CGMS450- Chapter 5- Scheduling the Project

Transcription

1 Prject schedule is simply the prject plan in an altered frmat Cnvenient frm fr mnitring and cntrlling prject activities Take n several frms Gantt charts PERT/ CPM netwrks Cnvert a prject plan r WBS int these frmats 5.1 PERT And CPM Netwrks PERT develped by US Navy, Bz-Allen Hamiltn and Lckheed Aircraft CPM develped by Dupnt De Nemurs Inc. When develped there were significant differences PERT used prbabilistic estimates f activity duratins CPM used deterministic estimates but included bth time and cst estimates t allw time/cst trade ffs t be used Bth emplyed netwrks t schedule and display task sequences Identified a critical path f tasks that culd nt be delayed withut delaying the prject Identified activities with slack that culd be smewhat delayed withut extending the time required t cmplete prject Anything ne can d with PERT, they culd d with CPM Traditinal PERT is used less ften than CPM CPM can be used with 3 time estimates We can d things with PERT that were restricted t CPM in lden times The Language Of PERT/ CPM Terms used in discussing PERT/ CPM Activity task r set f tasks required by the prject Use resurces and time Event an identifiable state resulting frm the cmpletin f ne r mre activities Cnsume n resurces r time Befre event is achieved all predecessr activities must be cmplete Milestnes identifiable and ntewrthy events marking significant prgress n the prject Netwrk A diagram f ndes cnnected by directinal arcs that defines the prject and illustrates the technlgical relatinships f all activities Drawn with a start nde n the left an a finish mde n the right Path series f cnnected activities between any 2 events in a netwrk Critical path set f activities n a ath frm the prjects start event t its finish event that, f delayed, will delay the cmpetitin date f the prject Critical time time required t cmplete all activities n the critical path Building The Netwrk 2 ways displaying the prject netwrk Depicts the activities as arrws and events in ndes Activity-n-arrw (AOA) netwrk usually assciated with PERT Create an activity-n-ndes (AON) netwrk by shwing each task as a nde and linking the ndes t arrws that shw their technlgical relatinship Usually assciated with CPM Easy t draw mre then are mre difficult t draw by hand Often d nt shw events but it is simple enugh t add them by shwing the event exactly as if it were an activity but with zer time duratin and n resurces Avids lines crssing each ther Dummy activity used in situatins where 2 activities have the same starting and finishing ndes r where a single activity cnnects 2 r mre ndes Require n time r resurces Prblem: difficult t distinguish the tasks frm ne anther 1

2 Slutin: add an extra ending nde fr ne f the tasks and then draw a dummy task frm the new nde t the previusly shared nde ensures that the tasks have unique identifies while at the same time maintaining the crrect technlgical precedence relatinship Finding The Critical Path And Critical Time We can add infrmatin t the ndes in the netwrks Abve each nde earliest start time (ES) and earliest finish time (EF) Belw each nde latest start time (LS) and latest finish time (LF) All activities and thus all paths must be cmpleted t finish the prject Shrtest time fr cmpletin f the netwrk is equal t the lngest path thrugh the netwrk Frward pass ES ad EF is fund fr each activity by beginning at the start nde and mving frm left t right thrugh the netwrk calculating as we g frm nde t nde Backward pass t calculate LS and LF we begin by assuming that we wuld like t cmplete the prject within the critical time identified in the frward pass Calculating Activity Slack Within limits if activities n the critical path cannt be delayed withut causing the entire prject t be delayed, it fllws that activities nt n the critical path can be delayed withut delaying the prject Slack r flat- the amunt f time a nncritical task can be delayed withut delaying the prject Slack= LS-ES=LF-EF Any task n the critical path LS must be the same is EF therefre 0 slack If it finishes later than EF, the activity will be late, causing the prject t delay in the prject Equally true fr its LS and ES Assumptins When calculating slack fr a set f activities n a nncritical path, the calculatin fr any given activity assumes that n ther activity n the same path uses any f the slack Once activity s underway, if a predecessr activity uses cme if its slack, its EF is adjusted accrdingly and the ES f successr activities must be crrected Critical time fr the prject is als the prjects due date it is nt cmmn fr a prject t have prject slack Milestnes may be added t the display quite easily: add desired milestne event as a nde with zer duratin ES=EF, LF=LS Immediate successrs f the activities that result in results Cmmn t shw actual dates fr EF, ES, LS, LF PM primary attentin must be paid t activities n the critical path Ding It The Easy Way Micrsft Prject (MSP) See pages fr steps in creating netwrks with MSP Free slack the activity can be delayed withut affecting the start time f any successr activity Ttal slack= LF-EF r LS-ES 2

3 5.2 Prject Uncertainties And Risk Management Calculating Prbabilistic Activity Times All pssible duratins fr sme tasks can be represented by a statistical distributin Optimistic (a) estimates fr a task duratin, a, such that the actual duratin f the task will be a r lwer less than 1% f the time Pessimistic (b) estimatin duratin fr the same task such that the actual finish time will be b r greater less than 1% f the time Mst likely r nrmal duratin (m) Mean f distributin- expected time T E=(a+4m+b)/ 6 Apprximatin f the mean f a beta distributin Beta distributin far mre flexible than the mre cmmn nrmal distributin Mre accurately reflects actual time and cst utcme Calculatin weighted average f 3 time estimate, a,m,b, using weights f Standard deviatin σ=(b-a)/6 6 is nt a weighted average but rather an assumptin that the range f the distributin cvers 6 standard deviatins (6σ) Variance f this distributin is estimated as Var=σ 2 =((b-a)/6) 2 Range f distributin, b-a, cvers 6 standard deviatins is imprtant Assumes that the estimate actually attempted t judge a and b s that 99.7$ f all cases were greater than a and less than b Less then 1% lay utside f these estimates Estimatrs are nt s uncmfrtable making estimates at the 90-95% levels a is estimated s that 5-10% f all cases are less than a and 5-10% are greater than b These levels d nt cver 6σ instead we use 95% level σ=(b-a)/3.3 90% level σ=(b-a)/2.6 The Prbabilistic Netwrk, An Example Lk at page fr further explanatin f the prbabilistic netwrk Once Mre The Easy Way MSP can easily handle the prbabilistic netwrk des nt d sme calculatins that we demnstrated Calculatins can be easily dne by excel The stchastic (synnym fr prbabilistic) netwrk used fr the preceding discussin is shwn as a prduct f MSP The Prbability Of Cmpleting The Prject On Time E.g. what is the prbability that a prject will be cmpleted 50 days r less Answered with the infrmatin available cncerning the level f uncertainty fr the varius prject activities There is an assumptin that shuld be nted Individual variances f the activities in a series f activities may be summed t find the variance f the set f activities n the path itself if the varius activities in the set are statistically independent E.g. if a is a predecessr f b and if a is early r late, it will nt effect the duratin f b Times when assumptin f statistical independence is nt met Re-estimating the duratin f tasks This shuld be dne anytime the resurces supplied t a prject are different frm thse presumed then he duratin f a prject activities was riginally estimated 3

4 T cmplete a prject by a specific time requires that all the paths in the prjects netwrk by cmpleted by a specified time Determining the prbability tat a prject is cmpleted by a specific time requires calculating the prbability that all paths are finished by a specific time D= the desired prject cmpletin time μ= the sum f the T E activities n the path being investigated σ 2 μ= the variance f the path being cnsidered (the sum f the variances f the activities n the path Simplify the task calculating the prbability that a prject is cmpleted by a specific time practical purpses it is reasnable t cnsider nly thse paths whse expected cmpletin time have a reasnable chance f being greater than the specified time Calculate the prbability that a prject will take lnger than any specified time first calculate the prbability that it will take less than the specified time Subtracting hat prbability frm 1 gives us the prbability that it will take lnger than the specified time Excel s NORMDIST functin used as an alternative fr calculating the prbability that a path finished by a specified time, D Syntax f this functin = NORMDIST (D,μ,σ μ, TRUE) Selecting Risk And Finding D Client asked when the prject utput will be available Wrking this prblem backwards Syntax f this functin = NORMDIST (prbability, μ,σ μ) The Case Of The Unreasnable Bss PAGE N negative Zs indicates that D is less than μ The Prblem With Mergers Page Simulatin Different apprach t risk Build n prbabilistic fundatin established in the previus sectin and use simulatin t handle the arithmetic as well helps us understand the cnsequences f uncertainty Simulatin analysis can prvide insight int the range and distributin f prject cmpletin times Simulate a prject instructins fr excel pages Simple mdel fr a cmplex prblem Hardest part f cnstructing this mdel is t identify all paths thrugh the netwrk Small prblems are easier Traditinal Statistic VS Simulatin Difficulty making sure that the risk analyst understands risk analysis Irrespective f hw the arithmetic is perfrmed, by human r machine, the analyst must understand the nature f the calculatins what they mean and what they dn t mean Examining many paths thrugh a large netwrk and finding thse that may turn ur t be critical r near-critical verwhelming task Sme cmmn issues Bth prcedures assume that task duratins are statistically independent May nt be the case when resurces are shared acrss tasks but prblems can be handled by re-estimating task duratins based n altered resurces Bth prcedures assume that the paths are independent even when paths have tasks in cmmn, the cmmn task duratins are the same fr every path in which they are an element during any given estimate f path duratin by whatever prcedure As if a cnsent is being added t each path Simulatin has a direct way f circumventing the assumptin f statistical independence if the assumptin is nt realistic 4

5 Simulatin ne simply includes the activity r path dependencies as part f the mdel Dependencies are mdeled by expressing the functinal relatinship between the activities paths alng with its distributin, mean, and variance becmes variable in the simulatin Statistical prcess analyst must find the T E and variance fr each path Simulatin cmputers selects a sample rm the distributin f activities fr each activity and then calculates the path duratin fr each path enumerated N matter what methd is used rarely necessary t evaluate every path carefully Large netwrk many paths will have bth srt duratins and lw variances when cmpare t high duratin paths Even when it is technically pssible fr ne f the shrt paths t be critical ften unlikely PM will discver the duratin f each activity when the activity is cmplete after the fact, regardless f the methd used t estimate and calculate prject duratin Cannt knw till the prject is dne Which f the paths will take the lngest time Which path will be critical Hw much slack the ther paths will have Can nly have reasnable estimates Simulatin Recmmend as preferred tl fr risk analysis nly after the analysis understands the underlying thery f the analysis Crystal Ball user friendly sftware that can be used t facilitate and enhance the simulatin f the prject netwrks as well as simulate a wide variety f ther types f prblems Facilitates the task d selecting the best distributins t be used t mdel alternative activities if histrical data are available fr these activities Can determine the best distributin t se t characterize prject cmpletin times and ther utputs f the simulatin analysis Cmprehensive set f prbability distributins available and the selectin f these distributins s facilitated by graphically shwing the analyst the shape f the distributin based n parameters specified Display results in variety f frmats frequency charts, cumulative frequency charts, reverse cumulative frequency charts Prvides all relevant descriptive statistics Quick ability t calculate the prbability assciated with varius utcmes E.g. The prbability that a prject can be cmpleted by a specified time Immediately assess the impact that alternative parameter settings have the shape f the distributin 5.4 The Gantt Chart Henry Gantt develped the Gantt chard arund 1917 Displays prject activities as bars measured against a hrizntal time scale mst ppular way f exhibiting sets d related activities in the frm f schedules The Chart Simple Gantt chart Task names are usually descriptive Each task shws its name WBS number r ID number in rder t identify predecessrs 5

6 Prblems understanding chart Several tasks begin at the same time and have the same duratin If ne such task is n the critical path and the thers nt it may be difficult t find the critical path Even with sftware aid technical dependencies are harder t see n a Gantt chart than n a PERT/CPM netwrk Sftware such a MSP makes it easy t use Gantt chart r netwrk t view critical tasks and paths f prject One can experiment what if with the prject schedule immediately bserving results f experiments n screen Simple t enter alternate time estimates and instantly see the impact n the prject duratin Infrmatin can be added t Gantt charts withut making them difficult t read Symbls can indicate an event Milestne symbls, diamnds, in MSP are added t the charts with different shading r clurs t differentiate between schedule and cmpleted milestnes Advantages Easy t read Decrate the walls f the ffice Updated easily Anyne interested can read a Gantt chart, with little t n training with little n r n technical knwledge f the prject Weakness T interpret beynd a simplistic level what appears n the chart r t alter the prjects curse may require an intimate knwledge f the prjects technlgy Nt necessarily visible n the chart but available n the netwrk r the prject plan Deceptive in its apparent simplicity One must be cautius abut publicly displaying Gantt charts that include activity slack r LS and LF Members f the prject team may be tempted t prcrastinate and tackle the wrk based n LS r LF This makes the critical path ut f a nncritical path becme immediate surce f headaches fr the PM Lse the ability t reschedule the resurces used by tasks that nce had slack Device in mnitring a prject and/ r cmmunicating infrmatin n its current state t thers Nt adequate replacements fr netwrks Cmplementary scheduling and cntrl devices 6

7 5.5 Extensins T PERT/CPM Precedence Diagramming PERT/CPM netwrk methd desn t allw fr leads and lags between 2 activities withut greatly increasing the number f sub-activities t accunt fr this Finish t start activity 2 must nt start befre activity 1 had been cmpleted Typical arrangement f an activity and its predecessr Start t start Activity 5 cannt begin until activity 4 has been underway fr at least 2 days Finish t finish Activity 7 must be cmplete at lease 1 day befre activity 8 is cmpleted Start t finish activity 11 cannt be cmpleted befre 7 days after the start f activity 10 Relatinship is rare there are usually simpler ways t map the required relatinship AON netwrk methd that easily allws fr these leads and lags within the netwrk MSP handles leads and lags withut a prblem Netwrk ndes times are calculated in a manner similar t PERT/CPM times lead and lag restrictins- it is ften helpful t lay ut a Gantt cart t see what is actually happening Final Thughts On The Use Of These Tls Class ntes Prducing a schedule... create 1. Wrk breakdwn structure 2. Lgic diagram a. Relatinship between wrk effrts b. Reality peple skip this step 3. Gantt chart Cntingency time/ funds What is it? What yu plan if smething ges wrng Why is it imprtant? Creates cntrl Hw is it represented? Padded add extra time t each activity Reserve cntingency task Adds extra time at the end in a separate task Buffer represented as a arrw, flat technique Desn t shw as an activity Allw fr cntingency Dn t schedule 100% f yu time r allcate 100% f yur budget Use buffers f mney and time t deal with uncertainty Absrbs minr prblems leading up t key deliverables Avid using padding Differentiate between prject nice t have, must d deliverables 7