v Performance Measurement Capability A Data Warehouse Business Architecture
Nov 01, 2014
v
Performance Measurement Capability
A Data Warehouse Business Architecture
v
Balanced Scorecard Activity Based Management
Performance Measurement Approaches
Robert S. Kaplan & David P. Norton “Mastering the Management System”, HBR, Jan 2008.
v
Performance Management Capability
The performance management domain defines the set of capabilities supporting the extraction, aggregation, and presentation of information to facilitate decision analysis and business evaluation
Capability Description
Analysis
& Statistics:
Defines the mathematical and predictive modelling and simulation capabilities that
support the examination of business issues, problems and their solutions
Business
Intelligence
Defines the forecasting, performance monitory, decision support and data mining
capabilities that support information that pertains to the history, current status or future
projections of an organization.
Visualization: Defines the presentation capabilities that support the conversion of data into graphical or
pictorial form.
Reporting: Defines the ad hoc, standardised and multidimensional reporting capabilities that support
the organization of data into useful information.
Data
Management:
Defines the set of capabilities that support the usage, processing and general
administration of structured and unstructured information.
FEA Consolidated Reference Model Document v 2.3
v
Business Measures
%Revenue by market segment
%Revenue by top 20 clients
%Revenue by client relationship
Increase key account / high margin clients Customer
Perspective
£Sales revenue by market segment
Number of new projects by top 20 clients
Revenue by top 20 clients (client value)
Product
Time Period
Region
Employee
Customer
£ Sales Income / Revenue
Calc. = quantity price
Target =
Alert Threshold =
v
Da
ta W
are
ho
us
e A
rch
itec
ture
Presentation Layer
Std Reports Analytics
%Revenue by market segment
%Revenue by top 20 clients
%Revenue by client relationship
ODS Data Marts
1. BI Presentation
3. Data Warehouse
4. Reconciliation Process
5. Operational Systems
T
L
E
2. Metadata Repository
v
Da
ta W
are
ho
us
e A
rch
itec
ture
ODS Data Marts
1. BI Presentation
3. Data Warehouse
4. Reconciliation Process
5. Operational Systems
BI Presentation Layer
? Ad Hoc Query Metadata Std Reports Analytics
T
L
E
2. Metadata Repository
v
Reference Architecture Components
Component Description
Business Intelligence
Presentation Layer
The presentation layer is responsible for providing tools for delivering ad hoc, standard and
analytical reporting. The reporting tools available fall under the business intelligence umbrella
(BI). These tool support access to and analysis of information to improve and optimize
decisions and performance, i.e. data mining, analytical processing, reporting & querying data..
Information Catalogue The information catalogue (data dictionary) component is responsible for maintaining the
definition of data and its lineage from the source systems through to the data warehouse. This
incudes data definitions, data mapping and transformations conducted on the data.
Data Warehouse
Data Mart
The data mart component is responsible for delivering line of business, departmental and
individual information needs and key performance indicators. These information needs are
reported as facts, allowing the data to be reported against standard dimensions, such as,.
Customer segment, product, organisation structure, location and time.
Data Warehouse
Operational Data Store
The operation data store (ODS) component is responsible for holding historic atomic data
extracted from operational systems. This data is held in non-redundant third normal form
arranged by subject area. It contains static near current data which is refreshed on a regular
basis from the source operational systems, e.g. daily, weekly or monthly. It is used to support
all decision support reporting needs.
Data Acquisition
Extract, Transform & Load
Data reconciliation component is responsible for data acquisition and resolving consistencies
and discrepancies between common data elements stored across the source systems, e.g.
reference codes, spelling & field lengths. The reconciliation process is conducted in a separate
staging area where the extracted data is reformatted, transformed and integrated into an agreed
common data model.
Operational Systems The transactional processing systems used to support the business operations of the
enterprise. These operational systems provide the primary data used for decision support and
reporting. This data is dynamic and constantly changing with each business transaction. Bill Inmon and Gartner
v
BI: Data Quality Scorecard Specification Approach
Business Measure - Information Need
Business Measure: Data Quality
Types 1. Actual
2. Target ± tolerance
Dimensions: Agency Data Item Location
Channel Attribute Post code
Segment Entity Statistical Area
Organisation Data Collection
Outlet
Calculations: % Master data duplication
% Collection submission data completeness
% Data item accuracy
% Consistency across data sets
Statutory timeline aging of collection receipts
Time Dimension: Weekly
Monthly
Year to date
Atomic Data: Agency
Agent Collection
Data Item
Attribute
Entity
Reporting Period
Data Submission
Validation Result
Rule
v
Summarised Data Store: Modelling Approach
Business Measure
Data Model
• Identify business measure (fact)
• Define measure formulae
• Identify measure dimensions
• Identify measure source data • Entity
• Attributes
• Maintain measure dimension affinity matrix
Business Measure
Database Design
• Design summarised database • Star Schema
• Snowflake Schema
• Prepare use case specification
Ralph Kimbal
v
High Level Data Model
• List in scope entities • Object, place, resource or
event
• All entities at the same level of abstraction
• Entity relational model structured by subject areas
• Defines scope of integration
Mid Level Data Model (DIS)
• Third normal form ERD • Remove repeating groups
• All attributes are dependant on the primary key
• Resolve all M : M relationships
• Add sub types where relevant
• Includes all data elements (data item set)
• Primitive data elements only, no derived data
Low Level Physical Model
• Derived from the DIS
• Identify primary keys
• Add alternate keys
• Define physical fields • Desc, field type & size
• Default values
• Value constraints
• Null value support
• Identification of system of record for all fields (data mapping)
• Definition of access method (sequential or random)
• Process data mapping (frequency & fields used)
Operational Data Store: Modelling Approach
Bill Inmon, “Information Engineering for the Practitioner”, Yourdon Press, Englewood Cliffs, N.J., 1988
v
Reconciliation Process Data Acquisition Approach
Data Mapping
• Identify source system fields
• Map source fields to target data model
• Define data transformation rules
• Determine interface services
• Prepare use case specification
Data Quality
• Determine quality grading scheme, e.g. • Platinum
• Gold
• Silver
• Define data quality measures
• Define quality measure formulae
• Identify quality measure dimensions
• Identify quality measure source data • Entity
• Attribute
v
Data Validation ETL Use Cases
The Solution
Data Collection Custodian
Monitor Data Quality KPIs
Maintain Reference Data
Assign Agency
Collection
Maintain Agency
Map Entity Collection Data
Define Validation Rule
Load Data Submission
Validate Data Submission
Notify Late Collection Submission
Assign Data Item Rules
Turn Off Agency Rule
Agency Submission Due Date
Agency
Record Submission Exemptions
Help Desk
v
Contact
Technology architecture & solutions are justified at a strategic and financial level by preparing a business case.