What happened to "business-driven" BI? Why did the initial attempts not work? How did BI vendors respond? How did user organizations and IT departments adapt? What kinds of modeling can help understand the evolving situation and suggest next steps?
"Adapting to Uncertain and Evolving Requirements: The Case of Business-Driven Business Intelligence" Presentation at the RCIS 2013 conference. Eric Yu, Alexei Lapouchnian, and Stephanie Deng.
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Adapting to Uncertain and Evolving Requirements: The Case of Business-Driven Business Intelligence - RCIS 2013 yu-lapouchnian-deng
1. Eric Yu, Alexei Lapouchnian,
and Stephanie Deng
Faculty of Information and
Department of Computer Science
University of Toronto
May 29, 2013
Adapting to Uncertain and
Evolving Requirements
The case of Business-Driven Business Intelligence
2. Motivation
IS today function in dynamic, unpredictable world
Uncertainties, changing requirements and business and
technological landscapes.
Ongoing change and adaptation
Design/Try/Monitor/Evaluate/Redesign
Need new methods for doing analysis/design
Most current methods assume stable/predictable setting
Need to cope with ongoing change
Agility trend – frequent releases
Our Objectives
Motivate the need for new modeling techniques for such systems
Can existing modeling techniques be applied in these dynamic
environments?
In this paper, we concentrate on applying social and process models
in a socio-technical context
2
3. Presentation Overview
Motivation
Example Domain Introduction: BI Adoption
Applying social and process modeling techniques to
the evolving the BI Adoption scenario
Analysis of applicability of the existing notations
Discussion
Reflections on the use of the existing techniques
Wish list for future notations
Future Work
RCIS 2013, 29.05.20133
6. Typical Path in BI Adoption
RCIS 2013, 29.05.20136
Handling
PUs
Traditional
BI
Self-Service
BI
Tailoring to
CUs
Leveraging
SUs
1
2
3
4
5
*not shown in the presentation
*
…
(Further iterations
may follow)
*
7. Typical Path in BI Adoption
RCIS 2013, 29.05.20137
Handling
PUs
Traditional
BI
Self-Service
BI
Tailoring to
CUs
1
2
3
5
*not shown in the presentation
*
…
(Further iterations
may follow)
Business Intelligence is about
transforming raw data into
meaningful and useful business
information
• Historical, predictive, current views
of business data
Leveraging
SUs
4
*
8. Typical Path in BI Adoption
RCIS 2013, 29.05.20138
Handling
PUs
Traditional
BI
Self-Service
BI
Tailoring to
CUs
1
2
3
5
*not shown in the presentation
*
…
(Further iterations
may follow)
Business Intelligence is about
transforming raw data into
meaningful and useful business
information
• Historical, predictive, current views
of business data
Context: Dynamic business environment
Focus: Reporting – bringing BI to BUs
Example Domain
Iterative BI solution adoption path
• Common sequence of socio-technical
solution implementations – evolution
of BI solutions in enterprises
• Sources: industry reports [1,2], contacts
in the industry
Leveraging
SUs
4
*
9. Can Modeling Help? How?
RCIS 2013, 29.05.20139
Report
interactivity
D
D
CU
BI Team
D
Standard
Reports
Ad hoc
result
D
D
D
D
D
DD
D
D
Ad hoc
authoring
tool
Ad hoc request
be fulfilled Ad hoc
authoring tool
be provided
D
D
Perform ad
hoc BI query
Execute BI
query
D
D
Be
implementedD
D
SU
Timely
[ad hoc report]
Ad hoc
reports
D
D
D
D
Access pre-
built standard
reporting
Ad hoc analysis
be conducted
Monitor
business
Right
time
Fulfilled by
SU
Provide standard
reporting
DW
Meet business
information needs
Fulfill ad hoc
request
Interactivity
Help
Learn tool
Quick to learn
[tool]
Help
Empowered
Help
Analytical
aptitude
On-demand
tool feature
D
D
Help
Timely
[analysis]
Help
Help
Ad hoc analysis
be supported
Provide ad hoc
authoring tool
D
D
Belonging
[BI Team]
D
BI tool
expertise
Help
Governance
[data]
D
Gather input
Recognition
[SU]
Help
D
D
Help
Follow
enterprise
standard
Help
Formulate BI
query
On-demand
tool feature
D
D
Technical
skills
Accurate
requirements
User
behavior
Sense and
analyze user
needs
Help
Enterprise
standard be
followed
D
Tool
input
D
Tool training be
provided
Std. report
requirements
User
behavior
D
D
Evolving Socio-Technical Solutions
Process Models: processes, lifecycles, change
Social and Goal Models: functional/non-functional objectives,
actors, relationships
10. 1: Traditional BI
RCIS 2013, 29.05.201310
Handling
PUs
Traditional
BI
Self-Service
BI
Tailoring to
CUs
1
2
3
5
*not shown in the presentation
*
…
(Further iterations
may follow)
Typical Initial BI solution within
many BI adopters:
• BU – standard reports
• BI Team – repository of
standard reports and
development of ad hoc reports
Leveraging
SUs
4
*
11. 1: Traditional BI
RCIS 2013, 29.05.201311
Handling
PUs
Traditional
BI
Self-Service
BI
Tailoring to
CUs
Leveraging
SUs
1
2
3
4
5
*not shown in the presentation
*
…
(Further iterations
may follow)
Typical Initial BI solution within
many BI adopters:
• BU – standard reports
• BI Team – repository of
standard reports and
development of ad hoc reports
What actually happened:
• Business domain volatility → changing
analytical needs of BUs
• Gap b/w capabilities of standard reports and
BU needs grows → # of ad hoc report
requests increase → bottleneck
12. The Case for Modeling
RCIS 2013, 29.05.201312
Handling
PUs
Traditional
BI
Self-Service
BI
Tailoring to
CUs
1
2
3
4
5
*not shown in the presentation
*
…
(Further iterations
may follow)
Leveraging
SUs
4
*
Given what happened,
• Could modeling have anticipated the result?
• Can modeling be used to help understand what
happened?
• Can modeling help us find a new solution?
13. RCIS 2013, 29.05.2013
Using the i* social modeling notation to capture the scenario
13
1: Traditional BI: Modeling
14. RCIS 2013, 29.05.2013
Using the i* social modeling notation to capture the scenario
14
LEGEND
Task
Resource
Goal
Softgoal
Help
Hurt
Means-ends link
Decomposition link
Contribution link
Actor
boundary
D
Dependency link
?
Denied
Unknown
1: Traditional BI: Modeling
15. RCIS 2013, 29.05.2013
Analyzing the model. With i*, we are able to capture:
Actors, their functional and non-functional goals, and dependencies
15
1: Traditional BI: Modeling
16. RCIS 2013, 29.05.2013
Analyzing the model. With i*, we are able to capture:
Actors, their functional and non-functional goals, and dependencies
Unmet goals, the driving forces for change
16
1: Traditional BI: Modeling
17. RCIS 2013, 29.05.2013
Analyzing the model. With i*, we are able to capture:
Actors, their functional and non-functional goals, and dependencies
Unmet goals, the driving forces for change
Alternative ways of fulfilling goals – i.e., the possible adaptation paths
17
1: Traditional BI: Modeling
18. RCIS 2013, 29.05.2013
Analyzing the model. With i*, we are NOT able to capture:
Dynamics of the scenarios – i* model is just a snapshot.
Speed and rates of change.
Frequencies of occurrences (e.g., of dependencies).
18
i* Modeling
• i* could not have anticipated the failure of
Solution 1, can somewhat help understand
what happened, can help identify future
alternatives.
1: Traditional BI: Modeling
19. Process Models: Modeling Temporal
and Iterative Aspects
Social models
Can capture motivations and driving forces behind change
(+social aspects)
We use BPMN (with extensions) to:
Look at multiple layers of change in dynamic environments
Determine if we can
Visualize feedback loops
Capture the details of (re-)design cycles, analyze their relative
frequencies, duration.
RCIS 2013, 29.05.201319
20. BusinessMonitoring
Sell Products
Online
Analyze &
Interpret
Decide
Managerial
Action
Personalize
Create
(Query)
Execute Analyze
Generate
(Ad hoc
report)
Adjust?
Analyze
(Reporting
Requirement)
Design Develop Test
Deploy
(Standard
report)
Data
Warehou
sing
Drill Down
C
Continued
D= minutes
Operational
Process
AdhocReporting
Standard
Reporting
D= seconds ~
minutes
D= hours
D= days
Report interactivity
Managerial Action
Standard reports Ad hoc reports
Ad hoc request
Report requirements
Operational data
S
C
S
M S
M
M
S
Report requirements
S
Legend
Pool
Task
Sequence flow
Annotation
S
M
C
Sensing flow
Control flow
Mechanism flowStart
event
End event
Start message
DurationMessage flow D
Text
End message
Association
21. BusinessMonitoring
Sell Products
Online
Analyze &
Interpret
Decide
Managerial
Action
Personalize
Create
(Query)
Execute Analyze
Generate
(Ad hoc
report)
Adjust?
Analyze
(Reporting
Requirement)
Design Develop Test
Deploy
(Standard
report)
Data
Warehou
sing
Drill Down
C
Continued
D= minutes
Operational
Process
AdhocReporting
Standard
Reporting
D= seconds ~
minutes
D= hours
D= days
Report interactivity
Managerial Action
Standard reports Ad hoc reports
Ad hoc request
Report requirements
Operational data
S
C
S
M S
M
M
S
Report requirements
S
Legend
Pool
Task
Sequence flow
Annotation
S
M
C
Sensing flow
Control flow
Mechanism flowStart
event
End event
Start message
DurationMessage flow D
Text
End message
Association
Capturing Duration
22. BusinessMonitoring
Sell Products
Online
Analyze &
Interpret
Decide
Managerial
Action
Personalize
Create
(Query)
Execute Analyze
Generate
(Ad hoc
report)
Adjust?
Analyze
(Reporting
Requirement)
Design Develop Test
Deploy
(Standard
report)
Data
Warehou
sing
Drill Down
C
Continued
D= minutes
Operational
Process
AdhocReporting
Standard
Reporting
D= seconds ~
minutes
D= hours
D= days
Report interactivity
Managerial Action
Standard reports Ad hoc reports
Ad hoc request
Report requirements
Operational data
S
C
S
M S
M
M
S
Report requirements
S
Legend
Pool
Task
Sequence flow
Annotation
S
M
C
Sensing flow
Control flow
Mechanism flowStart
event
End event
Start message
DurationMessage flow D
Text
End message
Association
Capturing Control Flow
23. BusinessMonitoring
Sell Products
Online
Analyze &
Interpret
Decide
Managerial
Action
Personalize
Create
(Query)
Execute Analyze
Generate
(Ad hoc
report)
Adjust?
Analyze
(Reporting
Requirement)
Design Develop Test
Deploy
(Standard
report)
Data
Warehou
sing
Drill Down
C
Continued
D= minutes
Operational
Process
AdhocReporting
Standard
Reporting
D= seconds ~
minutes
D= hours
D= days
Report interactivity
Managerial Action
Standard reports Ad hoc reports
Ad hoc request
Report requirements
Operational data
S
C
S
M S
M
M
S
Report requirements
S
Identifying Sense-Control Pairs
24. BusinessMonitoring
Sell Products
Online
Analyze &
Interpret
Decide
Managerial
Action
Personalize
Create
(Query)
Execute Analyze
Generate
(Ad hoc
report)
Adjust?
Analyze
(Reporting
Requirement)
Design Develop Test
Deploy
(Standard
report)
Data
Warehou
sing
Drill Down
C
Continued
D= minutes
Operational
Process
AdhocReporting
Standard
Reporting
D= seconds ~
minutes
D= hours
D= days
Report interactivity
Managerial Action
Standard reports Ad hoc reports
Ad hoc request
Report requirements
Operational data
S
C
S
M S
M
M
S
Report requirements
S
Identifying Feedback Loops
Monitor
Analyze Plan
Execute
25. A Hierarchical View of
Design Processes
Adaptation loops reveal special relationships among
processes
Higher-level process – control/design/change process
Lower-level processes – target/use/etc. processes
Change though
Control – constrains the options for the target process. Adaptation.
Mechanism – changes the space of options for the target process.
Evolution
Result – hierarchy of processes reflecting their control
order.
These also help when change cannot be accommodated at runtime
(e.g., when we need to design new capabilities)
RCIS 2013, 29.05.201325
28. 28
Controller/Designer procs vs. controlled/designed procs
• Controller/Designer (C/D) procs – executed less frequently
• Decisions in C/D procs – higher-level
• Different responsible roles
• Different decision criteria
• …
29. 2: Moving to Self-Service BI
RCIS 2013, 29.05.201329
Handling
PUs
Traditional
BI
Self-Service
BI
Tailoring to
CUs
1
2
3
5
*not shown in the presentation
*
…
(Further iterations
may follow)
In response, vendors came up
with Self-Service BI Tools: shorten
ad hoc report design cycle
• BU – given the self-service tool
• BI Team – repository of
standard reports and tool
training
Leveraging
SUs
4
*
30. 2: Moving to Self-Service BI
RCIS 2013, 29.05.201330
Handling
PUs
Traditional
BI
Self-Service
BI
Tailoring to
CUs
Leveraging
SUs
1
2
3
4
5
*not shown in the presentation
*
…
(Further iterations
may follow)
In response: vendors came up
with Self-Service BI Tools, shorten
ad hoc design cycle
• BU – given the self-service tool
• BI Team – repository of
standard reports and tool
training
What actually happened:
• Wide-ranging needs of BUs + unrealistic
expectations of BUs’ capabilities → self-
service tool too complex or too restrictive.
Reverted to Solution 1.
31. 2: Moving to Self-Service BI
RCIS 2013, 29.05.201331
Handling
PUs
Traditional
BI
Self-Service
BI
Tailoring to
CUs
1
2
3
4
5
*not shown in the presentation
*
…
(Further iterations
may follow)
Leveraging
SUs
4
*i* and BPMN Modeling
• As we will see in the next iteration, social
models could have anticipated the failure
and can be used to “diagnose” it.
• BPMN models: can’t capture social
aspects, but can analyze shorter report
design cycle.
32. 3: Tailoring to CUs
RCIS 2013, 29.05.201332
Handling
PUs
Traditional
BI
Self-Service
BI
Tailoring to
CUs
1
2
3
5
*not shown in the presentation
*
…
(Further iterations
may follow)
In response: self-service solution
to Casual Users + customizable
standard reports + ample training
• CU – new role (BU with limited
report creation capabilities).
Needs simple and easy-to-learn
tools.
Leveraging
SUs
4
*
33. 3: Tailoring to CUs
RCIS 2013, 29.05.201333
Handling
PUs
Traditional
BI
Self-Service
BI
Tailoring to
CUs
Leveraging
SUs
1
2
3
4
5
*not shown in the presentation
*
…
(Further iterations
may follow)
In response: self-service solution
to Casual Users + customizable
standard reports + ample training
• CU – new role (BU with limited
report creation capabilities).
Needs simple and easy-to-learn
tools.
What actually happened:
• Assumption of analytical aptitude fails for
many CUs + the tool is difficult even with
training → limited adoption.
34. 34
With i*, we are able to capture:
Objectives from the p.o.v. of CUs w.r.t. the tool and assumptions about
their capabilities w.r.t. the tool and assumptions about their capabilities.
NOTE: Physical actors (agents) and logical actors (roles) are distinct in i*. A capability is a property of an
agent. Physical agents may fail to possess skills required for playing logical roles.
CU
BI Team
DStandard
Reports
Ad hoc
result
D
D D
D
D
Access pre-
built standard
reporting
Provide standard
reporting
DW
Ad hoc analysis
be conducted
D
D
Ad hoc
authoring
tool
Execute BI
query
D
D
Be
implementedD D
Monitor
business
Rely on self with
authoring tool
Meet business
information needs
Easy to use
[tool]
Perform BI
query
Easy to use
[tool]
D
D
Right
time
Help
Help
?
Learn tool
Analytical
aptitude
Quick to learn
[tool]
Help
Help
?
?
?
Zero request
backlog
Help
Ad hoc analysis
be supported
Provide ad hoc
authoring tool
Formulate BI
query
Std. report
requirements
Tool training be
provided
3: Tailoring to CUs: Modeling
35. 35
CU
BI Team
DStandard
Reports
Ad hoc
result
D
D D
D
D
Access pre-
built standard
reporting
Provide standard
reporting
DW
Ad hoc analysis
be conducted
D
D
Ad hoc
authoring
tool
Execute BI
query
D
D
Be
implementedD D
Monitor
business
Rely on self with
authoring tool
Meet business
information needs
Easy to use
[tool]
Perform BI
query
Easy to use
[tool]
D
D
Right
time
Help
Help
?
Learn tool
Analytical
aptitude
Quick to learn
[tool]
Help
Help
?
?
?
Zero request
backlog
Help
Ad hoc analysis
be supported
Provide ad hoc
authoring tool
Formulate BI
query
Std. report
requirements
Tool training be
provided
3: Tailoring to CUs: Modeling
With i*, we are able to capture:
Objectives from the p.o.v. of CUs w.r.t. the tool and assumptions about
their capabilities.
Failures to achieve CUs’ goals and wrong capability assumption.
36. 36
CU
BI Team
DStandard
Reports
Ad hoc
result
D
D D
D
D
Access pre-
built standard
reporting
Provide standard
reporting
DW
Ad hoc analysis
be conducted
D
D
Ad hoc
authoring
tool
Execute BI
query
D
D
Be
implementedD D
Monitor
business
Rely on self with
authoring tool
Meet business
information needs
Easy to use
[tool]
Perform BI
query
Easy to use
[tool]
D
D
Right
time
Help
Help
?
Learn tool
Analytical
aptitude
Quick to learn
[tool]
Help
Help
?
?
?
Zero request
backlog
Help
Ad hoc analysis
be supported
Provide ad hoc
authoring tool
Formulate BI
query
Std. report
requirements
Tool training be
provided
3: Tailoring to CUs: Modeling
With i*, we are NOT able to capture:
The difference in frequencies of Learn Tool and Execute BI Query cannot
be captured.
Gradual learning and skill improvement and tool adjustment cannot be
easily represented.
37. 3: Tailoring to CUs
RCIS 2013, 29.05.201337
Handling
PUs
Traditional
BI
Self-Service
BI
Tailoring to
CUs
1
2
3
4
5
*not shown in the presentation
*
…
(Further iterations
may follow)
Leveraging
SUs
4
*i* and BPMN Modeling
• i* could have predicted some problems with
solution adoption and can help with post
mortem.
• BPMN models: can’t capture social
aspects, but can analyze shorter report
design cycle.
38. Conclusions
Based on our experience with the case study
The need for modeling ongoing adaptation and change in the face
of uncertain/evolving requirements is a fact of life for enterprises.
Co-evolution and alignment of the social and technical components
of solutions is important.
Neither the i* social modeling notation nor the (augmented) BPMN
notation is adequate.
Aspects of modeling and analysis that are important to
support:
Variability Modeling and Binding, criteria for alternative selection,
barriers to adoption/change.
Social Modeling – physical vs. logical actors, skills/capabilities,
personal goals, incentives.
Feedback (failures, changes in context, requirements), multiple
levels of design – changes within/across levels, iterations.
Temporal and dynamic aspects – frequencies, duration, etc.
38
39. Future Work
Further case studies.
We are aiming at proposing a modeling and analysis
approach that addresses the challenges identified in this
paper.
Working on the idea of enterprise decision space
Requirements-driven, goal-oriented
High-variability, explicit capturing of decision criteria
Many linked decisions at multiple levels
Feedback
Informed by BI
RCIS 2013, 29.05.201339
41. Bibliography
[1] W. Eckerson, “Business-driven BI using new technologies to foster
self-service access to insights,” TechTarget, 2012.
[2] B. Evelson, “The Forrester wave: self-service business intelligence platforms,
Q2, 2012”, Forrester Research, 2012.
41
Notes de l'éditeur
Transforming raw data into meaningful business information. Gives companies critical competitive advantage. Organizational/socialAnd technical components.
We will see how augmented BPMN models could be used to model the dynamics of this solution.
To evaluate enterprise adaptiveness, it’s important to evaluate monitoring and change processes w.r.t. to their operational processes. E.g., the operational process runs in seconds (and we may have 1000’s of them) and if the monitoring/change process takes days, than the org. will not be able to adapt the operational process when needed.
Control flow – usually constrains the options for the target process. Adaptation.Mechanism flow – changes the space of options for the target process. Evolution.
Note: 2 processes involved in the feedback loop. One controls another.
Need to focus on physical actors, not just on abstract roles.