is mentioned in section 4.4). An interesting phenomenon was discovered. We found out two factors affect the number of error obviously, one is delay time - d & d SOA, the other is coupled coefficient ρ, Measure Error under Delay Time Parameters are set as: : 37 I : 15 : 0.3 SW1: : 50 I : 28 : 0.3 T = 52weeks SW2 : : 40 : 20 I : 35 : 0.3 SOA, ρ = 0.36 N = 5 T = 52weeks d & d SOA is varied from 2 weeks to 18 weeks as axis y. Axis x means the numbers of error 1&2. Errors numbers is the times errors happened among simulation paths. Figure 5 7 Sensitivity Analysis of d in Switch1 48

10 Figure 5 8 Sensitivity Analysis of d in Switch2 SOA y figure 5-7 and 5-8, we find that the implementation delay is caused by more numbers of errors Measure Errors under Coupled Coefficient We also conduct analysis of coupled coefficients in Switch2. Parameters setting: Original: 37 I Original : 15 Original: 0.3 : 40 : 20 I SOA : 35 : 0.3 dsoa : 6weeks T = 52weeks N = 5 The coupled coefficient only exists in Switch2 (SOA). Hence, the error number analysis is performed for Switch2. ρ, is set from 0.1 to 1 as axis x, and Axis y present the same as which in section

11 Figure 5 9 Sensitivity Analysis of ρ, in Switch2 y figure 5-9, it showed that the loosely coupled (i.e. low ρ, ) system can reduce more error number than tightly coupled (high ρ, ) system. 50