Moss L.T., Atre S. Business intelligence roadmap: The complete project lifecycle for decision-support applications

Подождите немного. Документ загружается.

through data mining.

- Marketing costs should be reduced as a

result of data mining efforts.

- New customers should be acquired as a

result of data mining efforts.

However, it is challenging to know when to attribute

these changes to the data mining efforts and when

these changes would have occurred naturally

without the data mining efforts.

14. Meta Data

Repository Development

Provide interactive and context-sensitive meta data

query capabilities. Using the meta data repository

should be easy and intuitive.

Keep the meta data repository in synch with the

meta data contained in other tools and in the DBMS.

Unfortunately, meta data repositories are still

passive, which means that the synchronization has

to be performed manually or through custom-

written programs.

Actively maintain the meta data repository. If the

content of the meta data repository becomes stale,

it will affect the business people who are relying on

the completeness and accuracy of the meta data. It

will make the meta data repository questionable,

and the business people may stop using it.

Don'ts

Don't underestimate the effort required to build and

maintain a meta data repository, especially if you

choose a decentralized or distributed approach.

Building gateways and portals are not trivial

activities.

Don't attempt to develop a meta data repository

with part-time people who will leave after the meta

data repository is implemented. You need at least

one dedicated person to maintain the meta data

repository on an ongoing basis.

Don't overlook existing sources of meta data.

Organizations are sometimes drowning in

documentation. The challenge will be to separate

the valuable current information from outdated

information that is no longer valid.

Tips and Rules of Thumb

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

591/631

Developing a meta data repository solution is not an

event—it is an evolution. It starts with the first BI

decision-support initiative and continues to evolve

on an ongoing basis. Whenever new business data

is added to the BI target databases, new meta data

is added to the meta data repository (for example,

names, definitions, and domains of the new

business data). Whenever new functionality is

added to the BI applications, new meta data is

added to the meta data repository (for example,

calculations and reconciliation totals for the new

functions).

Be prepared to develop two types of meta data

repository interfaces:

- Access interface for business people and

technicians

- Tool interface for sharing meta data between

the meta data repository and the tools where

meta data is originated

Be sure to understand the import/export features of

the tools from which you will extract meta data, for

example, CASE, ETL, OLAP. Depending on how

much or how little commonality there is among the

tools, you may have to write several different tool

interface programs.

The estimated size of a meta data repository could

reach 5 percent of the total size of BI target

databases, if done correctly and diligently.

Keeping the meta data repository up to date and

synchronized with other tools, databases, and

programs is a challenge. It will require collaboration

from all stakeholders on all BI projects. The bigger

challenge is to get the operational systems people

to communicate any changes to their source data

immediately. Changes to source data may also

affect the ETL process and the design of the BI

target databases.

Remember that for every day of coding, you will

probably spend at least three days on testing.

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

592/631

15. Implementation

Work closely with the operations staff in creating

the production program libraries and moving all

programs into the production environment.

Include regular database maintenance activities in

the job schedule, such as database backups and

reorganizations.

Use appropriate monitoring and alert utilities to

detect and diagnose resource problems early.

Work with your organization's security officer to

prepare a security gap analysis matrix, and make

sure that only authorized persons can access the

data that is intended for their use.

Don'ts

Don't try to roll out a BI application to the entire

organization all at once. Use an incremental

implementation approach, learn as you go, and

make adjustments where needed (for example,

provide training earlier in the process or change the

log-on procedures).

Don't skimp on support staff. A BI decision-support

environment is very complex, and the business

people will need mentoring.

Don't offer open Internet access to the BI target

databases without properly tested security

measures. Improper security measures could allow

competitors to view sensitive organizational data,

allow unauthorized people to see customer data, or

allow someone to misuse or abuse displayed

information. In the worst case, you may also be

liable for fines, legal consequences, and so on.

Tips and Rules of Thumb

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

593/631

The standard engineering maxim for sizing

databases is to estimate the volume of your

business data, then triple it to get a realistic size (if

a lot of indices are created, quadruple it). The

allocation of space is often distributed as follows:

- 30 percent business data

- 30 percent indices

- 30 percent summaries, aggregates

- 10 percent miscellaneous

Based on the experience of other organizations, you

can assume that your BI decision-support

environment will double in size every two years.

Some believe this is a conservative estimate.

Many BI target databases are by their nature

VLDBs. Therefore, stop conserving disk space if the

BI application is helping the organization to make a

profit.

Because of the immense size of VLDBs, backing up

entire tables all at once may not be possible.

Consider incremental ("net change") backups, high-

speed mainframe backups, or partial backups of

partitions.

Monitor computer utilization, network utilization,

and personnel utilization closely to avoid

unexpected bottlenecks.

16. Release Evaluation

Conduct the post-implementation review meeting

offsite to eliminate interruptions.

Invite all project team members (from the core

team as well as the extended team). You may invite

additional stakeholders as observers if they have a

vested interest in the lessons learned from the BI

project.

Stay focused on the agenda during the meeting.

Schedule a second meeting if the topics cannot be

covered adequately in the time allocated.

Send out a preliminary agenda early and invite

attendees to add items to it.

End on a positive note and with assigned action

items for future development improvements and

possible business process improvements.

Don'ts

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

594/631

Don't pretend you are using the release concept if

you don't follow the BI application release

guidelines. To simply hurry an incomplete project to

implementation with little thought, no structure,

and poor quality is not in the spirit of the release

concept.

Never accept the excuse that "nobody has time for a

one-day post-implementation review." This

indicates a lack of commitment from management

to improve the quality and speed of the

development process, which they claim they want.

Don't review voluminous documentation because

that is unproductive. However, if some documents

will be referenced during the meeting, send those

documents out ahead of time so that the attendees

can read them before the meeting.

Don't let the meeting deteriorate into a "finger-

pointing" session.

Don't postpone the meeting beyond two to three

months after implementation because people will

become disinterested and will forget the issues.

Don't use anyone from the core team to be the

facilitator or the scribe. Core team members are

major players and contributors during the post-

implementation review. Their level of participation

would be diminished if they had to lead the session

or take notes. Have a trained third-party facilitator

conduct the meeting, and have a third-party scribe

take notes during the meeting.

Tips and Rules of Thumb

Implement your BI applications using the release

concept. It is much better to deliver high-quality,

partially functioning application releases over time

than to deliver a low-quality, completed application

that is fraught with many defects and with dirty

data. If the first release is successful, new

requirements will emerge as the business people get

used to the iterative development process.

Post-implementation reviews should always be

performed after each BI project. The purpose for

these reviews is to document "lessons learned" and

to improve the BI development approach. These

lessons could also be shared with other project

teams and business managers.

[ Team LiB ]

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

595/631

[ Team LiB ]

Appendix Work Breakdown

Structure

The work breakdown structure in this appendix reflects the contents of the enclosed

CD-ROM.

[ Team LiB ]

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

596/631

[ Team LiB ]

Task_Data

Task_Name

Predecessors

Your Project Name

Step 1: Business Case Assessment

Determine the business need

Identify the business need

Determine current financial consequences of the

business need

Assess the current decision-support system (DSS)

solutions

Assess current usage of the existing DSS

Determine the shortcomings of the existing DSS

Perform gap analysis

Assess the operational sources and procedures

Assess the data quality of operational systems

Review file structures and databases

Review content (domain) of source data elements

Assess the current data movement

Review data entry practices

Review data extraction practices

Review data manipulation practices

Review data duplication practices

Assess current operational procedures

Identify poor data entry practices

Identify lack of edit checks

Identify defective program code

Identify lack of training

Assess the competitors' BI decision-support

initiatives

Determine the competitors' successes and failures

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

597/631

Determine whether competitors gained market

advantages

Determine the BI application objectives

6, 10, 24

Identify the strategic business goals of the

organization

Define the overall BI decision-support objectives

Define the project-specific BI application objectives

Match the overall BI decision-support objectives to

the strategic business goals

Match the project-specific BI application objectives to

the strategic business goals

Propose a BI solution

Review current DSS solutions

Review DSS gap analysis

Determine how the BI application will lessen the

business pain

Create a high-level architecture for the proposed BI

solution

Consolidate and prioritize unfulfilled requirements

from previous BI projects

Create a high-level (conceptual) logical data model

Perform a cost-benefit analysis

Determine costs

Determine benefits (tangible and intangible)

Identify short-term benefits to the organization

Identify long-term benefits to the organization

Calculate the projected return on investment (ROI)

41, 42

Perform a risk assessment

Create a risk assessment matrix

List the technology risks

List the complexity risks

List the integration risks

List the organization risks

List the project team risks

List the financial investment risks

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

598/631

Assign weights to the risks

Rank the risks: low, medium, or high

Determine the risks (ramifications) of not

implementing a BI solution

Write the assessment report

33, 40, 46

Describe the business need

Describe lost opportunities

Describe the proposed BI solution

State the cost justification and expected ROI

Include risk assessment results

Write recommendations (include operational business

process improvements)

Obtain project approval from business sponsor

Step 2: Enterprise Infrastructure Evaluation

Section A: Technical Infrastructure Evaluation

Assess the existing platform

Review hardware

Review operating systems

Review middleware, especially DBMS gateways

Review custom interfaces

Review network components and bandwidth

Review the DBMS

Review tools (CASE, ETL, OLAP, etc.)

Review the meta data repository

Perform gap analysis

Evaluate and select new products

Identify the product categories you need to evaluate

(hardware, DBMS, tools)

List all products being considered for each category

Itemize your requirements for the products

Weigh each product requirement (scale of 1 to 10)

Rank each product against the weighted

requirements (scale of 0 to 10)

Determine the total score for each product

北斗成功社区 BeiDouWeb.com 教育音视频/电子书/实用资料文档/励志音乐影视仅供免费试用/版权原著所有

599/631

List all vendors of all products

Itemize your requirements for the vendors

Weigh each vendor requirement (scale of 1 to 10)

Rank each vendor against the weighted requirements

(scale of 0 to 10)

Determine the total score for each vendor

Evaluate the product scores and vendor scores

83, 88

Create a short list of products and vendors in each