Critical Chain Project
Management
BCS Nottingham & Derby
Winter School 2006
Steven Wray, BA, MBA, MBCS, C Eng
Phone: 07951 727 490
e-mail: steve@wrayassociates.com
web: www.wrayassociates.com
What makes a Project ?
• One end-point
• At least two tasks, linked by dependency
• Significant inherent unpredictability in how long
the tasks will take.
What do we want from Project Management ?
What do we want from Project Management ?
•
Reliable on time in full to budget delivery performance
More revenue, more Profit, happy customers
•
A stable plan
More Productive use of resources
•
Simple, objective measures of Project progress
Shorter meetings, better informed
stakeholders - less waste, more productivity
•
•
Simple, objective measures of Project health status
Shorter meetings, better informed
stakeholders - less waste, more productivity
Clear signals for when corrective action is - and is not - necessary
Better directed recovery efforts - less waste,
more productivity
•
Direction for ongoing improvement efforts
The future brings more revenue, more profit, happier
customers than the present
What is our normal experience from today's
way of Project Management ?
What is our normal experience from today's
way of Project Management ?
• Reliable on time in full to budget delivery performance ?
Or
A continuous struggle with time, cost and scope ?
• A stable plan
Or
Repeated rescheduling ?
• Simple, objective measures of Project progress ?
Or
Clarity at the start and end, thick fog in between ?
• Simple, objective measures of Project health status ?
Or
Subjective assessments compounded by human factors ?
• Clear signals for when corrective action is - and is not - necessary ?
Or
Intervening too much too early, and too little too late ?
• Direction for ongoing improvement efforts ?
Or
"We'll improve our methods when things get better"
'Normal' Practice in the planning phase
• We identify the tasks in the Project and specify the
resources needed for each one
• We allocate to each task sufficient time that we are
confident will allow it to be completed with those
resources. That is, the time the task should take on
average, plus some contingency to give us the
confidence we seek
• We apply task dependencies, and work out the longest
path of tasks in the Project
• The time along this path is the time-line of the Project
Normal practice in the execution phase
• As long as every task completes on time (within its
contingency), its successors will be started on time
• As soon as any task finishes late (outside its contingency),
its successors will start late, and this normally means they
will finish late
• In order to rescue a Project which shows any lateness, we
have to squeeze the remaining tasks in the Project
• Typically we have to compromise on time, cost or scope
and reschedule
The Estimating Dilemma
• If I allow more time for every task in the plan, each task is
less likely to be late, but the Project end date will be
later...
• If I allow less time, the end date will be earlier, but the
Project is more likely to overrun
• BUT I have to deliver the Project on time
How long should we allow in the plan for
a task ?
•
•
•
•
•
Some staff work more quickly than others
Sometimes staff are distracted or interrupted
Sometimes necessary resources are delayed
Some staff are risk-averse in their commitments
Some organisations reinforce risk-aversion
How long do we allow in the plan for a task ?
The average time the task ought to take an average
performer who focuses on it
PLUS The time we expect to be spent on distractions
PLUS A contingency time to take account of:
•
The spread between average and low performers
•
Our uncertainty in the average time
•
Our uncertainty in the time for distractions
•
How risk-averse we are or have to be
Some simple maths
If we set our estimates so that we are 90% sure that any
one task will be completed on time
If we have 20 tasks in our Project
The Probability that all the tasks
will be on time is: 0.920 = 12%
For 50 tasks, the Probability of all on time is: 0.950 = 1%
Emphasising
• Doing things by the text-book, with 20 tasks, each of
which we are 90% sure will complete on time, we have an
88% probability of being late for the Project
• With 50 tasks, its a lot worse: its a 99% probability of
being late for the Project
What if we change our estimating to a 95%
confidence level ?
• NB This will inflate the time for every task maybe by 25 50% because we will need more contingency
• For 20 tasks the probability of being late is now 64% (was
88%)
• For 50 tasks we are late now 92% (was 99%)
We have very much extended our Project time-line, and
increased our chances of success from 12% to 36% for
the small Project, and from 1% to 8% for the large Project
Surely, there's a better way...
• Plan A - invest our energy in reducing the extent of the
variability:
-
Allowing longer in Project planning stage for preparing estimates
Training staff in estimating
Use of formal estimating methods
Measuring progress (CSC tools) and feeding results back into estimating
practice
More detailed specifications
Less flexibility over changes to specifications
Training the staff better in their job content
Using individual performance measures to identify poor performers
Keeping projects short (c 6 months), breaking larger undertakings into
several short Projects
Doing this can help, but doesn't solve the problem
Plan B - Coping behaviours
•
•
•
•
•
•
•
•
•
Project Managers fight to be assigned the most viable Projects
Project Managers fight for the best staff
Project Managers fight to keep the Project scope down
Project Managers exploit changes in scope to unduly extend timelines
and budgets
Project Managers sandbag Project plans to create headroom
Project Managers quit long Projects well before the delivery date
Project Managers disregard targets they know to be impossible
Staff work double shifts in the final weeks / months
Dumping the blame elsewhere
Doing these may help the individual, but not the organisation
Plan C: Approach the problem in a different way
We can reduce variability, but we cannot
eliminate it, because it is inherent to the nature
of a Project
We must manage the variability that remains
How we handle variability in Critical Chain
• We do not build in any contingency at the Task level
• We move all the contingency to the Project level - we call
this the Completion Buffer
Individual Tasks can now be late without affecting the
completion date of the Project
The Project due date is protected as long as
the accumulated lateness along any one
chain is less than the completion buffer
What difference does this make to our
probability of being late ?
Under 'normal' practice, if any task is later than its
contingency allows, we have a problem
Under Critical Chain, we only have a problem if the total
lateness exceeds the total contingency
This second condition is much less likely than the first [ Law
of averages / Central limit theorem] and increasingly so
as the number of tasks increases
The Completion Buffer
• A Buffer is a block of time which protects a deliverable
from being affected by delays upstream. The Completion
Buffer protects the Project completion date
• Over the course of the Project we expect our buffers to
be used up, in proportion to progress made
Scarcity of Resources
• In putting together the plan, we must take into account
scarcity of resources
• In particular, if two tasks want exclusive use of the same
resource, at the same time, they have to be staggered
• This affects the plan in a similar way to the task
dependencies
Scarcity of Resources
Task C depends on both A and B
Each task uses a different resource
Task A – 10 d
Task C – 10 d
Task B – 10 d
Project Time required - 20 d
Scarcity of Resources
Task C depends on both A and B,
Both A and B need exclusive use of the same resource
Task A – 10 d
Task C – 10 d
Resource conflict
Task B – 10 d
Project Time required - 30 d
The Critical Chain
• We identify the longest chain of dependent tasks by
resource through the Project - this is the Critical
Chain, at the end we place the Completion Buffer
The time taken to complete the
Project is the time taken to complete
the Critical Chain
Any delay in the Critical Chain
delays the Project completion
Completion Buffer
Task A
Completion Buffer
Task C
Task B
Committed end date
In Practice
Task A
Task C
Task B
Project duration held constant
Task A
Completion Buffer
Task C
Task B
The buffer is 25-33% of
chain length
Feeding Chains
• All the other chains of tasks we call Feeding Chains ,
because each one at some point feeds into the Critical
Chain
• Every task in the Project is part of either the Critical
Chain or a Feeding Chain
Feeding Buffers
• We must not allow anything to delay the Critical Chain
• We must protect the critical chain from being delayed by
lateness in the Feeding Chains
• We start the feeding chains a little early, and insert a
block of time to decouple the Critical Chain from each
Feeding Chain
• We call these blocks of time Feeding Buffers
Feeding Buffers
Task A
Completion Buffer
Task C
Task B
FB
Task D
Planning Phase Summary
• There is no contingency at task level
• The Project due date is protected by the a block of time
called the Completion Buffer
• The Critical Chain is the longest chain of tasks through
the Project
• All other chains of tasks are Feeding Chains
• We place Feeding Buffers to decouple the Critical Chain
from the feeding chains
Execution phase....
• We have a great plan - what can happen in execution ?
Critical Chain in the Execution phase....
• No multi-tasking - when someone starts a task they
stick to it until it's completed
Bad Multi-tasking
Whenever I put down one task and pick up another, I lose
productive time
How much time is lost switching depends on how deep or
shallow the task in hand is
Putting a stop to multi-tasking in effect creates extra
capacity
More Bad Multi-tasking
If we have two people, each available 50% to our Project,
and they have to work together, they are effectively 25%
available
If we have four such people, it's effectively 6%, so we'll wait
on average 8 working days to get them together for half a
day
Real-life case: Ten team leaders, each on 50% availability:
Out of two Project meetings held, 3 came to both, 5 to
one or the other, 2 to neither
Very Bad Multi-tasking
When someone stops doing a task on the Critical Chain,
and starts doing something else, they are delaying the
entire Project
Critical Chain in the Execution phase....
• No multi-tasking - when someone starts a task they stick
to it until it's completed
• We begin each task as soon as resource is available and
prerequisite tasks are complete
The timings in the plan are for planning,
not a commitment on execution
Critical Chain in the Execution phase....
• No multi-tasking - when someone starts a task they stick
to it until it's completed
• We begin each task as soon as resource is available and
prerequisite tasks are complete
• We finish each task as soon as we can
Early finishes on Critical Chain tasks bring
forward the whole Project
Early finishes on Feeding Chains increase the
protection of the Critical Chain
Sometimes we call these three together 'Roadrunner' style
Summarising Practical differences...
• Planning phase:
– The Project is analysed into the Critical Chain and Feeding
Chains
– Contingency is aggregated into a Completion Buffer
protecting the Project end-date, and Feeding Buffers
protecting the Critical Chain
• Execution Phase
– No Multi-tasking
– Early finish / allowing early start of following tasks
– Subordination to the Critical Chain
Measures and Management
• How do we know how the Project is doing ?
Measures and Management
Making less progress than planned will eat into the Completion buffer
Making more progress than planned will add to the Completion buffer
A
C
B
At day 5, task A has 8 days
remaining (of 10) - Completion
Buffer is eroded by 3 days
D
Task D is completed - no effect
on Completion Buffer
Measures and Management
• Question 1: How many days work until the Project is
completed ?
• Answer = the number of days left on the Critical Chain
It is the time on the Critical Chain that
determines the time required to
complete the Project
Measures and Management
• Question 2: How certain are we about the answer to
Question 1 ?
• Answer = the proportion of the Completion Buffer that we
have left, compared to the proportion of the Critical Chain
still to do
The Completion Buffer protects the end
date. The less (more) it is eroded,
the less (more) the end date is at risk
Corrective Action
We compare the percentage of the Completion Buffer
Remaining (%CBR) with the percentage of the Critical
Chain Remaining (%CCR)
We set trigger points for corrective action, for example:
• When the ratio %CBR / %CCR is 1or more, Project
status is GREEN - Watch
• When %CBR / %CCR is between 1 and 2/3, Project
status is AMBER - Prepare a recovery plan
• When %CBR / %CCR is less than 2/3, Project status is
RED - Implement recovery plan
Measures in 2-D
Completion Buffer Remaining
0%
100 %
100 %
Critical Chain Remaining
0%
Don't overreact to buffer erosion
• We expect our buffer to be used up over the course of
the Project
• Our date is not threatened if the buffer is used up in
proportion to progress
– If we have 2/3 of the completion buffer left and only 1/3 of
the Project to do we are doing fine
• Our date is threatened if the buffer is used up
disproportionately
– If we have 2/3 of the Project still to do but have only 1/3 of
the completion buffer left we have a problem
Measures Summary
• We have simple, objective measure of Project progress
• We have a simple, objective measure of Project health
status
• We have a simple rule for triggering - and not triggering corrective action
We can redefine the Project progress
meeting as the Buffer Management
Meeting
Buffer Management Meeting
Attendees: Project Manager, Project Sponsor / Owner, Task Managers,
Resource Managers
Agenda:
1.
Reminder of what tasks are on the Critical Chain.
2.
Review Project status ( % Critical Chain outstanding).
3.
Review Completion buffer status (Red, Amber, Green).
If necessary, initiate corrective actions.
4.
Review Feeding buffers status (Red, Amber, Green).
If necessary, initiate corrective actions.
5.
Review tasks in progress to ensure earliest completion in full.
6.
Review tasks not started to ensure earliest start where
appropriate.
Using Buffer Management to drive ongoing
improvement
•
•
•
•
•
Buffer Management measures are fact-based and
objective
Buffer Management meetings highlight buffer erosion
/ Project delays
Preventing the causes of delay will speed up your
Projects
Your process of ongoing improvement is simply to
eliminate the causes of delay by following up on the
issues highlighted in Buffer Management meetings
As your Projects run faster and more reliably,
continue to eliminate more and more causes of delay
How Does Critical Chain help our Project
Management ?
Aggregated contingency and 'Roadrunner' style
 Effective due-date protection and extra effective capacity
 More reliable on time in full to budget delivery performance
Identification of Critical Chain and Feeding Chains  Focus
Buffers  Flexibility  A stable plan
Buffer measures
 Objectivity and clarity
 Better focused meetings, better directed recovery efforts, better informed
stakeholders, less waste and more productivity
 Less encouragement of 'coping behaviours‘
Buffer Management
 Focus for ongoing improvement efforts
Without CCPM
• If the plan is not based on the Critical Chain, it may be infeasible
• If the Critical Chain is not known, the Project Manager cannot focus
on it
• Without contingency the plan is not robust
• If contingency is all in the task estimates, it's still not robust it’s just
longer, and there is no sense of urgency
• Without feeding buffers, an early completion will not help the end
date, because the next task has to wait for its other prerequisites
• With multi-tasking and interruptions > 25% capacity is wasted
• Interruptions on Critical or Feeding Chains delay Project Completion
• Measures are backward-looking or misdirect attention
• Improvement processes lack focus
Multi-Project Pipelines
The Goal is to maximise Project throughput over
some timeframe
The constraint is one of the resources available
The pipeline can be overcommitted
Resources can be used on the wrong Project
Projects can delay one another
Multi-Project Pipelines
• Starting point is that each Project has a fully
defined CCPM plan, and
• We have a defined set of global resources
• Assumption that one of the resource constraints
dominates the others over the timeframe (Pacing
Resource)
• Execute a single pipeline plan which is feasible
and robust
Example
We run the engineering department of a manufacturing company making
custom products
Each product ordered is a new variation on a few standard designs
Typically, for each product ordered, we spend 5 days on the specification, 10
days on the design, 2 days on the test plan, and 3 days on the manual
We have 2 people who do specifications, 3 who do designs, one test planner
and one manual-writer
Which is the Pacing Resource ?
How many Products can we deal with in a 50 day quarter ?
Example
We run the engineering department of a manufacturing company making
custom products
Each product ordered is a new variation on a few standard designs
Typically, for each product ordered, we spend 5 days on the specification, 10
days on the design, 2 days on the test plan, and 3 days on the manual
We have 2 people who do specifications, 3 who do designs, one test planner
and one manual-writer
Specification: 2 x 50 = 100/5 gives 20 products
Design: 3 x 50 = 150, 150/10 gives 15 products
Test Plans: 1 x 50 / 2 = 25 products
Manuals: 1 x 50 /3 = 17 products
Design = Pacing Resource
Example
We run the engineering department of a manufacturing company making
custom products
Each product ordered is a new variation on a few standard designs
Typically, for each product ordered, we spend 5 days on the specification, 10
days on the design, 2 days on the test plan, and 3 days on the manual
We have 2 people who do specifications, 3 who do designs, one test planner
and one manual-writer
What is the consequence if:
The test planner is off for two weeks ?
One of the designers is off for two weeks ?
The documentation task for a product overruns by 50% ?
The design task for a product overruns by 50% ?
The specification task for a product overruns by 50% ?
Answers
Loss of output on the pacing resource has a permanent,
pervasive effect
Loss of output on the other resources has a temporary, local
effect
We need to protect the Pacing Resource
Planning the Pipeline
Stagger Projects on the Pacing Resource
Buffer the Pacing Resource
A feasible and robust due-date for each
Project
Priorities in the pipeline
• In order of net profit per hour of time on Pacing
Resource
• In sequence of strategic priority
• Committed Projects first, then new Projects
• Look at the effects of different sequences
Product 793
Schedule
Pacing Resource RB
Schedule
RB
Design
793
Product 794
Schedule
Spec
Design
792
Test plan /
Manual
Design
Spec
RB
Design
794
Design
RB
Test plan /
Manual
Execution Phase
Hierarchy of subordination:
• Never interrupt the Pacing Resource
• Only interrupt a Critical Chain task to start the Pacing
Resource
• Only interrupt a Feeding Chain task for a Critical Chain
task
Measures in 2-D
Completion Buffer Remaining
0%
3
5
2
4
1
100 %
100 %
Critical Chain Remaining
0%
Buffer Management
Management team (Project Managers, Resource Managers)
looks at:
• Pacing Resource schedule
• Resource Buffer status
• Critical Chains vs Completion Buffers picture
• Resources: best resources allocated to most important
Projects ?
• Pacing Resource tasks in Progress - early completions ?
• Pacing Resource tasks due to start - early starts ?
Record buffer erosion to focus improvement effort
Summary - Multi- Project
• Start with all Projects fully-defined in CCPM
• Identify Pacing Resource
• Load the pipeline by staggering Projects on the Pacing
resource
• Buffer the Pacing Resource
• Apply hierarchy of subordination
• Manage by Buffer Management
• Use Buffer Management to drive improvement
Without CCPM
• If we haven't identified the Pacing Resource:
– our pipeline may be infeasible
– we will waste capacity
• Without a resource buffer, the pipeline won't be
robust
• Without good measures, management is
unfocused or wrongly focused
• Without Buffer Management, improvement is
unfocused
In Practice
•
•
•
•
The logic of CCPM works
Many success stories: A50, Lucent, Phillips, US Navy
Implementation is not a trivial undertaking
As CCPM brings Projects under control, the constraint
moves to management
• Aggressive reductions in cycle time are both possible
and necessary
Books
"Critical Chain" by Eli Goldratt 0-88427-153-6
"Critical Chain Project Management" by Lawrence P
Leach ISBN 1-58053-903-3
"Enterprise-Focused Management" by Ted Hutchin
ISBN 0-72772-979-9
Questions ?