Critical Chain Project Management (CCPM) Ron Henry Project Management Journal Club June 3, 2009 1 What is CCPM? Billed as new project management paradigm Very different management approach Term “critical chain” does not fully capture the shift Really a philosophy built around a buzzword (like Six Sigma, TQM, EV) Key Ideas Resource management is central to planning Manage to project finish date, not interim milestones Aggressive (median) activity duration estimates Safety time removed from activities and aggregated into project buffer Focus on critical chain and protect it from delays in non-critical activities Discourage multitasking 2 CCPM Origins Developed by Eliyahu Goldratt Israeli physicist turned business consultant Developed “Theory of Constraints” (TOC) Popularized CC in 1997 with publication of Critical Chain Intellectual Roots Analysis of problems with traditional project management There must be a systemic reason for this Traditional management can create perverse incentives Statistics Same PM problems occur across a wide range of organizations Insight into human nature Most projects fail to finish on time or within budget Less margin is needed to protect project as a whole than to protect each individual task Business administration - CC is application of TOC to project management 3 Definitions and Metaphors Critical Chain vs. Critical Path Critical Path: The path in a network schedule with zero slack (or float) Note that slack only appears in a schedule if non-critical tasks are scheduled early Usually only task dependencies are considered If tasks are scheduled as late as possible, longest path is most critical Critical Chain: The longest chain of consecutively dependent activities, considering both task and resource dependencies Metaphor shift is as important as the technical difference Gets PM thinking about the people involved (links of the chain) A chain is only as strong as its weakest link Project execution as a relay race Different athletes, different skills Each leg should be finished as quickly as possible Coordination between resources to minimize overhead of passing baton 4 Pathologies with Traditional Scheduling (1) Philosophy Brooks’ Law (1975) “Adding manpower to a late software project makes it later” Goldratt makes a similar assertion - Deadline pressure on a task causes it to take longer Why task deadlines are counterproductive Resources held accountable for meeting their estimates have strong incentive to pad those estimates with “safety time” to assure ~90% success probability Once estimates have been padded, perverse incentives arise - Student syndrome – When people think they have extra time on a task, they tend to begin it late (if busy) - Parkinson’s Law – Work expands to fill available time (if less busy) - Murphy’s Law – People tend to underestimate the amount of safety time they need, then overcompensate – No time to cope with unexpected obstacles if safety time has already been used up by starting late or increasing scope Trivial example: commuting to make an early meeting - Ask for meeting to start later, then leave later and may be late anyway 5 Pathologies with Traditional Scheduling (2) Little incentive to finish tasks early Early finish may lead managers to discount future estimates Similar to syndrome of government agencies using up their budgets If a task does finish early, usually little benefit to project - Resources to begin successor task not available until scheduled start date Reasons for bad multitasking Each project manager wants to see some progress, even if small If one has safety time on a task, one may decide to interrupt it to work on something else that may have an earlier deadline Bad multitasking leads to inefficiency Repeated context switching overhead due to loss of focus Tasks not completed, so successor tasks can’t start Large amount of work-in-process (WIP) which hasn’t earned any revenue Upshot Accountability for estimates causes people to waste time, increase scope and work inefficiently Safety time nearly always used up Even padded estimates often exceeded (or met only through heroics) 6 What Makes CCPM Different? Focus Area Traditional Approach CCPM Approach Duration estimation Requested without clear guidelines and often padded Insist on average-case estimates; aggregate safety time into project buffer Planning approach Optimizing Satisficing Handling of uncertainty Planning phase only; estimates treated as deterministic once made Comfortable with uncertainty during both planning and execution Resource leveling Optional technique to improve schedule Integral part of scheduling Resource management Attempt to utilize all resources to optimum level Non-critical resources subordinated to critical ones Scheduling bias Forward from project start; start activities as early as possible Backward from need date; start activities as late as possible Execution and control Manage based on interim milestones Manage based on buffers; only date that matters is project completion Schedule accountability Team members accountable for making realistic estimates and meeting them Team members accountable for overall contribution to project success, not for meeting individual deadlines Project monitoring Backward looking (how much value have I earned?) Forward looking (buffer management, lead time on task completion) 7 Aggregation and Statistics Task durations have a skewed distribution Compare assuming 2-point estimates “Aggressive” “Safe” (50%) (85-95%) Time Suppose project P consists of just two independent tasks A and B P 2 A2 B2 Activity/Path Aggressive Estimate (days) Safety Margin Safe Estimate A 6 3 9 B 8 4 12 P 14 5 19 Conventional project duration: 21 days Little chance of finishing early CCPM project duration: 19 days including 5-day buffer ~10% reduction in lead time (savings increase with length of chain) Real chance to finish early 8 CCPM Planning Methodology 1. Define activities (same as CP) 2. Estimate activity durations (ask for average-case estimates) 3. Build network schedule by working backward from completion date 4. Level resources and resolve any resource contention issues 5. Identify critical chain 6. Add project buffer (up to ~50% of length of critical chain) 7. Insert feeding buffers for non-critical paths where they feed into critical chain 8. Baseline schedule and enter tracking mode 9 Buffer Management Project Buffer Buffer penetration is expected to increase during project Green, red, and yellow zones as function of date within project - If penetration reaches yellow zone, think about contingency plans If penetration reaches red zone, put them into effect Feeding buffers These protect critical chain from a delay in a non-critical activity If feeding buffer is consumed, that just uses up some time in project buffer Resource buffers (shorthand for a coordination system) Resources asked to estimate how much lead time they need before they can get up to speed on a task Monitor expected time to predecessor task completion Notify resource needed for successor task when within lead time of completion Resource then prepares to start task and shifts to interruptible work - Getting ready to “receive baton” Exact time when “baton will be passed” remains uncertain Number of buffers to manage depends on complexity of project 10 Fever Chart Example 11 CCPM and Earned Value Problems with Earned Value Management Deterministic (invests baseline estimates with more certainty than actually exists) Backward-looking Does not distinguish between critical and non-critical activities May be more useful to the sponsor deciding whether to kill a poorly performing project than to the PM trying to fix it EV generally not considered a good fit for CCPM methodology Requires estimated task completion dates, which strictly speaking don’t exist in CCPM Buffer management considered a better technique for monitoring and control - Avoids illusion of precision Less data to collect and analyze Focuses attention on critical chain EV may still have a role, however Collect data to satisfy contractual requirements Provides generic metrics to facilitate assessment of different PM approaches across projects (including CCPM itself) 12 Management of Multiple Projects Using CCPM Applies “Theory of Constraints” to project portfolio management Development organization is a system in itself The system has a goal That goal is to maximize throughput Value( i) Throughput Duration( i) Project i Techniques for Maximizing Throughput Identify critical constraint(s) that are limiting throughput - Typically one or more bottleneck resources Subordinate non-critical resources to critical ones Concept of “drum resource” that is most in demand across projects Keep critical resource busy by ensuring it never has to wait for inputs Everyone marches to beat of drum Avoid forcing resources to multitask - Simple project priority scheme based on “first in” Use capacity buffers to avoid starting projects before resources are available 13 CCPM in Practice Used successfully on many large projects Boeing 777 airframe design project (Johnson 2004) is an example - Pilot project for CCPM at Boeing Involved nearly 1000 people at its peak Significant improvement in EV metrics after CCPM was adopted – Evidence for significant learning curve – Performance improved only after project team learned to use CCPM effectively Some authors claim major project performance improvements for CCPM Wikipedia references 1998 New Zealand study by Balderstone & Madin Study claims a reduction of 69% in lead time from 32 observations However, study was on Theory of Constraints in general, not specific to CCPM Difficult to evaluate this without more research More recent data would be helpful 14 CCPM Problems Practice of requiring resources to give average-case duration estimates is “very controversial” (Lechler et al 2005) Can lead to behavioral problems of its own Claim that Parkinson’s Law and student syndrome are major causes of wasted time in projects is unproven Very difficult to get objective data Requires major change in organization’s PM culture (Patrick) May be difficult to implement piecemeal or as a pilot project if management has not bought in Johnson paper offers some useful lessons on pilot implementations Benefits appear proportional to project size (Johnson) May be best suited for large and complex projects Technical issues Appropriate buffer sizing Possible shifts in critical chain How to identify bottleneck resources across projects 15 Conclusions CCPM is a different set of techniques from traditional PM, built around a different management philosophy Seems to work best in a well-managed organization with the ability to coordinate across projects Need to identify and resolve resource contention problems quickly Things that make CCPM well suited to STScI environment We often know need dates on projects well before they start, so “backward scheduling” seems a natural technique to use Project buffer concept is just an extension of the way NASA manages other key resources by holding reserve at project level - Mass, power, cost ... why not time? Things that may make it less well suited Nonprofit environment Little benefit to finishing projects early How could we define throughput and how would the organization benefit from maximizing it? 16 CCPM Resources Journal papers Stephen Johnson paper on use of CCPM on large Boeing project (2004) Comparison of CC and CP methodologies by Lechler et al (Engineering Management Journal, December 2005) - Provides a more objective perspective Contains an extensive list of references for follow-up study Books by Eliyahu Goldratt (presented as “business novels”) The Goal (1986) Theory of Constraints (1990) Critical Chain (1997) Other books Goldratt’s Theory of Constraints: A Systems Approach for Continuous Improvement (William Dettner, ASQ Quality Press, 1997) Highly recommended in Johnson’s paper Software Commercial software is available to facilitate project management using CCPM - Prochain Project Scheduling from ProChain Solutions Project Scheduler 8 (PS8) from Scitor Corp. (discussed in Patrick material) 17