Swan(sea) Song – personal research during my six years at Swansea ... and bey...
12209508.ppt
1. BI – Preliminary Information on
the course
Prof. Paola Velardi
Dipartimento di Informatica
Sapienza Università di Roma
2. Why this course
• Enterprises today are driven by data. "Business Intelligence allows
people of all levels in organizations to access, interact with, and
analyze data to manage the business, improve the performance,
discover opportunities, and operate efficiently" (Cindi Howson,
Successful BI, McGrawHill).
• However, the degree to which BI solutions can be successfully
adopted within organizations depend to a great extent on the
degree to which business and IT experts can partner together.
• The objective of this course is to form Advanced Business Users of
BI applications, with a deep understanding of the business needs
and a good understanding of technology.
• The Advanced Business User understands the business and how to
leverage technology to improve it, leads the interpretation of
business requirements and strategic objectives, and helps designing
reports to answer business questions.
3. Organization
• The course is in two parts:
– PART A: Business Intelligence and Social Analytics
(Instructor: Prof. Paola Velardi velardi@di.uniroma1.it)
6CFU (MON-TUE)
– PART B: Process modeling (Instructor: Prof. Paolo
Bottoni bottoni@di.uniroma1.it) 3CFU (WED)
• Course syllabus, material and slides available on
the web page:
http://twiki.di.uniroma1.it/twiki/view/BI/WebHo
me
• Subscribe to google group!
5. Example: supply chain management
• Supply chain management is the integration of the
activities that procure materials and services and
deliver them through a distribution system
• It typically implies the following actions/actors:
1. Transportation vendors
2. Credit and cash transfers
3. Suppliers & Distributors
4. Accounts payable and receivable
5. Inventory
6. Order fulfillment
7. The traditional ICT tools for handling
SCM
• ON-LINE TRANSACTION PROCESSING or
OPERATIONAL TRANSACTION PROCESSING)
• “On line transaction processing, or OLTP, is a
class of information systems that facilitate and
manage transaction-oriented applications,
typically for data entry and retrieval transaction
processing. “ (wikipedia)
• Operational, or online transaction processing
(OLTP), workloads are characterized by small,
interactive transactions that generally require
sub-second response times.
8. Example of a typical OLT interface
Shipment number, status, carrier name, shipping state, date, source,
destination, package type..
9. What kind of info can you get from this
database?
• Shipments (transactions)
• Destinations
• Earnings..
• Everything according to time, geographic areas,
etc. (e.g., “How many shipments to Roma in
July?”)
• Data come from inside the company and are
mostly manually entered. Data processing is with
traditional database systems (SQL). Interfaces are
sometimes well designed , but usually not so
exciting.
11. Study Case one: Magpie Vaccine
supply chain (1)
• Cold chain in healthcare is the
temperature-controlled supply
chain in trasporting and storing
vaccines and drugs
• Maintaining cold chain is
extremely important for these
products
• A study of CDC (Center for
Desease Control in US) revealed
that ¾ of providers experienced
serious cold chain violation
12. Study Case one: Magpie Vaccine
supply chain (2)
• Magpie sensing (a start-up
project) is using cold chain data
analytics and business
intelligence to improve vaccine
supply chain management
• They gather location-tagged
information about cold storage
units using a shippable wireless
temperature and humidity
monitor system
• Real-time, fixed-location
monitoring device then
safeguards products from their
arrival at a lab or clinic until the
moment of use
13. Study Case one: Magpie Vaccine
supply chain (3)
• At the core of Magpie Sensing's solutions are
analytics algorithms which leverage data from
monitoring devices to improve cold chain
processes and predict cold storage problems
before they occur.
14. Study Case one: Magpie Vaccine
supply chain (4)
• Three types of analytics are provided:
• Descriptive: set point of the system's thermostat, typical range of
temperature values in the system, the duty cycle of the system's
compressor, etc. This informs the user as to whether their storage
unit is properly configured to store a particular product.
• Predictive. Predictive analytics detect cold chain problems (e.g.
poor equipment configuration, human error, equipment failures)
and send an alert before temperature bounds are violated.
• Prescriptive: provide recommendations to improve cold storage
processes and business decisions during normal operation (e.g.,
optimal thermostat setting for a particular product)
16. What’s new?
• Data come from several sources (in addition to
company data, we have sensors , methereological and
geographical data)
• Data are aggregated to obtain additional information
(e.g., times of day that are busiest and the periods
where the system's door is opened the most. This can
guide staff training and institutional policies)
• Not simple queries (descriptive) but also predictive
and prescriptive information. System helps efficient
and timely decision making!
• Interfaces (visualization) is a key to efficient use and
understanding of data
17. Study Case two: Hokuriku Coca-Cola
Bottling Company (1)
• When you select a beverage on a vending
machine, you never think that there might be a
lot of hard work and careful planning by the
bottling companies, who have to decide which
drinks will sell best in their machines while at the
same time eliminating out-of-stocks and reducing
equipment failures.
• Hokuriku used big data and data analytics to
improve efficiency of vending machines. In 2005,
they reported to increase 10% sales and decrease
46% associated costs.
18. Study Case two: Hokuriku Coca-Cola
Bottling Company (2)
• The first step was to provide vending machines with
wireless connections to transmit data
• Data from vending machines and other traditional sources
are collected in a data wharehouse (will se what this is
later: for now, like “a large heterogeneous database of
structured and unstructured data”)
• Wharehouse collects data like: time location and date of
each sale, when a product sells out, or is short-changed,
and when machine is malfunctioning. They also collect
sentiment data from social networks.
• Big data technologies (SAP HANA, Hadoop, and Spark )
have been used to manage and intersect these data
providing descriptive, predictive and prescriptive analyticts
22. What’s new?
• Data come from several sources (in addition to company data, we
have near real-time data from single vending machines and social
data)
• Data are aggregated to obtain additional information (e.g., best
seller by time and location, opinions of users, user’s preferences by
profile, date, location... This can guide stocking policies and
marketing campaigns)
• Not simple queries (descriptive) but also predictive (a product will
shortly sold-out, fine-tuning policies for refilling: not only if close-
to-sold-out, but also based on historical data, consumers’
opinions..) and prescriptive information (targeting users for specific
campaigns). System helps efficient and timely decision making!
• Interfaces (visualization) is a key to efficient use and understanding
of data.
23. Case Study 3
• Lennox International is a $3B manufacturer and distributor
of heating, air conditioning and refrigeration products.
• Lennox describes how it doubled the size of its footprint
with
– Manhattan’s integrated Warehouse Management System
(WMS),
– Transportation Management System (TMS) and
– Labor Management System (LMS).
• See more at: http://www.manh.com/resources/videos/lennox-
internationals-supply-chain-integration#sthash.1kjODMBd.dpuf
24. Case Study 3: Lennox
CHALLENGES ICT SOLUTIONS
RESULTS
25. • As for Magpie and Coca Cola, use telemetry data (wireless transmission of
data), e.g. for tank volumes.
• Use also POS (Point Of Sale) data rather than shipment data only.
• POS data allows to see the real time data driving the supply chain.
Statistical methods can then be applied to bin and cluster the data for
further analysis.
Case Study 3: Lennox (2)
26. Why POS data are helpful?
• They enable us to balance either too much
optimism or pessimism that can come from
shipment trends alone.
• They provide more stable and consistent patterns
that can be easily used, as well as helps to
identify changes in the trend over time. Shipment
data are not true indicators of a change in the
trend because their data bounce around.
• They allow companies to do a “sanity check” on
the final forecast numbers by using trade
inventory (inventory by store) in the data stream.
27. • Determine which distribution centers should service
which stores for which specific products, especially in
large networks, in order to optimize time, fuel and
resources.
• Create the most productive delivery schedule for
fleets, taking into account factors such as opened and
closed stores, route disruptions and seasonal issues
• Factor in delivery window optimization, determining
the best time for product delivery based on store
hours, traffic and other variables.
• NOTE: integrating real time data from POS,
metereology, and local sensors is crucial here.
Case Study 3: Lennox (3)
28. Case Study 3: Lennox (4)
• Lennox almost tripled the percentage of orders that can be delivered the next
morning, yet reduced inventory by almost 20%, even with a 250% increase in
physical locations.
• They improved service levels by about 20%. They also reduced distribution costs as
a percentage of sales by over 15%, increased inventory turnover by more than 20%,
and reduced response time.
29. Analytic system example: view
seasonal demand
Seasonal demand at the item level is identified in specific
locations within regions (system identifyied over 200 micro-climates).
30. What’s new?
• Data come from several sources (in addition to company data, we
use POS data and telemetry)
• Data are aggregated to obtain additional information (e.g., high-
demand products and parts that must be available immediately;
lower-demand items need to be delivered the same day or next
morning. This can guide staff training and institutional policies,
such as accurate stock replenishment at the order-line level)
• Not simple queries (descriptive) but also predictive and
prescriptive information. Advanced analytical techniques to
accurately model and forecast complex and unpredictable demand,
e.g., according to micro-climates . System helps efficient and timely
decision making!
• Interfaces (visualization) is a key to efficient use and understanding
of data
34. Image from Alok Srivastava
Where can we apply Business Intelligence?
Tactical
Apps.
Strategic
Apps.
Suppliers Customers
ERP Applications
SCM Applications CRM Applications
Materials/Component
s
Consumers
Business Intelligence
Supply Chain Enterprise Resource planning
Customer Relations
36. Architecture of a BI system
OLAP: on-line analytical processing
Data selection,
extraction, transformation
and load
Data storage,
in relational or
multi-dimensional
tables
Data visualization
and analytics:
descriptive, predictive
prescriptive
37. Information Technology challenges in a
Business Intelligence System
• Data Sourcing: extracting information from heterogeneous
data such as texts, databases, images, sensors, media files and
web pages
• Data Analysis: Synthesizing useful knowledge from collected
data using IT techniques such as data mining, text
understanding, image analysis and signal processing
• Knowledge extraction: Extracting useful (= both relevant and
frequent) facts and rules and filtering out irrelevant
information;
• Decision Support: employ semi-interactive software to
identify good decisions and strategies
38. Benefits of Business Intelligence
for Business Users
• Improve Management Processes
– planning, controlling, measuring and/or
changing resulting in increased revenues and
reduced costs
– Improve Operational Processes
– fraud detection, order processing, purchasing..
resulting in increased revenues and reduced
costs
• Predict the Future (sales, out-of-stock,…)
39.
40. Yes, but why should you take this
course?
• Because business users are closest to the data going
into BI systems, their involvement is paramount to
successful business intelligence management efforts
• Companies should NOT let the IT department control
of BI projects, particularly if they're setting up a
centralized team to take charge of BI management
enterprise-wide.
• While IT's strengths in data preparation, management
and governance are critical for effective BI initiatives,
treating a BI program as yet another IT project
doesn't bode well for its long-term success.
41. Role of Business users in managing BI
projects
• Define needs: Business users sometimes have a hard time to
adequately translating their requirements for BI and IT developers:
so task 1 for BI users is learn accurately identify the BI goals;
• Be knowledgeable: Managing a BI program takes a special breed of
employee, one with a combination of technical, people and process
management skills. Don’t need to design algorithms nor to be a
programmer, but BE AWARE of the technology (and stay up to
date..)
• Help creating reports and dashboards: Interfaces and dashboards
for BI do not answer only pre-defined queries as “old” OLTP
systems, but even though they are much more flexible, they cannot
manage data they have not been designed for. Business Users
should learn defining the objectives of the analytics: which data and
“what for” (in cooperation with IT people).
42. Business people should NOT be data
consumers!
Easy to lose focus of the real vision and milestones to arrive at the final goal. Efforts can
easily be deemed a success by IT measurements (good infrastructure, performance,
coding, change management, etc.) but a failure to the business users!
44. What are the typical roles in a BI
project team (and who could YOU be)
• Advanced Business User
• Professional Report Author
• Modeler
• Administrator
• Analyst
• Business User
ABU: This person has a deep understanding of the business needs and a good
understanding of technology. The Advanced Business User leads the
interpretation of business requirements and creates reports to answer business questions.
46. Course Syllabus (6 CFU)
1. Data Wharehousing: describe the process used in
developing and managing a data wharehouses
2. Decision Support Systems: how to extract knowledge
from data (with some “contamination” with Machine
Learning students from MD in Computer Science)
3. Social networks and Social Analytics: social media
and web as new source data to better engage clients
4. Business Reporting and Visual Analytics: why visual
analytics is crucial for BU, and what are the new
visualization startegies
5. Lab Application: using IBM Cognos (with IBM tutors)
47. Final Exam
• Homeworks (case studies, tests and practical
exercises) 10-15%
• Written exam (50%)
• Project (35-40%)
48. Homework 1
• Select one BI use cases. Find material on the web
http://www.teradata.com/Resources?AssetType=Case+
Studies
• Answer the following (for each case):
– What was the key problem for the company, when
deciding to adopt BI?
– Which information is provided to the BI analytics tool?
– Give examples of descriptive, predictive and prescriptive
analytics supported by the BI systems.
– Which business performance features are highlighted as
success indicators?
– Any free comment you may want to add.
49. Next monday
• Teams will describe their business case first!
• All must handle two pages report via email
• Submit on shared folder