System Dynamics Transforms Fluor Project and Change Management

Transcription

1 System Dynamics Transforms Fluor Project and Change Management Franz Edelman Award Competition INFORMS Analytics Conference Chicago April 20

2 SYSTEM DYNAMICS TRANSFORMS FLUOR PROJECT AND CHANGE MANAGEMENT Edward Godlewski Vice President, Fluor Energy and Chemicals 2

3 Projects around the world 3

4 Introduction Fluor designs and builds many of the world s most complex projects. Our primary objective: develop, execute, and maintain those capital projects on schedule, within budget, and with operational excellence. One of the biggest challenges throughout the engineering and construction industry is controlling project cost and schedule performance in rapidly changing conditions. We have implemented a simulation model-based system that has transformed our management of projects and changes. We tailor a System Dynamics model to each major project, and use the model to quantify and diagnose the future impacts of changing project conditions and to test ways to avoid those future costs. In doing so over the first 5 years, we have identified over $800 million in savings for our company and for our clients around the world. 4

5 Fluor by the numbers $680 million net income on $22 billion revenue (2009) $26 billion backlog 36,000 employees worldwide Fluor is # in Fortune Engineering, Construction category (# overall) Engineering News-Record (ENR) ranks Fluor # among Top 00 Design- Build Firms and #2 among Top 400 Contractors 5

6 SYSTEM DYNAMICS TRANSFORMS FLUOR PROJECT AND CHANGE MANAGEMENT Peter Oosterveer Group President, Fluor Energy and Chemicals 6

7 Peter Oosterveer, Group President Fluor Energy and Chemicals 7

8 SYSTEM DYNAMICS TRANSFORMS FLUOR PROJECT AND CHANGE MANAGEMENT Gregory Lee Senior Vice President (retired), Fluor Corporation Consultant, Kenneth Cooper Associates 8

9 Challenges Historical context Projects routinely encounter changes from planned conditions scope/design changes, late information, delayed decisions, schedule changes changes that can account for 20-30% of costs. Much of the impact of these changes occurs as secondary impact, which goes by many names (ripple effects, disruption, cumulative impact, productivity loss). Whatever the label, the full cost impacts of changes are typically underestimated, under-recovered, and occur as project surprises later, when difficult to take preventive measures. So what do you get when projects cost millions or 0 s of millions more than planned (and contracted) unhappy clients legal disputes over cost responsibility 9

10 Two project management perspectives regarding change Traditionally: Reactive Accept that change impacts are inevitable, estimate costs after they occur, and present claims to recover added cost. With the new O.R.: Proactive Anticipate and mitigate potential change impacts, to control their future effect on project cost and schedule Scope additions ( increase plant capacity ) and deletions ( take some cost out ) FEED/Preliminary Engineering readiness (when to start Detailed Design) Engineering work schedules (e.g., acceleration) Construction work schedules ( get to the field early ) Supplier delays (data, designs, equipment) Staffing levels ( catch up ) Design change approvals (timing matters) 0

11 A fundamental challenge: Impacts are often widely separated in time and space from the precipitating changes Project X Labor Changes here cause impact here Actual People Original Time

12 More challenges t retrospective what happened, but advance what would happen if When timing of the future impacts Why causation of the future impacts t just construction and scope changes also, engineering changes, information delays, even schedule changes One change 000 Usable broadly, quickly, accurately 2

13 SYSTEM DYNAMICS TRANSFORMS FLUOR PROJECT AND CHANGE MANAGEMENT Kenneth Cooper Managing Principal, Kenneth Cooper Associates 3

14 Three-part analytical solution. System dynamics model of projects 2. Tools for rapidly tailoring 3. Deploying the project models to non-modelers 4

15 System Dynamics basics of the method System Dynamics is a methodology and simulation modeling technique for analyzing complex systems. Rooted in engineering control theory, the method has been used to model and analyze companies, industries, complex projects, and more. The fundamental principles emphasize Cause-Effect affects Overtime The effect of one factor on another may be time-delayed and non-linear, as with the effect of overtime on productivity. Feedback Loops Overtime Feedback among cause-effect relationships can be corrective, or self-reinforcing. Here, more overtime usage increases progress, lowering overtime needs (corrective); but sustained overtime lowers productivity, slows progress, and thus increases the need for more overtime (self-reinforcing). 5

16 System Dynamics basics of the method The strength and timing of cause-effect relationships are described by model equations that can simulate behavior over time, and accurately represent a project plan. Staffing What if experimentation with an accurate model helps identify future impacts, and can test the effect of different actions. Time Staffing Time 6

17 Recall Project X What we need is the ability to foresee and avoid but first we need to understand Actual People Original Time 7

18 So, what happened on Project X WorkToBeDone(t + dt) = WorkToBeDone(t) (t) dt (t) = People(t) (t) WorkDone(t + dt) = WorkDone(t) + (t) dt People Work To Be Done Work Done Our productivity was impacted. 8

19 So, what happened on Project X Work done and done and done Drawings Issued Rev Issues Rev 2 s Rev 3 s 4 s 5 s TIME 9

20 So, what happened on Project X So we added rework to the model: People Quality (0-: fraction not to be reworked) Work To Be Done Work Done Rework 20

21 A better way of looking at projects and secondary impacts People Quality Work To Be Done Work Done Known Rework Rework Discovery Undiscovered Rework 2

22 A story from one project, experienced by many Have you seen any of these next conditions on a project Quality Staff on Project Work To Be Done Work Really Done Known Rework Rework Discovery Undiscovered Rework 22

23 The customer added (+) and changed ( ) work so much, we staffed up more. Scheduled Completion Time Expected Completion Time Quality Work To Be Done Staff on Project Work Really Done Hours Expended to Date Staffing Requested Expected Hours at Completion + & Known Rework Rework Discovery Perceived Undiscovered Rework 23

24 We used lots of overtime and had to hire in tight markets. Hiring Staff Turnover Scheduled Completion Time Expected Completion Time + & Quality Work To Be Done Known Rework Rework Discovery Perceived Overtime Undiscovered Rework Staff on Project Work Really Done Hours Expended to Date Staffing Requested Expected Hours at Completion 24

25 Less skilled new hires also needed more supervision. Hiring Scheduled Completion Time Expected Completion Time + & Skills and Experience Quality Work To Be Done Known Rework Overtime Rework Discovery Perceived Staff Undiscovered Rework Staff on Project Work Really Done Turnover Hours Expended to Date Staffing Requested Expected Hours at Completion 25

26 Rework caused more rework. Hiring Scheduled Completion Time Expected Completion Time + & Out-of-Sequence Work Vendor Design Skills and Experience Engineering Errors Quality Work To Be Done Known Rework Staff Overtime Rework Discovery Perceived Undiscovered Rework Staff on Project Work Really Done Turnover Hours Expended to Date Staffing Requested Expected Hours at Completion 26

27 Under pressure, morale suffered. Hiring Scheduled Completion Time Expected Completion Time Morale Schedule Pressure Out-of-Sequence Work Vendor Design Engineering Errors Overtime Staff on Project Turnover Hours Expended to Date Staffing Requested Quality managers, a story of impacts on productivity, rework, and + & Skills and Experience Known Rework Staff A diagramed version of a story told by thousands of project the interacting conditions that drive those impacts. Work To Be Done Rework Discovery Perceived Undiscovered Rework Work Really Done Expected Hours at Completion These inter-related factors are coded in the model equations, to produce a timestepping simulation of the project. 27

28 Equations: Rework Cycle WorkToBeDone(t + dt) = WorkToBeDone(t) (t) dt WorkDone(t + dt) = WorkDone(t) + Quality(t) (t) dt UndiscoveredRework(t + dt) = UndiscoveredRework(t) + ( Quality(t)) (t) dt ReworkDiscovery(t) dt KnownRework(t + dt) = KnownRework(t) + ReworkDiscovery(t) dt ReworkCorrection(t) dt (t) = StaffOnProject(t) (t) ReworkDiscovery(t) = UndiscoveredRework(t) / ReworkDiscoveryTime(t) 28

29 Equations: (t) = rm Effect on (influence_(t)) Effect on (influence_n(t)) Effect on productivity from each influence is a monotonic function of the given influence. For example, productivity decreases as overtime use increases (see the graph below). Effect on from Overtime 0 Overtime Use 29

30 Equations: Staffing StaffOnProject(t) = Min(StaffRequested(t), StaffAvailable(t)) StaffRequested(t) = (ExpectedHoursAtCompletion(t) HoursExpendedToDate(t)) / (Scheduled Time Remaining(t) rmalstaffworkhours) ExpectedHoursAtCompletion(t) = HoursExpendedToDate(t) / Perceived(t) StaffAvailable(t + dt) = StaffAvailable(t) + (Hiring(t) Turnover(t)) dt Hiring(t) = Max(0, (StaffRequested(t) StaffAvailable(t)) / HiringDelay + ExpectedTurnover(t)) 30

31 Three-part analytical solution. System dynamics model of projects Cause/effect Feedback loops Time delayed and non-linear effects 2. Tools for rapidly tailoring 3. Deploying the project models to non-modelers 3

32 How to get to a tailored project model Hiring Worker Worker Experience Staff Turnover Morale Overtime Staffing Schedule Staff on Quality Requested Scheduled Pressure Project Completion Time Out-of-Sequence Core Hours Expended Work to Date Expected Hours at Work to be Work Really Expected Done Done Model Completion Completion Vendor Time Design Information Known Structure Undiscovered Engineering Rework Rework Rework Revisions Discovery + & Perceived Staffing Project Model & Base Effect of Overtime Several new system steps were required Effects n 32

33 How to get to a tailored project model Hiring Scheduled Completion Time Expected Completion Time + & Worker Morale Schedule Pressure Out-of-Sequence Work Vendor Design Information Surveyed Fluor Business Area & Office-Specific Factors Engineering Revisions Industry- Standard Factors Worker Experience Quality Work to be Done Known Rework Rework Discovery Perceived Staff Overtime Core Staff on Project Work Really Done Model Structure Undiscovered Rework Turnover Staffing Requested Hours Expended to Date Expected Hours at Completion Populated Project Model Hill-Climbing Algorithms Project Model & Base Staffing User Input: Project Teams Conditions User Input: Project s & Schedules Several new system steps were required Effect of Overtime Effects n 33

34 Interface embeds project plans as inputs Industry- Standard Factors Core Core Model Model Structure Structure Hill-Climbing Algorithms Surveyed Fluor Business Area & Office-Specific Factors User Input: Project Teams Conditions Populated Populated Project Project Model Model User Input: Project s & Schedules Project Model & Base PLAN: Start Construction Month 0 22 months duration 2.5M Craft hours 34

35 Interface combines reference standards, global surveys, and project-specific productivity conditions Surveyed Fluor Business Area & Office-Specific Factors Industry- Standard Factors Core Core Model Model Structure Structure Populated Populated Project Project Model Model Hill-Climbing Algorithms Project Model & Base Periodic global surveys update office standard productivity conditions, tapping 40,000 person years of project experience. User Input: Project Teams Conditions User Input: Project s & Schedules Set project specific productivity conditions to match those assumed in project plan (benchmarked against industry reference or Fluor office standard). 35

36 Hill-climbing algorithms guide tailoring of model to simulate the project plan Industry- Standard Factors Surveyed Fluor Business Area & Office-Specific Factors User Input: Project Teams Conditions Core Core Model Model Structure Structure Populated Populated Project Project Model Model Hill-Climbing Algorithms Project Model & Base Start Initialize norm Staffing timing influence_ influence_n SimulatedCum Effort ned Cum Effort Tolerance Deviation Adjust User Input: Project s & Schedules Simulate Simulated PeakTime ned PeakTime Tolerance Deviation ned Max Impacts () SimMax Impacts () Tolerance Deviation ned Max Impacts ( n ) SimMax Impacts ( n ) Tolerance Deviation End 36

37 Tailoring converges model to simulate correctly amounts and timing of staffing, progress, productivity, effects on productivity Start Initialize norm Staffing timing influence_ influence_n SimulatedCum Effort ned Cum Effort Tolerance Deviation Adjust Simulate Simulated PeakTime ned PeakTime Tolerance Deviation Experience ned Max Impacts () Sim Max Impacts () Tolerance Deviation ned Max Impacts ( n ) Sim Max Impacts ( n ) Tolerance Deviation Staffing End 37

38 Tailoring converges model to simulate correctly amounts and timing of staffing, progress, productivity, effects on productivity Start Adjust Initialize norm Staffing timing influence_ influence_n SimulatedCum Effort ned Cum Effort Tolerance Deviation Simulate Simulated PeakTime ned PeakTime Tolerance Deviation Experience ned Max Impacts () Sim Max Impacts () Tolerance Deviation Staffing ned Max Impacts ( n ) Sim Max Impacts ( n ) Tolerance Deviation End 38

39 Tailoring converges model to simulate correctly amounts and timing of staffing, progress, productivity, effects on productivity Start Adjust Initialize norm Staffing timing influence_ influence_n SimulatedCum Effort ned Cum Effort Tolerance Deviation Simulate Simulated PeakTime ned PeakTime Tolerance Deviation Experience ned Max Impacts () Sim Max Impacts () Tolerance Deviation Staffing ned Max Impacts ( n ) Sim Max Impacts ( n ) Tolerance Deviation End 39

40 Tailoring converges model to simulate correctly amounts and timing of staffing, progress, productivity, effects on productivity Start Adjust Initialize norm Staffing timing influence_ influence_n SimulatedCum Effort ned Cum Effort Tolerance Deviation Simulate Simulated PeakTime ned PeakTime Tolerance Deviation Experience ned Max Impacts () Sim Max Impacts () Tolerance Deviation Staffing ned Max Impacts ( n ) Sim Max Impacts ( n ) Tolerance Deviation End 40

41 Tailoring converges model to simulate correctly amounts and timing of staffing, progress, productivity, effects on productivity Start Adjust Initialize norm Staffing timing influence_ influence_n SimulatedCum Effort ned Cum Effort Tolerance Deviation Simulate Simulated PeakTime ned PeakTime Tolerance Deviation Experience ned Max Impacts () Sim Max Impacts () Tolerance Deviation Staffing ned Max Impacts ( n ) Sim Max Impacts ( n ) Tolerance Deviation End 4

42 Tailoring converges model to simulate correctly amounts and timing of staffing, progress, productivity, effects on productivity Start Adjust Initialize norm Staffing timing influence_ influence_n SimulatedCum Effort ned Cum Effort Tolerance Deviation Simulate Simulated PeakTime ned PeakTime Tolerance Deviation Experience ned Max Impacts () Sim Max Impacts () Tolerance Deviation Staffing ned Max Impacts ( n ) Sim Max Impacts ( n ) Tolerance Deviation End 42

43 Tailoring converges model to simulate correctly amounts and timing of staffing, progress, productivity, effects on productivity Start Adjust Initialize norm Staffing timing influence_ influence_n SimulatedCum Effort ned Cum Effort Tolerance Deviation Simulate Simulated PeakTime ned PeakTime Tolerance Deviation Experience ned Max Impacts () Sim Max Impacts () Tolerance Deviation Staffing ned Max Impacts ( n ) Sim Max Impacts ( n ) Tolerance Deviation End 43

44 Tailoring converges model to simulate correctly amounts and timing of staffing, progress, productivity, effects on productivity Start Adjust Initialize norm Staffing timing influence_ influence_n SimulatedCum Effort ned Cum Effort Tolerance Deviation Simulate Simulated PeakTime ned PeakTime Tolerance Deviation Experience ned Max Impacts () Sim Max Impacts () Tolerance Deviation Staffing ned Max Impacts ( n ) Sim Max Impacts ( n ) Tolerance Deviation End 44

45 Tailoring converges model to simulate correctly amounts and timing of staffing, progress, productivity, effects on productivity Start Adjust Initialize norm Staffing timing influence_ influence_n SimulatedCum Effort ned Cum Effort Tolerance Deviation Simulate Simulated PeakTime ned PeakTime Tolerance Deviation Experience ned Max Impacts () Sim Max Impacts () Tolerance Deviation Staffing ned Max Impacts ( n ) Sim Max Impacts ( n ) Tolerance Deviation End 45

46 Tailoring converges model to simulate correctly amounts and timing of staffing, progress, productivity, effects on productivity Start Adjust Initialize norm Staffing timing influence_ influence_n SimulatedCum Effort ned Cum Effort Tolerance Deviation Simulate Simulated PeakTime ned PeakTime Tolerance Deviation Experience ned Max Impacts () Sim Max Impacts () Tolerance Deviation Staffing ned Max Impacts ( n ) Sim Max Impacts ( n ) Tolerance Deviation End 46

47 Tailoring converges model to simulate correctly amounts and timing of staffing, progress, productivity, effects on productivity Start Adjust Initialize norm Staffing timing influence_ influence_n SimulatedCum Effort ned Cum Effort Tolerance Deviation Simulate Simulated PeakTime ned PeakTime Tolerance Deviation Experience ned Max Impacts () Sim Max Impacts () Tolerance Deviation Staffing ned Max Impacts ( n ) Sim Max Impacts ( n ) Tolerance Deviation End 47

48 Three-part analytical solution. System dynamics model of projects Cause/effect Feedback loops Time delayed and non-linear effects 2. Tools for rapidly tailoring Embed company reference and industry standards Embed project plans as part of model input Converge model to simulate correctly amounts and timing of staffing, progress, productivity, effects on productivity 3. Deploying the project models to nonmodelers 48

49 The project model user specifies changes to be tested via the system interface Input foreseeable direct impacts (e.g. added work hours, change timing, design content affected) 49

50 The system provides summary reports and automated cause-effect diagnostics Attribution Attribution of of craft craft productivity productivity and and quality quality impacts impacts % % Late/Changing Engineering Engineering Schedule Schedule Pressure Pressure Late Late materials materials Worker Worker experience experience Construction Construction rework rework Physical Physical overcrowding overcrowding This project cost grows 30% with 2% added hours of direct change impact (.5 secondary impact ratio); changing engineering is the biggest driver of secondary impact in construction People Engineering Labor Overtime Overtime and and worker worker fatigue fatigue Worker Worker morale morale (KC 4 3pm-2base) (KC 4 3pm-2base-5chg) Time 50

51 Benefits Summary 80 Applications to Date Number of applications grew rapidly after the second year Year 5

52 Testing changed conditions on projects may mean changed schedules One real example of benefits Engineering Labor For a client to whom cost was a high priority, what if analyses conducted before the project began anticipated the project performance under different contemplated schedules Equiv heads, Earlier schedule targets require increased labor Time 52

53 Diagnostics help explain why Experience Effect on Eng Time Late vendor info Effect on Eng More hiring in tight markets will dilute experience, reducing productivity Suppliers will not keep up with earlier data needs, hurting in-house productivity Time 80% 70% 60% 50% 40% 30% 20% 0% 0% Engineering revisions will climb in accelerated schedules 0% 20% 30% 40% 50% Faster Schedule And because of the increased late revisions, only 2/3 of the attempted acceleration would be achieved 53

54 Based on the analyses, this cost-conscious client chose a later schedule, to reduce costs % Cost Savings 30% This is the example that Group President Peter Oosterver mentioned the $23 million savings achieved by rescheduling. Dozens of other projects have benefited from the same type of analyses as well Acceleration Costs 20% 0% 0% 0% 20% 30% Faster Schedule Later Schedule Original Schedule $23 Million Savings vs. ned Schedule 40% 54

55 Quantitative benefits: savings achieved for Fluor and our clients bottom-line 00+ Proactive uses of system dynamics model and project simulation system Examples... Scope Additions......(the classic case) Several changes on a mining project added scope and reduced productivity. Model used to analyze and recover $0 million of future impact not otherwise quantifiable. Changes and full cost amount agreed, and a cost-saving schedule adjustment identified, saving an additional $2 million. 55

56 Quantitative benefits: savings achieved for Fluor and our clients bottom-line 00+ Proactive uses of system dynamics model and project simulation system Examples... Preliminary Engineering Readiness......Analyses of tightly scheduled "turnaround" project identified large productivity gains achievable with different overlap of adjacent engineering phases. Work rescheduled accordingly, saving $34 million. 56

57 Quantitative benefits: savings achieved for Fluor and our clients bottom-line 00+ Proactive uses of system dynamics model and project simulation system Examples... Staffing Levels......Client asked Fluor to evaluate cost of accelerating the staffing buildup they requested on engineering a refinery project. Model quantified cost, plan changed accordingly, saving $25 million. 57

58 Quantitative benefits: savings achieved for Fluor and our clients bottom-line 00+ Proactive uses of system dynamics model and project simulation system Examples... Design change approvals......project team redesigned process of reviewing changes to speed their definition and approval, based on model's what-if analyses of mitigation value, saving $0 million. 58

59 Quantitative benefits: savings achieved for Fluor and our clients bottom-line 00+ Proactive uses of system dynamics model and project simulation system Examples... Construction work schedules......small European project was running in parallel with a similar project, both encountering many changes. One project manager employed the model to choose when to "go to the field" (start construction). Manager credits model findings with this project being "much smoother", and saving $ million....on a change-filled project, Fluor's client sought lower cost. Based on model's analyses of change impacts under different schedule scenarios, client implemented a new construction schedule, saving $23 million. 59

60 Quantitative benefits: savings achieved for Fluor and our clients bottom-line 00+ proactive uses of system dynamics model and project simulation system with quantified impact on these and dozens of other projects since 2006 Greater than $840 million in savings Designation by Engineering Construction Risk Institute (ECRI) as Industry Best Practice 60

61 Peter Oosterveer, Group President Fluor Energy and Chemicals Thank you for the opportunity to share with you the Fluor story 6

62 Thank you Thank you for the opportunity to share with you the Fluor story 62