2. Project management 2
Overview
■ Project management approaches
■ CommonKADS spiral approach
■ How to do a risk analysis
■ Project planning and the concept of model states
■ Case Study: a project on reactor noise analysis
■ Experience: points of attention
3. Project management 3
Control versus Flexibility
■ knowledge projects often have learning character
■ structure of knowledge may turn out to deviate
■ requirements (even) more likely to change
➤ goals adjusted along project route
4. Project management 4
Waterfall project approach
S trategy
P hase
Information
Analysis
S ystem
Design
P rogram
and
Test
Operation
Maintenance
5. Project management 5
Merits and disadvantages of
waterfall LCM
■ strong handle for project control
■ early phases are document-oriented
➤ often need for operational (partial) results
■ changing needs and requirements difficult and costly
■ adequate for applications with clear route
➤ not often true for knowledge-intensive systems
6. Project management 6
Evolutionary or rapid
prototyping approach
Gather
E xpert
data
Implement
P rototype
Validate
Get
feedback
Iterate
7. Project management 7
Basic Ideas underlying
KBS Project Approach
■ It's not activities, but products that count
■ Project lifecycle must be configurable
■ Configuration is based on risk assessment
■ Quality is engineered in through (1) model suite
based development; (2) risk-based cyclic project
management
■ Development (1) and management (2) are linked
through the concept of model (product) states
8. Project management 8
Boehm's Spiral Model
R E VIE W
R IS KP L A N
MONIT OR
cycle-‐0
cycle-‐1
cycle-‐2
cycle-‐3
9. Project management 9
CommonKADS Spiral
Lifecycle Model
MONIT OR R E VIE W
R IS KP L A N
-‐
monitor
development
work
-‐
prepa re
a ccepta nce
a s s es s ment
-‐
eva lua te
cycle
res ults
-‐
review
prog res s
-‐s et
cycle
objectives
-‐
cons ider
cons tra ints
-‐
inves tig a te
a lterna tives
-‐
commit
-‐
pla n
cycle
ta s ks
-‐
a lloca te
res ources
-‐
a g ree
a ccepta nce
criteria
-‐
identify
ris ks
-‐
ca rry
out
ris k
a s s es s ment
-‐
decide
on
development
s teps
10. Project management 10
Project management cycle
1. Review
➤ current status of the project is reviewed
➤ objectives for upcoming cycle are established
➤ special case: cycle-0
– project plan + quality plan is developed
➤ ensure continued commitment of stakeholders
2. Risk
➤ obstacles identified
– significance assessed
➤ counteractions are decided upon
11. Project management 11
Project management cycle
3. Plan
➤ make detailed plan for next cycle
– work breakdown structure
– schedule of tasks
➤ allocating needed resources and personnel
➤ agreeing on acceptance criteria
4. Monitor
➤ evaluating outputs
➤ track progress
➤ meeting with stakeholders
12. Project management 12
How to do a
Risk Assessment
■ Risk = (Likelihood of Occurrence) x (Severity of
Effect)
■ Valuate both on a qualitative (five-point) scale
■ Subsequently rank all risks on this basis
■ Device countermeasures for each risk
■ Plan accordingly, and take the risks with high priority
rank first
■ Risk assessment: see Worksheet PM-1 (is part of
project documentation)
14. Project management 14
Knowledge-oriented quality
features
■ Knowledge capture:
➤ quality features of knowledge elicitation, modeling and
validation activities by knowledge engineers
■ Knowledge usability
➤ quality features of knowledge itself
15. Project management 15
The Concept of
Model States
■ Models are seen as (sub)products of KBS project
■ Project management: get products from one state
into another, desired next state
■ Qualitative range of state values: empty, identified,
described, validated, completed
■ Project planning is state transition based
➤ See Worksheet PM-2
■ May apply to separate models or even components,
depending on perceived risk
16. Project management 16
PM-2: Setting plan objectives
through model states
■ What model(s) are to be worked on in next cycle?
■ Which model component(s)?
■ To what degree?
➤ refers to model state
■ By what means, resources, development method or
technique?
■ Success condition
■ Quality metrics
17. Project management 17
CommonKADS Planning is
based on Model States
S et
objectiv es
Identify
ris ks
Define
targ et
model
s tates
R ev iew
objectiv es
Quality
C ontrol
current
model
=>
new
model
des cription
P lan
dev elopment
activ ities
-‐
unders tand
current
s ituation
-‐
problem
des cription
incomplete
O M:
problem
des cription
=
validated
O M:
s tructure
=
des cribed
T M:
decompos ition
=
des cribed
O M:
proces s
=
des cribed
T M:
time
load
=
des cribed
O M:
problem
=
des cribed
O M:
problem
=
validated
Dev elopment
P rojec t
Manag ement
18. Project management 18
Project Documentation
■ Project Plan
➤ Project motivation, background, scope, goals
➤ Project deliverables
➤ Work breakdown: cycles, resources
➤ Project organization
■ Quality Plan
■ Cycle Documentation
■ Project Close Down Report
➤ lessons learnt, recommendations and proposed guidelines
follow-up work, improvements
19. Project management 19
Case: a Project on Nuclear
Reactor Noise Analysis
steam generator
P
reactor vessel
core
neutron detectors
pumps
21. Project management 21
Risk: Cycle 1
Risk list (ranked in this order; see Table 12.9):
■ Knowledge engineer not acquainted with this (complex)
domain
– Countermeasure: do part of domain model first
■ Nature and complexity of noise interpretation task
unknown (classification, monitoring, assessment,
model-based diagnosis?)
– Countermeasure: scenario-based task modeling
■ Limited availability of the expert
– Countermeasure: develop contacts with other experts
22. Project management 22
Plan: Cycle 1 Gantt Chart
time
KM-‐a T M-‐b
T M-‐a T M-‐c
OM-‐a OM-‐b
840
88
24 60
C Y C L E -‐1
23. Project management 23
Expectation Management!
From an interview transcript:
■ Knowledge engineer:
– Question (shows noise spectra): if the reactivity differs from
the expected value, is it possible to tell what the potential
causes are? Does it also show up in or affect other physical
parameters?
■ Expert:
– Answer: In what language are you going to implement the
system? On a VAX or a PC?
24. Project management 24
Conclusion of First Cycle;
Risk Analysis Second Cycle
■ Reactor noise interpretation is assessment task
(similar to credit card fraud or housing!)
➤ Hence: (1) feasible for KBS; (2) assessment task template
can be reused (and it did work!)
■ Perceived major risk in second cycle: detailed insight
needed in economic costs and benefits
➤ Hence, second-cycle plan: focus on organization and task
model, esp. value, resource and performance components
25. Project management 25
Gantt Chart Cycles 2 and 3
Further cycles were rather waterfall-like
time
KM-‐b
DM
KM-‐c KM-‐e
2020
20
10time
T M-‐d
OM-‐c
40
40
C Y C L E -‐2 C Y C L E -‐3
KM-‐d
20
26. Project management 26
KBS Cost Estimation
■ KBS economy similar to other complex information
systems. Rough figures (cf. Boehm: S/w Eng. Economics)
➤ Project management: 10%
➤ Organization/Task/Agent Models: 10%
➤ Knowledge modeling (information analysis): 30%
➤ Design and Communication Models: 20%
➤ Implementation and testing: 25%
➤ Quality assurance: 5%
■ Notes: (1) high variability; (2) exchanges
possible,depending on project and approach (e.g.
GUI prototyping)
27. Project management 27
Requirements Engineering
■ Requirements engineering in software is still
underestimated and underdeveloped
– cf. books like Sommerville: Software Engineering
■ CommonKADS Model Suite can be seen as a
method to structure the requirements elicitation and
capture process
■ Analogy between requirements and knowledge:
– Tacit vs. explicit
– What is said may not coincide with what is really used
– Context dependence, organizational environment
– The human factor
28. Project management 28
Experience-based project
management guidelines
■ Identify and interview stakeholders right at the
beginning; keep them informed and interested
■ Be prepared for different viewpoints and interests
■ Manage expectations all the time
■ Risks are often not of a technical nature!
■ Requirements have a life of their own: they emerge,
grow and change, multiply, and die
■ Use the universal 80/20% rules
■ Learn from our recipes for failure !