© R. Barnes, 1998
Critical Chain• A fresh look at project management by Eliyahu
Goldratt• Claimed to do for Project Management what
“The Goal” did for Process Mgmt• Reported Successes include
– Israeli Aircraft - Turnaround from 3 months to 2 weeks
– http://www.goldratt.com/tocsuc.htm
– BOS (Midrange Software) - 5 mths cut from critical software project
– http://www.goldratt.com/bos.htm
© R. Barnes, 1998
Critical Chain - the potential
• To replace this project plan, which we probably won’t meet
• with one that is– Shorter– More likely to be met– easier to monitor
through intelligent use of explicit buffersProject, with NO margin
Project, with safety margins hidden
Explicit Buffer
Should complete somewhere within
this
© R. Barnes, 1998
“The Goal” - Review
• 1/ Identify the Goal• 2/ Identify the Constraint (“Herbie”)• 3/ Elevate the Constraint• 4/ Go back to 2.• 5/ “Improvements” elsewhere are
– at best, useless– probably counterproductive
• building inventory
© R. Barnes, 1998
Session[s] Objectives
• Review ideas in “Critical Chain”• Discuss
– Do we agree with Goldratt?– Which of the ideas can we adopt here?
• Action Plan– How?
© R. Barnes, 1998
Critical Chain, Session Outline
• Part 1 - Current Reality– Look at cause-effect relationships in
project management• Part 2 - Critical Chain Scheduling
– Key ideas - achieving “Miracle 1”• Part 3 - Global Viewpoint
– Dealing with Multiple Projects - keeping focus on The Goal
• Part 4 Implementation Issues– Overcoming resistance, making it all
happen
© R. Barnes, 1998
Assumptions
• General Familiarity with Project Management Concepts– Scope, organization– GANTT Charts, PERT, Critical Path, – Resources, Resource levelling– Slack, Early/late finish– Work-in-Progress
• Earned Value (BCWP etc)
• This topic builds on knowledge of these concepts.
© R. Barnes, 1998
Critical Chain• Key concept - A different way of -
– Using project buffers, and – handling uncertainty
• Definition of “Critical Chain”– The set of tasks which determines the overall duration of the
project
© R. Barnes, 1998
Critical Path
A
B
D
C
Critical Path - longest dependent path
A -> B -> D
Assumes flexible resource
Available Manpower
© R. Barnes, 1998
What determines whichorder is best?
Critical Chain - a levelled CP
A
B
D
C
Critical Chain - takes into account resource limits Either A -> B -> C-> Dor A -> C -> B -> D
Available Manpower
© R. Barnes, 1998
Joe’s Story - Reading 1.
• Take the “Reasonable Estimate” and multiply by fudge factor– necessary in order to meet promised delivery
• Work on several things at once– how else to keep busy?
• Both of these are the COMMON PRACTICE
• According to “Critical Chain” they’re BOTH WRONG
© R. Barnes, 1998
Issue 1 - estimates
• “Task will take 5 days”• What does this mean?
– Will take on average 5 days?– 50% probability that it will complete in 5
days?– Almost certainly (80%? 90%?) will complete
in 5 days?
© R. Barnes, 1998
Probability Curve
Where are you going to put
this line?
• Almost certainly estimates contain substantial buffers• Most workers are unaware of this, and can’t tell you “How much
buffer is allowed?”• Even if they could, they’d be too suspicious of your motives to
tell you
© R. Barnes, 1998
Combining Estimates - Parallel Tasks
Job 1(Wkr A)
Job 2(Wkr B)
Job 3(Wkr C)
Job 4(Wkr B)
• Delay in any one gets passed on
• Early finish of others doesn’t help
(No surprises here)
© R. Barnes, 1998
• Each job estimated as 5 days
How long will the project take?
Combining Estimates -Serial Tasks
Job 1(Wkr A)
Job 2(Wkr B)
Job 3(Wkr C)
Job 4(Wkr B)
© R. Barnes, 1998
Job 1(Wkr A)
Job 2(Wkr B)
Job 3(Wkr C)
Job 4(Wkr B)
• The “simple maths” answer (20) is invariably optimistic.
• Experienced Project leaders add their own fudge factors– “5 + 5 = 13”
• Yet we have already seen that each estimate has substantial safety
• Why do we have to add more?
© R. Barnes, 1998
Job 1(Wkr A)
Job 2(Wkr B)
Job 3(Wkr C)
Job 4(Wkr B)
Another Job(Wkr B)
Another Job2(Wkr B)
• If Job 1 finishes early, can Wkr B start on Job 2?• Perhaps - has he finished “Another Job”?• If not, does he know that “AJ” is not critical, and
should be set aside? What if this makes AJ2 critical?
• Result:-– Delays are passed on in full– Advances are usually wasted.
Non-critical Jobs
© R. Barnes, 1998
When should non-critical tasks start?
Job 1(Wkr A)
Job 2(Wkr B)
Another Job(Wkr B)
5 days
2 daysEarly
Late
5 -2 = 3 days slack
Safest - BUT• Loss of Focus (thinking about unimportant task)• Parkinsons Law will apply to AnotherJob (WkrB not busy)
Unnecessary Risk - NowAnother Job is also on theCritical Path!
© R. Barnes, 1998
Parkinson’s Law
Worker must slow downor make work to look busy
Worker is undercommitted Worker must keep busy
Predicted effect?
Reading 2
© R. Barnes, 1998
Current Reality- Reading 3
• Does this describe our situation?
• Is there any way out?– Improvements in one place cause
problems elsewhere.
• We need a miracle!
© R. Barnes, 1998
Critical Chain, Session IISession Outline
• Part 1 - Current Reality– Look at cause-effect relationships in
project management• Part 2 - Critical Chain Scheduling
– Key ideas - achieving “Miracle 1”• Part 3 - Global Viewpoint
– Dealing with Multiple Projects - keeping focus on The Goal
• Part 4 Implementation Issues– Overcoming resistance, making it all
happen
© R. Barnes, 1998
The Miracle• 1. Use an approach to scheduling and logistics that
protects us from Murphy’s Law.
• 2. Have people focus on Global improvements rather than Local ones.
• 3. Have everyone understands and accepts the policies, procedures, and measurements that apply to them.
• 4. Believe that we can make dramatic improvements.
© R. Barnes, 1998
An approach to scheduling and Logistics that protects against
Murphy
What we want What we doMake projects on time Try to make tasks on timeProduce more projects Try to make people more efficientShrink Project Times Try to shrink task timesProjects within Budget Detailed risk analysisCustomer satisfaction Make more detailed specifications
© R. Barnes, 1998
Critical Chain Scheduling
• 1/ Identify the key tasks– (Critical Chain)
• 2/ Exploit performance on the key tasks– focus on them, do everything you can to
make sure they’re not late• 3/ Subordinate to the key tasks
– don’t waste time worrying about the other tasks.
© R. Barnes, 1998
Critical Chain Example
• Reading 4
• How does this differ from conventional Critical Path project scheduling?
© R. Barnes, 1998
Buffers & Managing Uncertainty
• Buffers are – more than a planning tool. – an EXTREMELY VALUABLE monitoring
tool• Reading 5 - “Managing Janet’s Buffers”• Risk Placement• Buffer Types
– How to treat resource buffers
© R. Barnes, 1998
Resolving the WIP Conflict
A (Objective)Maximize Profits
B (Requirement)Quote Short Lead Times
C (Requirement)Give Reliable commitment
dates
D (Prerequisite)Don’t pad task times
D’ (Prerequisite)Pad task times
Conflict!
Arrow AssumptionA<-B Short quoted lead times are important for customersB<-D Safety increases lead times significantlyA<-C Customers care about commitment datesA<-C Profitability depends on customer satisfactionC<-D’ There are statistical fluctuations and unanticipated
problemsC<-D’ We must deal with uncertainty by padding task timesD<-D’ All tasks need safety time
Attackthis link
© R. Barnes, 1998
Resolving the WIP Conflict - 2
A (Objective)Maximize Profits
B (Requirement)Quote Short Lead Times
C (Requirement)Give Reliable commitment
dates
D (Prerequisite)Don’t pad individual
task times
D’ (Prerequisite)Put in aggregate buffers
• Key Concepts– Any conflict can be diagrammed explicitly– There are hidden assumptions behind any conflict that can
be challenged– Conflict between more and less WIP can be resolved
through buffers– Buffers are not optional
© R. Barnes, 1998
Identifying the Critical Chain
• Reading 6• Create initial plan
– Average durations, late as possible• Level Load• Identify tasks with no slack
– this is the critical chain
© R. Barnes, 1998
Add Buffers
• Identify buffer points
• Decide on buffer sizes. Perhaps– Project Buffer - 1/2 project (= 1/2 padding saved)
– Feeding Buffers - 1/2 feed path (= 1/2 padding saved)
– Resource Buffers (Wake-up calls) - say 2 weeks
© R. Barnes, 1998
Buffers and Schedule Pressure
• Should project due-date be pushed to make room for buffers?
• YES!!!!!• BUFFERS ARE NOT OPTIONAL!!!!!• YOU HAVE ALREADY CHOPPED 25% OUT
OF THE SCHEDULE - DON’T LET MANAGEMENT FORCE YOU TO HIDE YOUR BUFFERS, AS YOU USED TO!
© R. Barnes, 1998
Critical ChainSession Outline
• Part 1 - Current Reality– Look at cause-effect relationships in
project management• Part 2 - Critical Chain Scheduling
– Key ideas - achieving “Miracle 1”• Part 3 - Global Viewpoint
– Dealing with Multiple Projects - keeping focus on The Goal
• Part 4 Implementation Issues– Overcoming resistance, making it all
happen
© R. Barnes, 1998
Global Viewpoint,Global Leverage
• We often think locally when a wider perspective gives better results
• Example - software company“We have 20% excess capacity. Market is
saturated, we could only sell more by selling below cost. Obvious answer - improve productivity by cutting costs”
See spreadsheets 1 and 2
(Reading 7, pages 1 and 2)
© R. Barnes, 1998
Cost vs Throughput
• What if we used excess capacity to sell below cost?– See Page 3
• Example illustrates– Local solutions can be sub-optimal– Throughput is Number 1(Ichiban)
• Not lowest cost, but best value!
© R. Barnes, 1998
The Throughput World
• Leverage Points– a constraint that we can improve– 80/20 - 20% of the possible
improvements will produce 80% of the gain– but because of linkages (improvements not
independent - one improvement causes another), more like
– 1% of possible improvements -> 95% of gain
© R. Barnes, 1998
Possible Levers
• Earlier delivery/Response time• Higher quality• Due-date performance• Image• Features and OptionsAll of these may be preferable to • Lowest costWhat does the customer value?
© R. Barnes, 1998
Global Improvementsthe 5 Focussing Steps
(Reading 8)• Identify the leverage point(s)• Exploit the leverage point(s)• Subordinate everything else to the above
decisions• Elevate the leverage points• Go back to step 1 - don’t let inertia become
a constraintThis is simply “The Goal”. Remember Herbie?
© R. Barnes, 1998
Variation
Select the leverage point(s)• Exploit the leverage point(s)• Subordinate everything else to the
above decisions• Elevate the leverage points• Evaluate whether the leverage point
should change
Avoid excessive chop&change of focusConcentrate on best-payoff leverage points
© R. Barnes, 1998
TOC Accounting
• Throughput pricing– Standard pricing may lead to wrong results– Use a pricing model that focuses on the
constraint• Reading 9
© R. Barnes, 1998
Multiple Projects• Single-project scheduling works well, even in
multiple-project environments, if– Individual projects practically independent
• eg, contractors used
• But, if resources shared, then– Each project needs significant buffers
• By planning projects together, we can – estimate/evaluate impact of decisions– better coordinate
• Result, need less buffers.
© R. Barnes, 1998
How?• Approach 1. Plan projects together.
– very tricky in a complicated environment• Critical chain keeps changing• Lots of detail
• Approach 2. Successive Projects(Add new project at end)– Better, but still tricky
• can still get swamped with detail
• Approach 3: manage the strategic resource– simplest, but can get contention for non-strategic
resource
© R. Barnes, 1998
Conclusion
• It ain’t easy!!
• No clear answer!!
• My opinion - Suggest we concentrate on Miracle 1 (Critical Chain), get familiar with this before going the next step to Global View
© R. Barnes, 1998
Critical Chain, Session Outline
• Part 1 - Current Reality– Look at cause-effect relationships in
project management• Part 2 - Critical Chain Scheduling
– Key ideas - achieving “Miracle 1”• Part 3 - Global Viewpoint
– Dealing with Multiple Projects - keeping focus on The Goal
• Part 4 Implementation Issues– Overcoming resistance, making it all
happen
© R. Barnes, 1998
What is a schedule?
(Reading 10)• What is the schedule designed to
accomplish?Maximise throughput– get as many projects completed as
possible• Minimise inventory (work in progress)
– gives shortest lead times
© R. Barnes, 1998
The worker needs -
• Start times for jobs with no predecessors (“gating” tasks)
• Relative priority if alternatives• Who gets the work next• ? Approximately when next job is
coming, and what it is• How urgent• (also task description/requirements)
© R. Barnes, 1998
The worker does NOT need
• Start times for non-gating tasks• Finish times
– Start, finish times inhibit early starts• Task durations
– becomes self-fulfilling
(do we agree with this? I’m not sure that I do)
© R. Barnes, 1998
Resource Manager needs -
• When tasks are late or early, and by how much
• How important is this– impact on the buffer
• Currently-expected start/finish times• Status of resource buffers
© R. Barnes, 1998
Project Manager Information Needs
• Project Buffer Status• Feeding Buffer Status• Critical Chain tasks and resource buffer statusSome rules • Don’t reschedule frequently• Don’t worry about late task unless it’s important• Evaluate decisions not just on costs, but also
on performance of the project• Keep it simple
© R. Barnes, 1998
Weak Links
• NIH (Not Invented Here)• SEP (Someone Else’s Problem)• Boundless Optimism• Milestones• Good Enough
© R. Barnes, 1998
Now what?
• Implementation Checklist– Reading 11
• Discussion– How much do we agree with?
• (Is it all a load of rubbish - new terminology for old ideas?)
– How to proceed?