critical chain method

Critical Chain Method

These sides and note were prepared using 1. The book Streamlined: Building Lean Supply Chain Using the Theory of Constraints. Srinivasan. McGraw-Hell.2. The slides originally prepared by Professor M. M. Srinivasan.3. Presentation of Dr. Daniel Walsh, Vector Strategies. Continuous Process Improvement Summit. 2012, Northridge, CA.3. Presentation of Dr. Rob Richards, Stottler Henke. Continuous Process Improvement Summit. 2012, Northridge, CA.

Complex solutions do not work, the more complex the problem

the simpler the solution must be.Eli Goldratt.

2Ardavan Asef-Vaziri March-2012Critical Chain Method- Basics

Predicting the Future

It is tough to make perdition, especially about the future. Yogi (Lawrence P.) Berra, Baseball catcher and manager, 1925.

The best way to predict the future is to create it. Peter F. Druker.

Writer and professor, 1909-2005.


Motivation


Unavailability of resources.Not high enough prioritiesMurphy Law……..….They ran out of time

Why Projects are Late


Courtesy of Dr. Daniel Walsh, Vector Strategies.

Wrong Metrics - Gedanken


Project Management: Problems Usually Faced

Task durations are highly variable. A lot of uncertainty involved in estimating task durations

Project is not clearly defined : Known Work + Known Unknown Work + Unknown Work.

Existing project work is not complete before new projects shift priorities leading to multi-tasking.

Problems in a project cascade into another project. Constant pressure to increase staff for peak loads.


Task Duration

An activity is composed of three types of work: Known Work + Known Unknown Work + Unknown Work.

Due to unknowns, task durations are highly variable. People usually give a number that they expect to have a 10% chance or less of missing. There is 90% or more probability that the duration is less than the given number. Inflated task times reduce productivity and unnecessarily inflate budgets.

CCPM motivates people to give a number they have a high probability (50% or more) of missing. The idea is to pool task buffers and allow the tasks to use the buffer only if they need it.


Task Duration

Task 2Task 1 Task 3

Task 2Task 1 Task 3

The buffer for the critical chain is smaller than the sum of the buffers required for all the tasks on the critical chain. (i)Protecting task times with buffers will degrade on-time performance, (ii)Providing specific due date to complete each specific task is against a systematic approach, and (iii)Aggressive-but-possible task times increase on-time completion of the projects.


If each task has mean of 15 and StdDev of 5. With 95% probability each task is completed in 25 days. Project 50 days.

But Variance of the critical path is 25+25=50, StsDev of = 7.1We can have duration of 15 for each task and a 14.2 days project buffer. With 95% probability, the project is complete in 44.2 days.

Sequential Tasks

Task 1 Task 2

Task 1 Task 2 Buffer


How a Task is Done


Task Duration Syndromes

Parkinson’s syndrome : Work expands to fill the time available. People tend to continue working on a task that could have been completed earlier if they are given a pre-specified completion time.

Continue to Polish syndrome (aka: the 3-Minute Egg Rule): It’s not quality if it’s finished before time is up.

Student Syndrome: When people feel there is plenty of time to complete a task, other things become important and they procrastinate on the task.


Managing Projects Under Uncertainty

Task 1 Task 25 25

Task 2

Task 1

Task 3

Task 1

Task 2

Task 3

Task 4Task 5


Consider a simple project with 2 tasks performed by 2 different operators.

Sequential Tasks

Task 1 Task 2

5 25

Assume task durations are uniformly distributed (5,25). On average 15 days.

What is average project completion time?

Generate 1000 Projects

Task1 Task2 Project8.8 20.5 29.3

12.2 14.2 26.417.0 7.0 24.023.5 16.0 39.58.6 19.9 28.5

20.4 13.3 33.711.1 5.1 16.115.5 18.6 34.122.9 6.7 29.614.5 9.3 23.821.6 23.6 45.218.5 9.3 27.816.8 18.8 35.617.3 10.1 27.46.1 15.8 21.98.5 6.9 15.4

11.4 14.8 26.220.6 7.5 28.16.5 6.8 13.36.5 11.9 18.415.0 15.1 30.05.9 5.8 8.1

AverageStdDev

30 daysStdDev < StdDev1 + StdDev2


Consider a slightly more complex project.

Task 2

Task 1

Task 3

5 25

Merging Nodes

What is the average project completion time?

Task1 Task2 Start3 Task3 Project9.7 7.2 9.7 5.5 15.3

12.5 11.9 12.5 23.4 35.96.0 19.0 19.0 21.9 40.9

14.7 10.7 14.7 22.4 37.017.7 9.5 17.7 7.9 25.624.3 10.0 24.3 8.7 33.023.7 22.6 23.7 16.5 40.214.4 9.3 14.4 10.1 24.516.0 9.6 16.0 22.1 38.27.1 17.3 17.3 7.7 25.1

14.9 22.2 22.2 18.3 40.516.6 20.8 20.8 16.0 36.820.5 21.4 21.4 20.3 41.614.4 5.9 14.4 8.5 23.022.6 15.2 22.6 18.5 41.115.0 6.8 15.0 9.9 24.915.6 11.4 15.6 6.4 22.015.6 5.3 15.6 15.8 31.415.9 16.0 16.0 24.4 40.418.6 11.6 18.6 13.9 32.515.4 15.2 18.6 15.0 33.65.7 5.8 4.6 5.9 7.5

33 daysTask 3 = Max(Task 1, Task2)


33%

Probability of completing project in 30 days?

Merging NodesMean 33.21Standard Error 0.24Median 32.96Mode #N/AStandard Deviation 7.4Sample Variance 55.3Kurtosis -0.5Skewness -0.1Range 36.9Minimum 12.7

Bin Frequency Cumulative %15 3 0.30%18 14 1.70%21 29 4.60%24 63 10.90%27 88 19.70%30 136 33.30%33 136 46.90%36 132 60.10%39 122 72.30%42 128 85.10%45 86 93.70%48 54 99.10%

More 9 100.00%


The affect of resource interdependencies on a simple project.

Task 1

Task 2

Task 3

Task 4Task 5

5 25

Merging Nodes and Resources

What is the average project duration?What is the probability of competing the project in 45 days.


1000 InstancesTask1 Task2 Start3 & 4 Task3 Task4 Complete3 Complete4 Task5 Project

8.0 13.7 13.7 24.7 6.2 38.4 19.9 19.1 57.517.8 24.4 24.4 15.9 20.4 40.4 44.8 18.0 62.910.8 19.6 19.6 23.7 19.8 43.3 39.4 16.6 59.95.7 12.0 12.0 25.0 23.3 36.9 35.3 24.2 61.15.8 5.6 5.8 10.9 22.1 16.7 28.0 16.1 44.1

15.8 11.2 15.8 15.1 14.0 30.9 29.8 15.4 46.38.8 8.6 8.8 24.8 5.5 33.7 14.3 24.9 58.5

14.2 8.9 14.2 22.7 18.5 36.9 32.7 9.4 46.313.8 8.1 13.8 16.1 18.4 29.9 32.2 7.0 39.28.1 12.5 12.5 17.3 21.1 29.9 33.6 12.9 46.5

17.6 12.6 17.6 14.5 10.2 32.0 27.8 18.3 50.315.0 19.5 19.5 15.4 15.4 35.0 34.9 17.6 52.67.1 8.1 8.1 16.6 19.3 24.7 27.4 15.1 42.4

18.4 17.3 18.4 6.7 5.2 25.1 23.6 20.2 45.218.5 14.0 18.5 10.3 7.0 28.8 25.5 16.6 45.410.6 24.1 24.1 20.3 5.5 44.5 29.7 10.8 55.219.5 11.4 19.5 22.1 6.7 41.7 26.3 14.0 55.713.8 14.8 14.8 14.9 7.1 29.7 21.9 15.8 45.520.9 15.6 20.9 18.7 23.9 39.6 44.8 5.4 50.212.4 23.0 23.0 10.3 14.8 33.3 37.8 23.4 61.214.9 14.9 18.3 15.1 14.8 33.3 33.1 15.0 51.65.7 5.8 4.7 5.7 5.9 7.4 7.8 5.7 8.8


1000 Instances

Bin Frequency Cumulative %25 0 0.00%30 4 0.40%35 28 3.20%40 70 10.20%45 128 23.00%50 200 43.00%55 205 63.50%60 185 82.00%65 110 93.00%70 59 98.90%

More 11 100.00%

A project’s most likely completion time is much larger than the sum of the averages of the tasks making up it’s longest path (due to synchronization or due to task dependencies).

So, how do we quote estimated completion time of the project?

Do people give a number that they know has a high (50% or more) chance of missing?


Task and Project Durations – the Traditional Way

So, the average task times are “padded” to accommodate any possible delays. Instead of specifying a 50% time estimate (which fails half the time), a 98% confidence estimate is developed for the tasks and project duration.

98% of [5,25] = 5+.98*20 = 5+19.6 = 24.6

What is the chance the project will complete in 70 days?

5 25

Task 1

Task 2

Task 3

Task 4Task 5


Given CPM computation, the project will take 15 weeks, and all tasks are on the critical path. If the project was scheduled for 20 weeks, there is a 5 week project buffer.

Task Precedence Duration ResourceTask1 - 5 weeks Resource ATask2 - 8 weeks Resource BTask3 Task1 7 weeks Resource BTask4 Task2 4 weeks Resource ATask5 Tasks3 and 4 3 weeks Resource C

T1 T3

T5

T4T2

5

8

7

4

3

S


Project Buffer and Feeding Buffers

T1 T1 T1 T1 T1

T2 T2 T2 T2 T2 T2 T2 T2

T3 T3 T3 T3 T3 T3 T3

T4 T4 T4 T4

T5 T5 T5

a) Compute the project buffer.b) Compute and identify the feeding buffer(s).



Fever Chart


Buffer Management: Allocate Resources to Tasks Based on “Buffer Burn Rate”

Buffer

50% work completed 60% buffer consumed

Chain 2

Burn Rate: % of buffer consumed vs. % of work completed. Automatically calculated on an ongoing basis to assess how much buffer is still available for future uncertainties.

Task Priorities: Tasks that lie on chains with less safety remaining are given top priority. This ensures that buffers are not wasted, and also reduces pressure to multitask.

Buffer

33% work completed 20% buffer consumed

Chain 1


Multi Tasking


•Most heavily loaded shared resource (constraint), determines throughput

•Project starts are based on constraint’s capacity,

•Pressure to multitask also comes down

Most heavily loaded resource

Pipelining

Pipelining is more efficient than starting projects ASAP

Pipelining: Release Projects Based on Constraints Instead of Starting ASAP

Realization Technologies, Inc.


Erroneous assumptions: Protecting task times with buffers will improve on-

time performance (this is a biggie)

By finishing too soon, people hurt their future negotiating power.

Providing milestones for each task is good

It is good to induct work as soon as possible

Multitasking is beneficial

What to Change?


Summary: Sources of Project Delays

Synchronization Delays Integration (assembly) points Resources and tasks

Delays due to Behavioral Effects Parkinson’s Law Student Syndrome Continue to Polish Syndrome

Queuing Delays Induct work ahead of schedule Multitasking


50% chance od failure, not 1%. Avoid the 3 syndrome.

Aggregation, project buffer, feeding buffers

Integrate the resources into the CPM.

Proactive, leading, how much of PB is consumed.

Not how many days have you worked on the task, but how many days do you need to complete the task.

Team work culture at the project level not individualism at task level

Managing by exception, fever chart, what are todays priorities?

Avoid multi-tasking, follow constraint recourse based release.

Avoid inducting work ahead of schedule (ASAP).

Do not switch the Critical Chain

Summary


Courtesy of Dr. Rob Richards, Stottler Henke

critical chain method

Documents

ardavan asefvaziri

task buffers

specific task

unknown work unknown

critical path

existing project work

possible task times

inflated task times