518 Proceedigs of the 8th Iteratioal Coferece o Iovatio & Maagemet Risk Assessmet for Project Pla Collapse Naoki Satoh 1, Hiromitsu Kumamoto 2, Norio Ohta 3 1. Wakayama Uiversity, Wakayama Uiv., Sakaedai 930, Wakayama-city 640-8510, Japa 2. Professor emeritus, Kyoto Uiversity, Japa 3. IBM Japa, Ltd., Japa (E-mail: satoh@ceter.wakayama-u.ac.jp) Abstract Followig the adoptio of probabilistic risk assessmet (PRA), which has traditioally bee applied to the risk assessmet of physical systems like uclear power plats, chemical plats, railroad facilities, ad so o, to iformatio security, this study attempts to apply PRA to project maagemet (PM). This paper discusses the quatitative risk assessmet by PRA, especially focusig o the case i which cost plaig of a project collapsed. Key words PRA; Risk assessmet; Project pla collapse; Sceario maagemet 1 Itroductio Probabilistic Risk Assessmet (PRA) is a strog tool for assessig the safety risks of physical system such as uclear power plat, chemical plat, railway facilities, etc. The study of PRA was made public i 1975 with the codeame WASH-1400 as oe of the studies o the safety of the uclear reactors i the Uited States. Whe a accidet type has bee idetified as, for example, a explosio, there exist various steps, scearios, ad a series of evets before the occurrece of the accidet. I order to quatify the risk of the accidet, i the first place, it is ecessary as well as importat to eumerate the scearios ad quatify the scearios. I the field of iformatio security ad project maagemet, the same ca be applied. The previous study of mie attempted the applicatio of PRA to iformatio security [1], ad this study examies the applicatio of PRA to project maagemet. To be cocrete, based o the sceario, the quatificatio of the project pla collapse of cost pla, quality pla, ad staff pla is discussed. PRA is composed of evet trees ad failure trees. While failure trees are employed to aalyze the causes of fuctioal failure, evet trees are the key tool for eumeratig the accidet scearios. The evet tree is a kid of decisive tree that starts from the iitiatig evet ad fially reaches success or failure. The accidet sceario of each idividual iitiatig evet is eumerated with the evet trees that start from the iitiatig evet. 2 Sceario Maagemet As the prelimiary step for the applicatio of PRA, this sectio illustrates the course of a project pla collapse, eumeratig the courses of the pla collapse of cost pla, staff pla, quality pla, ad delivery pla. 2.1 Cost pla collapse The course of the cost pla collapse is as follows: 1)The i-house desig documet was haded to the cliet, who orally accepted the desig. 2)However, o documeted ispectio acceptace was obtaied from the cliet. 3)Bugs i the geerated file were detected by the system test. 4)I order to correct the bugs, modificatio program had to be geerated. As a result, icremetal ma-hour is required. 5)The egotiatio with the cliet broke dow due to the additioal fee, the delayed delivery, ad the additioal staff. 6)The budget was ot secured eough. Cosequetly, the cost pla collapsed. 2.2 Staff pla collapse The course of the staff pla collapse is as follows: 1)Sice the umber of system egieers who are familiar with the work was few, the project pla depeded o the parter compay staff. 2)Due to the flaw of the project, delivery to the cliet had ofte bee delayed. However, the delivery deadlie of the deliverables was at the begiig of April. O the other had, the ma-hour of the system egieers from the parter compay was overflowig due to trouble resolutio ad Q&A
Proceedigs of the 8th Iteratioal Coferece o Iovatio & Maagemet 519 correspodece. Moreover, due to the flaw of the project, ew tasks were accumulatig, which led to the lack of Key Me who are able to desig with good kowledge of specificatios. As a result, it was impossible to implemet the project with the staff pla. Thus, the staff pla collapsed. 2.3 Quality pla collapse The course of the quality pla collapse is as follows: 1)The implemetatio test was doe, but it did ot ecompass all the real jobs, i.e., it was doe by samplig the weekly, mothly, ad aual jobs o the ordiary olie. 2)The test was codig by usig the Ispect Istructios of PC Cobol. Four to five hours after the real olie started workig, the middleware caused Abed. Abed Dump was obtaied, ad Dump was aalyzed to idetify the cause. The, it tured out that the cause of Abed was the Istructio Ispect of PC Cobol, thus the patch was made. I this case, if the implemetatio test ecompassed all the real jobs, this Abed could be evaded. The, recovery measure was take by re-codig with PC Cobol without usig Ispect. Fially, operatio check ad operatio test were coducted. As a result, medical accoutig jobs recovered, but it took log to recover, ad the work stopped. Thus, the origial quality was ot obtaied, ad the quality pla resulted i collapse. 2.4 Delivery pla collapse The course of the delivery pla collapse is as follows: 1)The i-house desig documet was haded to the cliet, who orally accepted the desig. However, o documeted ispectio acceptace was obtaied from the cliet. 2)Bugs i the geerated file were detected by the system test. 3)I order to correct the bugs, modificatio program had to be geerated. As a result, icremetal ma-hour is required. 4) The egotiatio with the cliet for icremet of staff broke dow. 5)Despite the ecessity of icremetal ma-hour, it was ot possible to icrease staff members. As a result of the work doe by curret staff members, the delivery due was ot observed. Thus, the delivery pla collapsed. 3 Applicatio of PRA to Project Maagemet I applyig PRA to project maagemet (PM), firstly, it is ecessary to defie the accidet ad clarify the iitiatig evet for the occurrece of the accidet. Here, the iitiatig evet is the factor that triggers the accidet. The, the accidet sceario is eumerated with evet trees. Fially, evet probability of each sceario is calculated. 3.1 The merits of the applicatio of PRA to PM The coceivable merits of the applicatio of PRA to PM are as follows: 1)Accidet scearios ca be eumerated with evet trees. 2)The risk ca be assessed by both the combiatio of the evet probability of the sceario ad the degree of ifluece. 3) The risks ca be assessed for each sceario. 3.2 Problems of ucertaity I the previous sectio, the scearios of collapse of cost pla, quality pla, staff pla, ad delivery pla were eumerated. Also, the scearios with high evet probability to fail were able to be extracted. However, these evet probabilities of the mai collapse are oly qualitatively presumed, i.e., they have ucertaity. Therefore, this paper attempts to examie the ucertaity of the accidet scearios of the collapse of cost pla, quality pla, staff pla, ad delivery pla. 4 A Example of the Applicatio of PRA to PM I this paper, PRA is regarded as a approach to risk maagemet, a kowledge area of PMBOK, ad the focus hereafter is o cost pla collapse. I this case, from the evet trees, 10 scearios ca be idetified as is show i Figure 1. I discussig the applicatio of PRA to PM, the case that eough cost could ot be obtaied ad thereby the cost pla collapsed is take as a example. The merit of the applicatio of PRA is that it eables to draw up the scearios that lead to the accidet. The defiitio of the accidet is the collapse of the cost pla, ad that of the iitiatig evet is what causes the collapse of the cost pla.
520 Proceedigs of the 8th Iteratioal Coferece o Iovatio & Maagemet coditio1 coditio2 coditio3 coditio4 coditio5 iitiatig evet The i-house Bugs i the I order to correct The egotiatio The budget desig documet was geerated file the bugs, with the cliet was ot haded to the cliet, who orally accepted the desig. However, o documeted ispectio acceptace was obtaied from the cliet. were detected by the system test.(the detectio model based o Heirich's law) modificatio program had to be geerated. As a result, icremetal ma-hour is required broke dow due secured to the additioal eough fee, the delayed delivery, ad the additioal staff Frequecy result sceario# y0.85 0.575586 fail 1 y0.9 P11:0.15 success 2 y0.9 P10:0.1 success 3 y0.88 y0.95 P8:0.1 success 4 P7:0.12 success 5 iitiatig evet project cost pla collapse 0.5 0.95 0.0188 fail 6 P6:0.05 success 7 0.9 0.88 P5:0.5 success 8 0.3 P3:0.1 success 9 success 10 Figure 1 ET_Cost Pla Collapse 5 Scearios of the Cost Pla Collapse The course of the cost pla collapse is stated i 2.1, 1) to 5). I discussig the applicatio of PRA, let us take the case that eough cost could ot be obtaied ad thereby the cost pla collapsed. By illustratig the course of cost pla collapse with evet trees, 10 scearios ca be idetified from the evet trees i Figure 1. The probability that the i-house desig is delivered to the cliet, ad the documeted ispectio acceptace is obtaied from the cliet is supposed 0.05 (P1). The probability that the bugs i the geerated file are ot detected is supposed 0.12 (P2, P7). The probability that o correctio program, icremetal ma-hour, or icremetal cost is ecessary is supposed 0.1 (P3, P8). The probability that the egotiatio with the cliet cocerig icremetal fee, extedig delivery due, ad icremetal staff does ot break dow is supposed 0.1 (P5, P10). Fially, the probability that the cotigecy budget is ot secured is supposed 0.05 (P6, P11). Sceario 1 The i-house desig was delivered to the cliet, who agreed with it, but the ispectio acceptace documet was ot received from the cliet. I the system test, bugs of the geerated file were detected. Icremet of correctio program, icremetal ma-hour, ad icremetal cost were ecessary to complete the project. However, the egotiatio with the cliet with regard to icremetal fee, extedig delivery due, ad icremetal staff broke dow. The cotigecy budget is ot secured. Sceario 2 The i-house desig was delivered to the cliet, who agreed with it, but the ispectio acceptace
Proceedigs of the 8th Iteratioal Coferece o Iovatio & Maagemet 521 documet was ot received from the cliet. I the system test, bugs of the geerated file were detected. Icremet of correctio program, icremetal ma-hour, ad icremetal cost were ecessary to complete the project. The egotiatio with the cliet with regard to icremetal fee, extedig delivery due, ad icremetal staff broke dow. However, the cotigecy budget is secured. Sceario 3 The i-house desig was delivered to the cliet, who agreed with it, but the ispectio acceptace documet was ot received from the cliet. I the system test, bugs of the geerated file were detected. Icremet of correctio program, icremetal ma-hour, ad icremetal cost were ecessary to complete the project. However, the egotiatio with the cliet with regard to icremetal fee, extedig delivery due, ad icremetal staff did ot break dow. Sceario 4 The i-house desig was delivered to the cliet, who agreed with it, but the ispectio acceptace documet was ot received from the cliet. I the system test, bugs of the geerated file were detected. However, icremet of correctio program, icremetal ma-hour, ad icremetal cost were ot ecessary to complete the project. Sceario 5 The i-house desig was delivered to the cliet, who agreed with it, but the ispectio acceptace documet was ot received from the cliet. I the system test, bugs of the geerated file were ot detected. Sceario 6 The i-house desig was delivered to the cliet, who agreed with it, ad the ispectio acceptace documet was received from the cliet. I the system test, bugs of the geerated file were detected. Icremet of correctio program, icremetal ma-hour, ad icremetal cost were ecessary to complete the project. The egotiatio with the cliet with regard to icremetal fee, extedig delivery due, ad icremetal staff broke dow. Moreover, the cotigecy budget is ot secured. Sceario 7 The i-house desig was delivered to the cliet, who agreed with it, ad the ispectio acceptace documet was received from the cliet. I the system test, bugs of the geerated file were detected. Icremet of correctio program, icremetal ma-hour, ad icremetal cost were ecessary to complete the project. The egotiatio with the cliet with regard to icremetal fee, extedig delivery due, ad icremetal staff broke dow. However, the cotigecy budget is secured. Sceario 8 The i-house desig was delivered to the cliet, who agreed with it, ad the ispectio acceptace documet was received from the cliet. I the system test, bugs of the geerated file were detected. Icremet of correctio program, icremetal ma-hour, ad icremetal cost were ecessary to complete the project. However, the egotiatio with the cliet with regard to icremetal fee, extedig delivery due, ad icremetal staff did ot break dow. Sceario 9 The i-house desig was delivered to the cliet, who agreed with it, ad the ispectio acceptace documet was received from the cliet. I the system test, bugs of the geerated file were detected. However, icremet of correctio program, icremetal ma-hour, ad icremetal cost were ot ecessary. Sceario 10 The i-house desig was delivered to the cliet, who agreed with it, ad the ispectio acceptace documet was received from the cliet. I the system test, o bugs of the geerated file were detected. Now, focusig o Sceario 1, whose appearace probability would be the highest due to the fail, let us geerate ormal radom umbers of the appearace probabilities of pla collapse coditio 1, 3, 4, ad 5 i Sectio 2.1 of this paper, o the coditio that the probabilities take ormal distributio. As for coditio 1, ormal radom umber is geerated with 0.95 averages ad 0.01 SD. As for coditio 3, 4, 5, ormal radom umbers are geerated with averages of 0.9, 0.9, ad 0.85 ad with SD of 0.02, 0.02, ad o.1 respectively. As for coditio 2, adoptig the detectio model based o Heirich's law, detectio rate of 0.88 uder the static coditio is applied. The occurrece probability of Sceario 1 is give by the equatio (I) blow. (Sceario 1 occurrece probability) = ( coditio 1 occurrece probability) X ( coditio 2 occurrece probability) X ( coditio 3 occurrece probability) X ( coditio 4 occurrece probability) X ( coditio 5 occurrece probability) (I) I order to obtai the occurrece probability of Sceario 1, the simulatio to geerate ormal
522 Proceedigs of the 8th Iteratioal Coferece o Iovatio & Maagemet radom umbers was repeated 1,000 times. As a result, the probability distributio histogram illustrated i Figure 2 was obtaied, while the qualitative estimate value of the occurrece probability of Sceario 1 is 0.575. The histogram i Figure 2 shows that the occurrece probability falls i the iterval betwee 0.563 ad 0.623 with high frequecy. Therefore, it ca be cosidered that the qualitative estimate value of the occurrece probability of Sceario 1 is withi the valid iterval. As for the probability that this occurrece probability is below 0.573, the result of the calculatio of cumulative of this histogram from 0 to 0.573 is 0.443. This meas that the occurrece probability that the qualitative estimate value is below 0.575 is aroud 0.443. Next, i the simulatio that the occurrece probability of coditio 1 is 0.8 ad the occurrece probabilities of coditio 2, 3, 4, ad 5 are the same as those i Figure 2, probability distributio histogram illustrated i Figure 3 was obtaied. Although the qualitative estimate value is 0.485, the results of the simulatio fall i the iterval betwee 0.444 ad 0.484 with high frequecy. Thus, it ca be cosidered that the qualitative estimate value ted to be rather over estimate. 40 30 20 10 0 Histogram 0.373772524 0.408772524 0.443772524 0.478772524 0.513772524 0.548772524 0.583772524 0.618772524 0.653772524 0.688772524 0.723772524 0.758772524 0.793772524 Figure 2 Occurrece Probabilities Distributio of Sceario1(case1) Histogram 80 60 40 20 0 0.244404309 0.304404309 0.364404309 0.424404309 0.484404309 0.544404309 0.604404309 0.664404309 0.724404309 0.784404309 0.844404309 Figure 3 Occurrece Probabilities Distributio of Sceario1(case2) 5.1 The detectio model So far, detectio models focusig o huma kietic ad static states have bee developed ad improved [2]. As a result, aroud 88% detectio rate has bee obtaied for static state, while aroud 50% detectio rate for simple kietic state [2]. I this paper, the 0.88 detectio rate for static state is applied to the detectio of the bugs i the file geerated for the system test. 5.2 Applicatio to cost pla makig The resulted figures of this simulatio ca be used i the calculatio of cotigecy reserve. I copig with the pealty article i the cotract o a success-fee basis, it is possible to calculate the cotigecy reserve by usig the resulted figures of this simulatio. Cotigecy reserve is give by the equatio below: (Cotigecy reserve) = (cost for the threat risk) X (occurrece probability of the risk) Whe the cotigecy budget for the occurrece of Sceario 1 is 5 millio ye, i the case 1 i Figure 2, the cotigecy reserve is idicated as follows: 5,000,000 X 0.563 cotigecy reserve 5,000,000 X 0.623 2,815,000 cotigecy reserve 3,115,000
Proceedigs of the 8th Iteratioal Coferece o Iovatio & Maagemet 523 That is to say, the cotigecy reserve would be betwee \2,815,000 ad \3,115,000. With the qualitative estimate value 0.575, cotigecy reserve is calculated oly uiquely as 5,000,000 X 0.575 = 2,875,000 (ye). However, with the use of the simulatio, the cotigecy reserve ca be obtaied more specifically, i.e., betwee \2,815,000 ad \3,115,000. 6 Coclusios As a result of this study, by employig evet trees, the sceario with high occurrece probabilities of mai failure i the case of cost pla collapse was able to be extracted. By geeratig ormal radom umbers agaist cost pla collapse coditios 1, 3, 4, ad 5 (Sectio 2.1) of this sceario, ad by applyig the detectio rate uder the static coditio i Heirich's law (0.88) to coditio 2, histograms of probability distributio were obtaied. Furthermore, the occurrece probability of the sceario, which had bee qualitatively ad uiquely estimated, was able to be aalyzed ad assessed multilaterally because the frequecy of occurrece probability of each coditio was obtaied by the simulatio. I additio, calculatio of cotigecy reserve was able to be available. O the other had, project maagemet is the field where experiece is respected, ad thus computer support is difficult. By accumulatig the achievemet experiece of the vetera PM, the tedecy of each idustry applicatio ca be grasped, which could be a strog support to future PM work. Therefore, it is our future subject to promote improvemet i the efficiecy ad quality of project maagemet work. Ackowledgemet I would like to express my deep appreciatio to the various suggestios from the project members ad other people cocered, without which this study could ot have bee completed. Refereces [1] N. Satoh, H. Kumamoto, N. Ohta, Iformatio Security from the Poit of View of Quatitative Risk Assessmet. Promac2010, 2010:594-602 [2] K. Matsuoka. Achievemet Report 2001, Techological Developmet of Life Eviromet Suitable for Huma Behavior, AIST, 2001 [3] Y. Satoh. IT Project Maagemet Practice with PMBOK. SRC, 2009