Modelling Sensible Business...
Transcript of Modelling Sensible Business...
Pedro Antunes pedro.antunes @ vuw.ac.nz
Modelling Sensible Business Processes
Building rich process models for knowledge sharing
2016
IntroBusiness process management
!2
Business transformation
IT c
apab
ility
Workflow
Automation
Servitisation
Design of business
Intro
Process modelling is central to BPM approach Codifying organisational knowledge and behaviour
Enabling computational support
!3
IntroProcess modelling
!4
Business transformation
IT c
apab
ility
Workflow
Automation
Servitisation
Design of business
Functionalist view: control data, production data, flows, procedures, consistency
Flexibility: unique cases, loose activities, human interventions
Context awareness, decision making, agility
Coordination view: routing
Intro
Early days
Automation, rationalisation, and coordination
Nowadays
Design of business
Strategy, innovation, through-life services, collaboration, transparency, agility
!5
Intro
Early days
Easy elicitation-analysis-design cycle
Limited scope, known flows, clear procedures
Expected exceptions, predefined behaviour
Nowadays
Complex knowledge-sharing cycle
Unique cases, human discretion, changing goals
Unexpected exceptions, emergent behaviour
!6
Problem
How to capture and model process knowledge?
Lack of realism/detail/context
Lost in translation
Lack of flexibility
Lack of agility
!7
Illustration
Modeller: What do you do when event X arrives?
Domain expert: It depends on so many decisions…
Modeller: Sorry, I need to report ONE activity
Modeller: Which activity follows Y?
Domain expert: I often decide on the spot…
Modeller: Sorry, I need to know ALL conditions
!8
Example
!9
MeetingFirst morning flight
Expense report
“Happy” path
Evening flight
Example
!10
MeetingFly to meeting
Expense report
Unexpected path
Wait 3 days
Check alternatives
Take ferry
Fly back same day
Buy clothes
2 days at sea
Classification of Processes
!11Model
Rich Lean
Mac
hine
con
trol
Hum
an c
ontr
ol
Beha
viou
r
Classification of Processes
!12Model
Rich Lean
Mac
hine
con
trol
Hum
an c
ontr
ol
Beha
viou
r
1 Mechanistic processes
• Behaviour clarity • Predictable • No decision-making • No disturbances • Typical BPM
Classification of Processes
!13Model
Rich Lean
Mac
hine
con
trol
Hum
an c
ontr
ol
Beha
viou
r
2 Ad hoc
processes
• Exceptional events • Workarounds • Human ingenuity • Typical case handling
Classification of Processes
!14Model
Rich Lean
Mac
hine
con
trol
Hum
an c
ontr
ol
Beha
viou
r
3 Generative processes
1 Mechanistic processes
• Evolution • Breeding • Adaptation • Mining
Classification of Processes
!15Model
Rich Lean
Mac
hine
con
trol
Hum
an c
ontr
ol
Beha
viou
r
4 Sensible
processes
• Context sensitivity • Sensemaking • Augmentation • Meta-design
Classification of Processes
!16Model
Rich Lean
Mac
hine
con
trol
Hum
an c
ontr
ol
Beha
viou
r
3 Generative processes
1 Mechanistic processes
2 Ad hoc
processes
4 Sensible
processes
Sensible Business Processes
A definition
Processes that leverage both the human capacity for sensemaking and decision making and the processing capacity of complex information systems
!17
Sensible Business Processes
Our research questions
What process knowledge would be captured?
What methods and tools would be needed to capture such knowledge?
What would be the effectiveness of these tools and methods?
!18
Research Paradigm
Design science
Building artefacts
Advancing the knowledge base
Demonstrating utility
!19
Research Paradigm
!20
Build EvaluateArtefact
Artefact
Process stories
!21
Actors
DialogueScenes
Information resources
Narrative
Process Stories
!22
Process storytelling tool
Process Stories
!23
A process story in storyboard view
Process Stories
!24
A process story in structure view
Process Stories
Theoretical foundations
Sensemaking theory
“Sensemaking involves the ongoing retrospective development of plausible images that rationalise what people are doing”
Organisational storytelling theory
“Stories communicate complex ideas and spring people into action using narrative mechanisms”
“Stories bring detailed explanations, contextual information, values, and what-if considerations to knowledge sharing”
!25
Build-Evaluate Cycles
!26
Build Evaluate
We needed 3 build-evaluate cycles to answer our research questions
Artefact
Build-Evaluate Cycles
!27
Build Evaluate
Features Purpose Participants Cases
Cycle 1 Initial tool FormativeLarge organisation
Model existing process
Cycle 2Improved editing
Summative Small teamModel new process
Cycle 3Improved storytelling
SummativeLarge organisation
Model existing process
Cycle 1
!28
Case Size Data gathering
Cycle 1
Model existing process - Client relationship management
27 participants
1 stage: - Individual storytelling - Informal feedback
Cycle 1
Results (disappointing)
Many difficulties composing scenes (too many steps)
Problems viewing process stories
Wide variations in level of abstraction
Lack of contextual details
Lack of confidence expressing stories
Consequence
Improved tool with fewer clicks and automatic if-then-else
Improved training and instructions to users
!29
Cycles 2 and 3
!30
Case Size Data gathering
Cycle 2
Model new process - IT service provisioning
Small team - leader plus 5 members
2 stages: - Individual storytelling - Convergence meeting moderated by leader
Cycle 3Model existing process - administrative
Large organisation - 20 participants
3 stages: - Define reference process (sanctioned
by the organisation) - Modelling sessions (mainly individual) - Conversion of individual stories into
BPMN and comparison with reference
Research Questions
RQ1. Did the subjects create detailed stories?
RQ2. Can workflow be derived from user stories?
RQ3. Did the stories portray emotion?
RQ4. Were unexpected situations depicted?
RQ5. Was contextualised knowledge applied?
RQ6. Did sharing stories help the team better understand the process?
RQ7. Did the gathered stories influence the final adopted practices?
!31
Detail (Size & Complexity)
!32
0%
25%
50%
75%
100%
Structural complexity
None Low Medium High Very high
Valu
e A
xis
0%
25%
50%
75%
100%
Number of stories with “x” scenes
None Low Medium High Very high
New Existing
Low [1-4] Medium [5-9] High [10-19] Very high [20-∞]
Detail (Dialogue & Narrative)
!33
0%
25%
50%
75%
100%
Use of dialogue
None Low Medium High Very high
New process Existing process
0%
25%
50%
75%
100%
Use of narrative
None Low Medium High Very high
Detail
Did the subjects create detailed stories?
Yes - Regarding both size and complexity
Interestingly, size and complexity were higher with a new process
Yes - Regarding narrative
However, existing process had more narrative
Depends - Regarding dialogue
New process had less dialogue
!34
Workflow
!35
0%
25%
50%
75%
100%
Workflow elements
No Yes
New and Existing process
Can workflow be derived from user stories?
Yes - Every story had at least one workflow element (activity, condition, flow)
Emotion
!36
0%
25%
50%
75%
100%
Emotion
No Yes
New process Existing process
Did the stories portray emotion?
Divided results (50% mark) Existing process generated slightly more emotional elements
Surprise
!37
0%
25%
50%
75%
100%
Surprise
No Yes
New process Existing process
Were unexpected situations depicted?
Depends on type of process Existing process very often described with surprising situations
New process completely focussed on “happy path”
Context
!38
0%
25%
50%
75%
100%
Context
None Low Medium High Very high
New process Existing process
Was contextualised knowledge applied?
Depends on type of process Existing process highly contextualised
New process had low contextual cues
Understanding & Influence
RQ6. Did sharing stories help the team better understand the process?
RQ7. Did the gathered stories influence the final adopted practices?
To answer these questions, we had to adopt a different approach to data analysis
Chunking the process in logical segments and analysing individual contributions to each segment
!39
Understanding
!40
0
4
8
12
16
20
Number of scenes per story
Chunk 1 Chunk 2 Chunk 3 Chunk 40
4
8
12
16
20
Number of scenes in converged story
Chunk 1 Chunk 2 Chunk 3 Chunk 4
New process
Understanding
!41
0
6
12
18
24
30
Number of activities per individual process
Chunk 1 Chunk 2 Chunk 30
6
12
18
24
30
Number of activities in reference process
Chunk 1 Chunk 2 Chunk 3
Existing process
Understanding
Did sharing stories help the team better understand the process?
Yes - In both cases
The converged story had more scenes than most individual stories (new process)
An excessively detailed story was toned down by the group
The reference process had fewer activities than most individual processes (existing process)
!42
Influence
!43
0%
25%
50%
75%
100%
% individual activities appearing in individual stories and converged story
Chunk 1 Chunk 2 Chunk 3 Chunk 4
New process
Influence
!44
0%
25%
50%
75%
100%
% individual activities appearing in individual stories and converged story
Chunk 1 Chunk 2 Chunk 3 Chunk 4
New process
This story shaped chunk 3 of final story
This story had no influence on chunk 2
Influence
!45
0%
25%
50%
75%
100%
Influence of individual activities on reference process
Supports Contradicts
No Yes No Yes
Existing process
Influence
Did the gathered stories influence the final adopted practices?
Yes - In both cases
The converged story was significantly influenced by some stories
Those stories exhibited high expertise in specific chunks of the whole process
The reference process was supported by all stories, but also contradicted by some stories
!46
In a GlanceRQ1. Did the subjects create detailed stories?
YES (size, complexity): New > Existing
YES (narrative): Existing > New
DEPENDS (dialogue): Existing > New
RQ2. Can workflow be derived from user stories?
YES
RQ3. Did the stories portray emotion? Around 50% did
RQ4. Were unexpected situations depicted? DEPENDS: Existing = Yes, New = No
RQ5. Was contextualised knowledge applied? DEPENDS: Existing = High, New = Low
RQ6. Did sharing stories help the team better understand the process? YES
RQ7. Did the gathered stories influence the final adopted practices?
YES
!47
Discussion
Arguments in favour of sensible processes
Enriching knowledge of existing processes
85% process stories had high context reasoning
75% expressed surprising events
67% expressed emotion
100% activities supported the established process
43% activities contradicted the established process
Enriching knowledge of new processes
83% stories had high detail
66% had high complexity
!48
Discussion
Enriching knowledge
!49
This… …versus this
Discussion
Arguments in favour of sensible processes
Balancing the picture
Converged process was balanced when compared to individual process stories
More detail in some chunks, less in others
Reference process was simultaneously supported and contradicted by individual process stories
Better account of reality in some chunks
Integration with existing approaches
Every story contained workflow elements (mechanistic)
Stories identified alternative activities (surprises)
!50
Discussion
Integration with existing approaches
!51
Discussion
Domain experts, instead of modelling experts No specialised language or skills required
Collaboration Domain experts can effectively collaborate in building process stories
Process elicitation versus modelling No need for separate activities
!52
Discussion
Regarding design science Anonymous reviewer
“I understand what you intend, but the RQs are too fuzzy, 'can' can never be falsified. Sure, all you present CAN improve business processes, but it is more relevant whether and to which amount it really improves. There is no baseline/control group for comparison, still I think the results are interesting and should be reported”
Answer?
Design science is more about problem solving
“what”/ “can” type of research questions
Comparison with baseline was possible in one case (229% increase in number of process nodes), but not in others
!53
Future Work
Alternative ways for capturing process stories
Generic purpose tools?
Quality assessment of process stories Utility/understandability versus correctness/completeness?
Elicitation/modelling method Best practices?
More research needed for theory building
New versus existing processes
Big versus small organisations
Consensus versus divergence
!54
ReferencesSimões, D., P. Antunes and L. Carriço (2018). “Eliciting and Modelling Business Process Stories.” Business & Information Systems Engineering, 60 (2), pp. 115-132.
Simões, D., P. Antunes and J. Cranefield (2016). Enriching Knowledge in Business Process Modelling: A Storytelling Approach. Innovations in Knowledge Management: The impact of social media, semantic web and cloud computing. L. Razmerita, G. Phillips-Wren and L. Jain. Heidelberg, Springer. 95: 241-267.
Simões, D., N. Thuan, L. Jonnavithula and P. Antunes (2015). Modelling Sensible Business Processes. 2nd International Conference on Future Data and Security Engineering (FDSE). Ho Chi Minh City, Vietnam. T. Dang, R. Wagner, J. Küng, N. Thoai, M. Takizawa and E. Neuhold. Heidelberg, Springer. 9446.
Jonnavithula, L., P. Antunes, J. Cranefield and J. Pino (2015). Organisational Issues in Modelling Business Processes: An Activity-Based Inventory and Directions For Research. 19th Pacific Asia Conference on Information Systems (PACIS), Singapore, AIS.
Antunes, P., D. Simões, L. Carriço and J. Pino (2013). “An End-User Approach to Business Process Modeling.” Journal of Network and Computer Applications, 36 (6), pp. 1466-1479.
Simões, D., P. Antunes and J. Pino (2012). Humanistic Approach to the Representation of Business Processes. 16th IEEE International Conference on Computer Supported Cooperative Work in Design (CSCWD), Wuhan, China, IEEE.
Antunes, P. and H. Mourão (2011). “Resilient Business Process Management: Framework and Services.” Expert Systems With Applications, 38 (2), pp. 1241-1254.
Antunes, P. (2011). “BPM and Exception Handling: Focus on Organizational Resilience.” IEEE Transactions on System, Man, and Cybernetics Part C: Applications and Reviews, 41 (3), pp. 383-392.
!55