the risk driver approach to project schedule risk · pdf file3 . limitations with the...

© 2013 Hulett & Associates, LLC 1

The Risk Driver Approach to Project Schedule Risk Analysis

A Webinar presented by

David T. Hulett, Ph.D.

Hulett & Associates, LLC

To the

College of Performance Management

April 18, 2013


Risk Drivers Method Agenda

Limitations of traditional 3-point estimating

Explain “Risk Drivers” approach, start with the risks themselves – One Driver per activity

– Multiple Risk Drivers per Activity

Correlation is modeled in the Monte Carlos simulation

Assess the schedule against CPM scheduling best practice

Simple space vehicle development schedule as an example

Prioritize risks to schedule

Mitigation of high-priority risks


Limitations with the Traditional 3-point Estimate of Activity Duration

Typical schedule risk analysis starts with the activity that is impacted by risks, not the risks themselves

– Estimates the 3-point estimate for optimistic, most likely and pessimistic duration

– Creates a probability distribution for activity duration

– Performs Monte Carlo simulation

Which risks cause the most overall schedule risk? These questions are typically answered by:

– Sensitivity to activity durations

– Criticality of activity durations


Some Problems with Traditional Approach

Can tell which activities are crucial, but not directly which risks are driving

Makes poor use of the Risk Register that is usually available

Cannot decompose the overall schedule risk into its components BY RISK – Ability to assign the risk to its specific risk drivers helps with

communication of risk causes and risk mitigation


We Propose a Different Approach: Start with the Risks Themselves

Drive the schedule risk by the risks already analyzed in the Risk Register

For each risk, specify: – Probability it will occur

– Impact on time if it does

– Activities it will affect

Starting with the risks themselves gives us benefits – Links qualitative analysis to the quantitative analysis

– Estimates the impact of specific risks for prioritized mitigation purposes

– Correlations between activities happen automatically – never have to guess at these coefficients again, never get impossible matrices


Simple Example of Risk Register Risks

Use the Risk Drivers feature in Pertmaster 8

Collect probability and impact data on risks

Load the risks

Assign risks to schedule activities


Risk Drivers Mechanics (1)

The risk Driver is assigned to one or several activities,

affecting their durations by a multiplicative Driver

– E.g., the Driver may be .90 for optimistic, 1.0 for most likely and

1.25 for pessimistic

– These Drivers multiply the schedule durations of the activities to

which they are assigned

Risks can be assigned to one or more activities

Activity durations can be influenced by one or more risks


Risk Drivers Mechanics (2)

Risk Drivers are assigned a probability of occurring on any

iteration.

– When the risk occurs, the Driver used is chosen at random from

the 3-point estimate and operates on all activities to which it is

assigned

– When not occurring on an iteration the risk Driver takes the value

1.0, a neutral value

When an activity is influenced by more than one risk, their

Drivers are multiplied together, if they happen on an

iteration


Risk Driver Probability is 100%, Driver can be + or -

Here the

Ranges are

based on

deviations +

and – from the

Plan.

Probability is

100%

For the examples

we use an activity

with 100 days in

the schedule

90 95 100 105 110 115

Distribution (start of interval)

0

50

100

150

200

250

300

350

400

Hit

s

0% 90

5% 94

10% 95

15% 96

20% 97

25% 98

30% 99

35% 99

40% 100

45% 101

50% 101

55% 102

60% 103

65% 104

70% 104

75% 105

80% 106

85% 108

90% 109

95% 111

100% 115

Cu

mu

lati

ve F

req

uen

cy

Entire Plan : Duration

, 10

Risk Driver Prob. = 100%, Driver is all Overrun

Here the

Plan is the

Optimistic

Value.

Probability is

100%

100 105 110 115 120 125 130


0

50

100

150

200

250

300

350H

its

0% 100

5% 104

10% 105

15% 107

20% 108

25% 109

30% 109

35% 110

40% 111

45% 112

50% 113

55% 114

60% 115

65% 116

70% 117

75% 118

80% 119

85% 121

90% 122

95% 125

100% 130

Cu

mu

lati

ve F

req

uen

cy



Assigning a Probability Less than 100%

The essence of a “risk” is its uncertainty in two dimensions:

– Uncertainty of its occurrence, specified by a probability

– Uncertainty of its impact, specified by a range of durations

If the risk may or may not occur, we specify the probability

that it will occur

– The risk occurs and affects the activities it is assigned to on X% of

the iterations, chosen at random, the multiplicative Driver used is

chosen at random from the range of data input by the user

– On (1 – X)% of the iterations, Driver takes 1.0 value


Assigning a Probability Less than 100%

Spike

contains

40% of

the

probability

Spike

contains

70% of

the

probability

100 105 110 115 120 125 130


0

200

400

600

800

1000

1200

1400

1600

1800

2000

Hit

s

0% 100

5% 100

10% 100

15% 100

20% 100

25% 100

30% 100

35% 100

40% 103

45% 106

50% 107

55% 109

60% 110

65% 111

70% 113

75% 114

80% 116

85% 118

90% 120

95% 123

100% 130

Cu

mu

lati

ve F

req

uen

cy


90 95 100 105 110 115


0

500

1000

1500

2000

2500

3000

3500

Hit

s

0% 90

5% 97

10% 99

15% 100

20% 100

25% 100

30% 100

35% 100

40% 100

45% 100

50% 100

55% 100

60% 100

65% 100

70% 100

75% 100

80% 100

85% 101

90% 104

95% 107

100% 114

Cu

mu

lativ

e F

req

uen

cy



Assigning More than One Risk to an Activity

If more than one risk is acting on an activity, the resulting

ranges are the multiplication of the percentages

This is reality – an activity is often affected by multiple risks

Two cases are shown next:

– When both risks are 100% likely to occur

– When both risks are < 100% likely to occur

In each case, the computer simulation creates the

uncertainty range on an activity’s duration – it is not

estimated


Parallel and Series Risks – Multiplicative – Used with Risk Drivers (RiskDrivers)

Risk 1 1.2 Driver

Risk 2 1.05 Driver

If these two risks are parallel, they can be recovered

simultaneously

Risk 1 1.2 Driver Risk 2 1.05 Driver

Use 1.2 Driver, the largest

Driver only

Use (1.2 x 1.05 = 1.26)

Driver, multiply the two

If these two risks are series, they can not be

recovered simultaneously

, 15

Two Risks affect One Activity using Drivers that Occur 100% - placed in Series

Risks in

series, P80

is 123 days

100 110 120 130 140


0

20

40

60

80

100

120

140

160

180

200

220

240

260

Hit

s

0% 92

5% 102

10% 104

15% 106

20% 108

25% 109

30% 110

35% 111

40% 112

45% 114

50% 115

55% 116

60% 117

65% 118

70% 119

75% 121

80% 123

85% 124

90% 127

95% 131

100% 146

Cu

mu

lati

ve F

req

uen

cy



Two Risks affect One Activity using Drivers that Occur 100% - in Parallel

100 105 110 115 120 125 130


0

50

100

150

200

250

300

350

Hit

s

0% 100

5% 105

10% 106

15% 107

20% 108

25% 109

30% 110

35% 111

40% 111

45% 112

50% 113

55% 114

60% 115

65% 116

70% 117

75% 118

80% 119

85% 121

90% 122

95% 125

100% 130

Cu

mu

lati

ve F

req

uen

cy


Risks in

parallel, P80

is 119 days


Two Risks with Less than 100% Probability Affecting one Activity – Risks in Series

The spike at 100

days represents (1)

the likelihood that

neither risk occurs

[60% x 50% =

30%] and (2) the

chance that 100

days is picked

when one or both

occur.

90 100 110 120 130 140


0

200

400

600

800

1000

1200

1400

1600

1800

Hit

s

0% 90

5% 96

10% 99

15% 100

20% 100

25% 100

30% 100

35% 100

40% 100

45% 101

50% 103

55% 105

60% 107

65% 108

70% 110

75% 112

80% 114

85% 116

90% 119

95% 123

100% 144

Cu

mu

lati

ve F

req

uen

cy



Two Risks with Less than 100% Probability Affecting one Activity – Risks in Parallel

100 105 110 115 120 125 130


0

200

400

600

800

1000

1200

1400

1600

1800

2000

2200

Hit

s

0% 100

5% 100

10% 100

15% 100

20% 100

25% 100

30% 100

35% 100

40% 100

45% 101

50% 103

55% 105

60% 107

65% 109

70% 110

75% 111

80% 113

85% 115

90% 118

95% 122

100% 130

Cu

mu

lati

ve F

req

uen

cy


With one parallel

risk’s having a

minimum range of

100%, it cannot be

less than 100 days


Risk Drivers Model How Correlation Occurs Coefficients are Calculated (1)

Risk Probability = .5,

Range .95, 1.05, 1.15

Activity 2 Activity 1

Correlation = 100%


Scatter showing 100% Correlation


Risk Drivers Model How Correlation Occurs Coefficients are Calculated (2)

Correlation is modeled as it is caused in the project

Correlation coefficients are generated, not guessed


Range .95, 1.05, 1.15

Activity 2 Activity 1

Correlation = 37%


Range 1.0, 1.10, 1.20 Risk Probability = .25,

Range .8, .95, 1.05


Scatter showing 37% Correlation


Sensitivity to the Risk Drivers

Risk 1 is more

important since it

affects both Activity

A and Activity B


Schedule Check Report in Pertmaster

25 (C) 2010-2013 Hulett &

Associates LLC


Using Acumen FUSE for Best Practices

(C) 2010-2013 Hulett &

Associates LLC

26


Simple 2-Stage Space Vehicle Schedule

Software used: Pertmaster v. 8.7


Two Types of Risk

Inherent variability including duration estimating error – uncertainty – Probability = 100% – Used Quick Risk of -5% and +10%, could be reference ranges for

different types of activities

– Could use reference ranges that would differ by type of activity

Discrete risks derived from Risk Register – Probability < 100% – Summarized from detailed Risk Register

– These have a probability of occurring and an impact on specific activities if they do

– Parallel to their Risk Register information


Standard 3-point Range Representing Inherent Variability and Duration Estimating Error

Inherent variability

and estimating

error: Optimistic -

5% Pessimistic

+10%


Results with Inherent Variability and Duration Estimating Error Only

Deterministic: 13 APR

2020 is 1%

P-80 is 30 JUL 20, about

3.5 months later than

planned

Spread from P-5 to P-95

is 5 MAY 20 to 27 AUG

20 for 3.7 months

15 Mar 20 04 May 20 23 Jun 20 12 Aug 20 01 Oct 20


0

50

100

150

200

250

300

350

400

Hit

s

0% 11 Mar 20

5% 05 May 20

10% 15 May 20

15% 25 May 20

20% 01 Jun 20

25% 08 Jun 20

30% 11 Jun 20

35% 17 Jun 20

40% 22 Jun 20

45% 26 Jun 20

50% 01 Jul 20

55% 06 Jul 20

60% 10 Jul 20

65% 15 Jul 20

70% 20 Jul 20

75% 24 Jul 20

80% 30 Jul 20

85% 06 Aug 20

90% 14 Aug 20

95% 27 Aug 20

100% 30 Oct 20

Cu

mu

lati

ve F

req

uen

cy

Spacecraft for PMChallenge 2009Entire Plan : Finish Date


Risk Analysis on Space Vehicle Project Risk Drivers are from Risk Register

Seven risk Drivers have been identified and quantified.

Each Risk has probability assigned

Five have optimistic ranges possible, two are pure threats

Risk Min Most Likely Max Likelihood

Requirements have not been decided 95% 105% 120% 70%

Several alternative designs considered 95% 100% 115% 60%

New designs not yet proven 90% 103% 112% 40%

Fabrication requires new materials 95% 105% 115% 50%

Lost know-how since last full spacecraft 100% 100% 105% 30%

Funding from Congress is problematic 90% 105% 115% 70%

Schedule for testing is aggressive 100% 120% 130% 100%


Risks Assigned to Activities (1)

Risk Requirements

Definition

FS Preliminary

Design

FS Final

Design

FS

Fabrication

Test FS

Engine

Requirements Not

Complete X

Alternative Designs

Possible X

Designs Not Proven X

New Materials in

Fabrication X

Lost Know-How X

Funding Problematic X X X X

Testing Schedule

Aggressive X

32


Risks Assigned to Activities (2)

Risk US Preliminary

Design

US Final

Design

US

Fabrication

US

Test Integration

Integration

Testing

Requirements Not

Complete

Alternative Designs

Possible X

Designs Not Proven X

New Materials in

Fabrication X

Lost Know-How X X

Funding Problematic X X X X X X

Testing Schedule

Aggressive X X

33


Results Adding Risk Drivers to the Background Risk

Baseline 13 APR 20 is only

3% likely

80th percentile is 7 APR 21,

11.8 months later

Spread P-5 to P-95 is

13May20 to 6 Aug 21, for ~

15 months

25 Jan 20 12 Aug 20 28 Feb 21 16 Sep 21


0

50

100

150

200

250

300

350

400

450

500

550

Hit

s

0% 12 Nov 19

5% 13 May 20

10% 30 Jun 20

15% 28 Jul 20

20% 14 Aug 20

25% 31 Aug 20

30% 11 Sep 20

35% 28 Sep 20

40% 13 Oct 20

45% 29 Oct 20

50% 16 Nov 20

55% 03 Dec 20

60% 22 Dec 20

65% 13 Jan 21

70% 05 Feb 21

75% 03 Mar 21

80% 07 Apr 21

85% 10 May 21

90% 16 Jun 21

95% 06 Aug 21

100% 08 Mar 22

Cu

mu

lati

ve F

req

uen

cy

Spacecraft for PMChallenge 2009Entire Plan : Finish Date


Activity Tornado Chart from All-In Simulation

Risky Activities:

Fabrication, Integration,

Final Design, Preliminary

Design, Testing

All are correlated with the

finish date > 60%

These are activities /

paths, NOT RISKS

61%

61%

63%

70%

70%

75%

76%

79%

81%

81%00025 - US Fabrication

00011 - FS Fabrication

00028 - Integration

00023 - US Final Design

00009 - FS Final Design

00021 - US Preliminary Design

00007 - FS Preliminary Design

00029 - Integration Testing

00026 - US Test

00012 - Test FS Engine

Spacecraft for PMChallenge 2009Duration Sensitivity: Entire Plan - All tasks


Risk Driver Tornado from All-In Simulation

The main RISK, however, is

funding from Congress,

which affected all activities in

this model.

This is the main risk to

mitigate, if possible

6 - Funding from Congress is problematic

4 - Fabricaton requires new materials

3 - New designs not yet proven

7 - Schedule for testing is aggressive

2 - Several alternative designs considered

5 - Lost know-how since last full spacecraft

1 - Requirements have not been decided

8 - Cost Risk is based on immature data

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80

Correlation

Risk Factors Driving Project Schedule


Contribution of Each Risk to the Contingency (1)

Explain the Contingency to the P-80

P-80 Date Take Risks Out:

All Risks In 7-Apr-21 Days Saved % of Contingency

Specific Risks Taken Out in Order

Funding 23-Nov-20 135 38%

Testing Schedule 5-Oct-20 49 14%

New Materials 3-Sep-20 32 9%

Alternative Design 21-Aug-20 13 4%

Requirements 14-Aug-20 7 2%

New Design 6-Aug-20 8 2%

Lost Know How 31-Jul-20 6 2%

Uncertainty

Natural Variation &

Estimating Error 13-Apr-20 109 30%

Total Contingency 359 100%


Contribution of Each Risk to the Contingency (2)

ia ion:1 ariation:45 :11 3 5 4 ariation:3 Variation:50

Variation:134 Variation:49 ation:32 n:13 n:7 :8 Variation:6 Variation:108

19 20 Jan 20 10 Mar 20 29 Apr 20 18 Jun 20 07 Aug 20 26 Sep 20 15 Nov 20 04 Jan 21 23 Feb 21 14 Apr 21 03 Jun 21 23 Jul 21 11 Sep 21 31 Oct 21 20 Dec 21 08 Feb 22

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Cu

mu

lativ

e P

ro

bab

ilit

y

Graphic of the effect

of taking the risks

out in order of

priority at the P-80


Mitigating the Most Important Risk

Mitigating the risk is estimated to reduce the probability from 70% to 30%

but, if the risk happens, the impact range remains the same. This saves

92 calendar days at the P-80 target level of confidence

Effect of Partially Mitigating the Risk with the Highest Priority

Min Most Likely Max Likelihood P-80 Date

Funding from Congress is problematic 90% 105% 115% 70% 7-Apr-21

Mitigation: Situate your suppliers strategically by Congressional District of the members of the Committee. Make the “jobs” case for continuous funding.

Impact on the parameters 90% 105% 115% 30% 5-Jan-21


Summary (1)

The focus is on the risks, not their impact

Risks “explain” the need for a contingency

Management appreciates this focus on risks

Risk interviews are conducted at 5,000 foot level, where

people typically think of risk

Interviews go faster, stick to the substance


Summary (2)

Use Risk Register for quantitative analysis

Specific risks can be quantified and assigned to schedule

activities

– Quantification is probability and impact

– A risk can affect several activities

– An activity can be affected by several risks

Risk Drivers can be combined with other more traditional

approaches such as 3-point estimates for inherent variability

risk, estimating error or for probabilistic branching


The Risk Driver Approach to Project Schedule Risk Analysis

A Webinar presented by

David T. Hulett, Ph.D.

Hulett & Associates, LLC

To the

College of Performance Management

April 18, 2013

the risk driver approach to project schedule risk · pdf file3 . limitations with the...

Documents