Transcript Presentation

Beyond Data Warehousing:
What’s Next in Business Intelligence?
Matteo Golfarelli, Stafano Rizzi
DEIS – University of Bologna
Iuris Cella
Gruppo Formula S.p.A.
Summary:
1. Metric Driven Management
2. Business Performance Management
3. Proposed Architecture
4. Research and Technological issues
5. Conclusions
DOLAP 04
History table for DWing
• Late ’80s: the request for efficiently and effectively
analyzing the enterprise data becomes a pressing matter
for managers
• Early ’90s: the birth of DWing and OLAP in the industrial
world
• Mid ’90s: academic world becomes interested in Dwing
• Late ’90s: metrics driven management becomes a
pressing matter for managers
• Early ’00s: the birth of Business Performance Management
in the industrial world
• Mid ’00s: What happens in the academic world?
Metric Driven Management
I
• Process-oriented management couples the organizational
structure with a set of inter-division processes.
– The organizational structure is a hierarchy of divisions aimed at
defining duties and responsibilities.
At this level the global strategy of
the enterprise is decided
Strategic
level
Tactical
level
Operational
level
Information System
It is composed by multiple
divisions, each controlling a set of
functions; the decisions taken here
are related to the corresponding
functions and must comply with the
strategy defined at the upper level
Where the core activities are
carried out; the decision power is
limited
to
optimize
specific
production activities in accordance
with the main strategy.
Metric Driven Management
II
• Process-oriented management couples the organizational
structure with a set of inter-division processes.
– Processes identify a set of related tasks performed to accomplish
a defined goal. Tasks are carried out by different divisions and at
different levels.
Strategic
level
Tactical
level
Operational
level
Information System
Processes focus on the global
business goal rather than a single
divisional
task.
Of
course,
employees involved in processes
must share the business strategy in
order to synchronize their behavior
Metric Driven Management
III
• Metrics (KPI) are used to:
– continuously measure process performances.
– share the top-level business strategy with the employees in order
to synchronize their behavior.
Strategy
target values
for indicators
Strategic
level
Daily export sales
=10000$
Tactical
level
Calls to prospects
= 200 per day
Order latency
= 2days
actions/
decisions
Operational
level
actions/
decisions
DW
Information System
current values
for indicators
Business Performance
Management (BPM)
• The term BPM defines the new approach to management.
The peculiarities that distinguish BPM from classical DWbased BI applications are:
– Users: are decision-makers but at the tactical and operational
levels.
– Delivery time: decision at the lower levels must be faster than the
strategic ones. BMP systems are not supposed to operate in realtime, but rather in right-time.
– Information coarseness and lifetime: information are usually more
detailed since concerns events related to specific tasks. Lifetime of
information is limited according to right-time requirement.
– User interface: information are mainly accessed via dashboards
and reports as well as through automated alerts activated by
business rules.
Architecturual sketch for BPM
• The technology implementing the BPM flow is often called: Business
Activity Monitoring (BAM)
OLAP
Right-Time Integrator (RTI):
integrates data from operational
data sources at right-time
Dynamic Data Store(DDS): is a
repository capable of storing shortterm data for fast retrieving
report
dashboar
d
user interface
alert
BAM
Data
Warehouse
Rule engine
metadata
repository
KPI Manager: computes all the
indicators necessary at different levels
to feed dashboards and reports
Mining tools: capable of extracting
relevant patterns out of data streams
Rule engine: that continuously monitors
the events filtered by RTI or detected
by mining tools to deliver timely alerts
metadata
repository
KPI
manager
Mining tools
ODS
DW
ETL
RTI
ERP
other DBs
source data
business domain
EAI
DDS
data
streams
Research and technological issues I
Data latency reduction
• Right-time constraint makes classical ETL and ODS
approaches unfeasible
• On-the-fly techniques are needed:
– Cleaning techniques devised so far rely on the presence of a
materialized integrated level.
– Data stream manipulation still presents many challenges (i.e
complex queries over data are performed offline, real-time queries
are usually restricted to simple filters).
Research and technological issues II
Informative power
• Complex KPI systems have been devised in the economic field (e.g.
balanced score card) but the BI community has only partially faced the
problem of their modeling and handling
• System dynamic tools are used within the economic field but their
exploitation within the BI tools has not been considered yet.
Input
3
KP2
KP3
Input
2
KP1
Input
1
KP4
KP5
• When the set of KPI is large (i.e. hundreds or thousand) it is hard to
understand which their relationships are and if the targets are
consistently defined. To this end simulation tools are needed.
Research and technological issues III
Informative power
• Most of the mining techniques devised so far are inapplicable in the
right-time assumption.
• Indicators and mining tools may need short-term information to
compute complex information. Since simple buffering techniques are
not appropriated, main-memory databases (real-time databases) must
be considered.
Research and technological issues IV
Interface
• The classical paradigms of DW systems, namely reporting and OLAP,
are no more sufficient to give users a full picture of the trend of their
business in the short and medium term.
• New paradigms, with different characteristics, must be merged into a
common interface
Interface
Structure Freshness Interaction
Information
Report
Static
Short-time
Pull
Measures/Indicators
OLAP
Dynamic
Short-time
Pull
Measures
Dashboard
Static
Right-time
Pull
Indicators
Alert
Static
Right-time
Push
Event
Research and technological issues V
Design
• Right-time: the designer must determine what is the meaning of righttime for the specific business domain
• Light-architecture: the requirements for KPIs change quickly
depending on the pursued strategies that will bring to the foreground
new functions and behaviors monitored by KPI.
• Process design: with respect to DW, BPM also requires to understand
processes and their relationships in order to find out relevant KPI and
Business rules.
• KPI design: capturing and modeling the relationships between
different indicators has a primary role in this phase to ensure that
effective and reliable information is delivered.
Conclusions and discussion
• We summarized the requirements emerging from modern
companies and discussed how they meet into a new
architecture that promises to lead Business Intelligence
beyond data warehousing.
• We saw that different sophisticated technologies are
involved in BPM. Most of these fields are not mature
enough in terms of commercial products.
• Gartner group indicates BPM and BAM as the emerging
market in Business Intelligence till 2008.
Is it just a passing fad ?
Is it a technology matter only ?
Is it the second era in BI ?