Mitigate Business Intelligence Project Risks With Rule -PDF Free Download

2019 | 1 views | 21 Pages | 376.98 KB

Mitigate Business Intelligence Project Risks With Rule-Based Audits and Proof-of-Concepts ... Conducting A Data Quality Audit, Intelligent Enterprise, July 10, 2004 Gonzales, Michael L., The No-Sacrifice Affordable Data Warehouse APP, Intelligent Enterprise, October 30, 2004 2. Executive Summary Analytical applications such as those found in Business Intelligence (BI) and warehouse-centric ...

---

Table of Contents
Executive Summary 3
Introduction 4
BI Real World Risks
Real World Example Disparate Data Sources 6
Risk Mitigation 8
The Spiral Approach 8
Rule Based Audits and Proof of Concepts 11
The Purpose of an RBA and POC 11
Rule Based Audit and Proof of Concept Steps 12
Technology That Supports Your Risk Mitigation 14
Rule Based Audits Using BRE 14
Proof of Concepts Using DMExpress 15
Conclusion 19
Appendix A DMExpress Case Study ACNielsen 20
The Challenge 20
The Solution 20
The Benefits 21
Figures and Tables
Figure 1 Analysis Thrashing Page 6
Figure 2 Risk mitigation round Page 9
Table 1 0 RBA and POC comparison Page 11
Figure 3 Sample data flow Page 14
Figure 4 BRE rules definition Page 15
Figure 5 Complex rules with DMExpress Page 16
References
Boehm Barry A Spiral Model of Software Development and Enhancement IEEE May 1988
Gonzales Michael L How To Make BI Less of a Gamble Intelligent Enterprise February 1 2005
Gonzales Michael L Conducting A Data Quality Audit Intelligent Enterprise July 10 2004
Gonzales Michael L The No Sacrifice Affordable Data Warehouse APP Intelligent Enterprise
October 30 2004
Executive Summary
Analytical applications such as those found in Business Intelligence BI and warehouse
centric projects are often plagued with a lack of clear definitions of source data as well
as comprehensive application specifications This creates a significant amount of project
Successfully implemented BI and data warehouse projects have one thing in common
explicit consideration of risk To address BI project risk this author recommends using as
a guide the seminal work of Barry Boehm and his process model the Spiral Approach1
This approach
Fosters the development of specifications that are not necessarily uniform
exhaustive or formal
Incorporates prototyping as a natural part of risk reduction
Encourages re work in the early development stages as better alternatives are
discovered
The Spiral Approach is the only process model that is risk driven All other process
models and software development methods are document driven
Techniques for risk mitigation are woven into the project itself and include Rule Based
Audits RBA and Proof of Concepts POC Use RBAs to prove that sample source
data coupled with known business rules can create your target table s Then follow up
with a POC in order to prove that the results of the RBA can be achieved at production
scale for both data volumes and platform capacity
Business Rules Engine BRE is a tool specifically designed to conduct inexpensive
Rule Based Audits on a laptop For larger more complex audits and especially Proof of
Concepts this author recommends DMExpress2
Boehm Barry A Spiral Model of Software Development and Enhancement IEEE May 1988
Syncsort Inc www syncsort com
Introduction
Successful BI and data warehouse projects share at least one common characteristic
explicit consideration of risk Nothing addresses BI project risks as well as a Rule Based
Audit or Proof of Concept Nothing Not a detailed project plan not expensive
technology not high priced talent
BI projects are peppered with risks from data quality to analytic value and scalability
Furthermore these risks often bring entire projects to a halt leaving planners scrambling
for cover sponsors looking for remedies and budgets wiped out Simply put analytic
applications are some of the trickiest to develop3 They are stuffed with concepts like
slice and dice ad hoc data pivoting and drill through Users often don t know exactly
what they want you to build until they start seeing part of the application This often
requires BI teams to build an application before it s fully defined and specified Couple
this challenge with the data quality problems inherent when sourcing operational systems
scalability in terms of data volumes refresh rates and the potential for risk is very real
This paper describes two methods for minimizing the risk associated with BI iterations
Rule Based Audits and Proof of Concepts Independently used each will help you clarify
and better understand risk points associated with your project Implemented together
they represent a continuum of risk mitigating techniques Their presence in projects is a
demonstration of the professionalism and discipline found only in seasoned successful
BI Real World Risks
At the beginning of a project BI planners might encounter the following response when
asking users to specify parameters for strategic analysis and reporting Well I need a
weekly report that tells me this or that Frustratingly the moment the report is
delivered they might hear Good But can you summarize this column or add another
column Why does this happen It s because users are not entirely sure of the decision
support requirements until they have something in front of them to work with Then and
only then are they better able to clarify their needs
Then there is the matter of the data being sourced The Extraction Transformation and
Loading ETL process is still the most underestimated under budgeted part of most
BI DW iterations And the biggest reason why the ETL portion of a project often raises
more questions than it resolves has to do with a lack of understanding of the source data
Gonzales Michael L The No Sacrifice Affordable Data Warehouse APP Intelligent Enterprise October
During data extraction and transformation there are numerous issues that your team will
encounter Many of these will require you to go back to the user group for further
clarification They include
Multiple meanings for the same data element Once you have isolated the business
requirement and have started to gather the source requirements you may find that a value
defined by one user audience is different for other users who may be accessing the same
data store An example would be values like profit cost or revenue Each of these
values could have different meanings to different user communities Consequently you
will need to document definitions from all groups and see if you can implement
transformation processes to satisfy each
Multiple sources of data elements It is often the case that specific data elements exist
on multiple source systems Your job will be to identify the various sources and discuss
with the users which one is most applicable
Differing levels of history Another challenge you may encounter has to do with
availability of history What if your business requirement calls for four years worth of
history but the best most recent data contains only one year The other three years would
need to be extracted from other data sources possibly of questionable quality
Data cleanliness and accuracy Warehouse data is never perfect Instead you need to
ask yourself is the data clean enough
De integration for audits and validation So now that you have done a wonderful job
at integrating transforming and cleansing the data how do you separate the data for audit
and validation purposes
The descriptions above should not suggest a defined sequence of events The message is
that any single data element can quickly go from a simple transformation into an
explosion of new requirements regardless of the mix or order in which the issues are
confronted
And with each data quality issue uncovered there is a significant amount of thrashing
between the ETL team project designers and end users Refer to Figure 1 Often it
requires everyone to evaluate the new transformation issues before a solution can be
established and sent back to your ETL programmers
The project team designs a system to support a pretty pie chart But if the project team
and its designers never assess the quality of source data necessary for that pie chart then
they are exposing the entire project to great risk Consider this carefully if no one spends
any time to assess the source data quality then it is entirely possible that you will
purchase and install all the technology do all the analysis write all the source to target
code to populate target tables and still fail
Figure 1 Analysis Thrashing
Analysis Thrashing
Multiple Meanings
Multiple Sources
Problems with History
Cleanliness Accuracy
Audit Validation
Real World Example Disparate Data Sources
A project I worked on several years ago convinced me of the value of risk mitigation
The account had 20 different sales applications dispersed around the world leaving
executives unable to report current sales accurately Their goal was not just to correctly
report current sales but all chronological history of sale order line detail changes as well
The company hired one of the Big Six consulting firms to create a single sales data mart
on a Windows platform After spending nearly 1million on the effort and not achieving
their goal the account decided to stop the project The problem was not data volume or
technology it was data quality As it turned out a few of the sales applications restated
history anytime a change was made Consequently you would never be able to accurately
report all reversing entries and changes to every sales order line simply because the
application did not maintain that information But they did not need to spend a million
dollars to find out Consider the following options
Option One Spend 1million to bring in a high priced BI team conduct planning
sessions to create and agree to an elaborate project plan conduct business requirements
gathering sessions document and formalize all requirements in professional binders
build a fantastic entity relationship model gather and map source data to that model
purchase and install your platform start writing transformation scripts only then to find
out that the source data cannot be transformed into the required target table
Option Two Take a laptop with sample source data apply your business rules and see if
you can create the target table needed for less than 50k Do this before you commit to
the full scale project
Risk mitigation is all about saving money time and grief You be the judge Spend
1million to find out you have problems or 50k
Risk Mitigation
The peculiarities of analytical applications plus the lack of understanding of the source
data create significant amount of project risk And it is the project risk that must be
addressed as opposed to naively attempting to build detailed formal documentation based
on a best guess To address BI project risk this author recommends using as a guide the
seminal work of Barry Boehm and his process model the Spiral Approach This
Fosters the development of specifications that are not necessarily uniform
exhaustive or formal
Incorporates prototyping as a natural part of risk reduction
Encourages re work in the early development stages as better alternatives are
discovered
This section will briefly describe the Spiral Approach and more specifically how it can
be an integral part of your BI project planning
The Spiral Approach
The Spiral Approach is a unique risk mitigation process that can be used to drive the
entire BI project iteration or to supplement your in house life cycle development steps
and warehouse planning procedures The Spiral Approach contains four sections as
outlined below and illustrated in Figure 2
Quadrant 1 Determine Objectives Constraints This quadrant is designed for
project planners to examine the objectives and any constraints that might be associated
with this particular round
Quadrant 2 Risk Analysis Alternatives Prototypes Here is where risk is
explicitly addressed The quadrant is formally defined in order to ensure project planners
identify means by which risk can be mitigated specific to the round being implemented
For example if we have a known data quality issue in our data warehouse iteration then
we can define and initiate a Spiral round to address just that risk up front Perhaps we
know that integrated sales data from 20 disparate locations is going to be difficult To
address the integration risks we would conduct a Rule Based Audit or Proof of Concept
to see if we can achieve the level of integration necessary or come up with alternative
Quadrant 3 Development This is where we would blend the results of our risk
analysis with development requirements in order to create the needed solution In the
example above we may have identified an alternative approach to integrating sales data
which at this stage is being developed into a formal ETL process
Quadrant 4 Plan the Next Phase As you conduct the risk analysis and adapt those
results into your formal development you potentially change how the next tasks of your
overall project may be conducted Let s say our initial plans called for nightly updates of
integrated sales And after we conducted a test of the integration we determined that
updates could only be done weekly This new information changes how subsequent tasks
of the overall project iteration may be approached This is the quadrant where you adjust
your next step
The Spiral Approach is the only process model that is risk driven All other process
models and software development methods are document driven What s the difference
Document driven processes assume that complete formal documentation can be
obtained Unfortunately to obtain clear concise documentation the solution must be
clearly understood and defined Therein lies the problem Anyone with experience in BI
and warehouse centric iterations knows that solutions are seldom clearly and accurately
defined prior to development
Figure 2 Risk mitigation round
Q2 0 2 Risk Analysis Alternatives And Prototypes
Readiness Assessment
Data Quality Audit
Rules based Audit
Proof of Concept
GAP Analysis
Spiral START
Q1 0 1 Objectives And Constraints Q3 0 3 Development
Create Preliminary Scope
Risk Mitigation DIF Assessment
Examples Commit to Feasibility Study DW Strategy
Establish Objectives High level WBS
Outline Issues and Constraints Training Plan
Q4 0 4 Plan For Next Phase
ETL POC Effort
Examples Data Quality Strategies
Establish Requirement Priorities
In our 1million example the project was based on a document driven approach
Consequently they had very detailed professional documents and only encountered the
data quality problem in development If they had taken a risk driven approach to the
project the risky aspects of the effort would have been identified in advance and
addressed first In this manner they would have established whether the level of
integration could be achieved and if not what alternative solutions might be available
The power of the Spiral Approach is not just the fact that it is risk driven it is also
adaptable The Spiral Approach can be adapted for various BI iterations Outlined in
Figure 2 is a risk mitigation round with example tasks for each phase The round is
designed as a discovery effort to clarify understand and otherwise address the risky
points of any particular BI iteration The round can be used to examine issues ranging
from iteration priority data quality access alternatives cultural aspects of the
organization as well as the skill of IT and user communities This gives the organization
a chance to better understand its BI and warehouse iteration requirements strategies and
their relevant impact on costs and other resources
The most important thing to remember when implementing the Spiral Approach is to be
flexible creative and to adapt the approach to the circumstance The Spiral Approach is
a process model Its purpose is to help you define the entry and exit criteria between tasks
and or phases Most importantly it consciously blends risk assessment into the process
before you develop or create anything
Rule Based Audits and Proof of Concepts
Proof of Concepts and Rule Based Audits are examples of risk analysis techniques used
in the second phase of a risk mitigation round These techniques might be used to
evaluate source data quality identify alternative data sources or formalize new cleansing
strategies They can also help you address scale risks For instance you may have
decided that loading large data sets within a limited window might be of greater concern
and therefore a POC is used to test ETL processes for an optimum load time Regardless
of the type of risk mitigating technique used its scope must be kept within the range of
the Spiral round itself and agreed to by planners in the first phase of the round
In this section we examine specific steps necessary to conduct a successful RBA or POC
for your risk mitigation rounds
The Purpose of an RBA and POC
There are any number of questions issues doubts and unknowns about BI projects that
need answers Conducting an RBA or POC is done in order to obtain those answers to
add clarity as well as to understand the scale and scope of the project at hand
Specifically the RBA is designed to answer a single fundamental question can we take
known sources add explicit business rules and create the target data necessary for
subsequent analysis If you cannot answer this question with confidence then you have
no business risking company resources for a project
A POC on the other hand answers more questions regarding a BI project A POC takes
the results of your business rule audit and scales the testing to prove the feasibility of
production issues such as actual data volumes processing time constraints and platform
stress testing to name just a few
Refer to Table 1 0 for a comparison between Rule Based Audits and Proof of Concepts
Table 1 0 RBA and POC comparison
Risk Mitigating Techniques
Rule Based Audit Proof of Concept
Source Data Sample data only Sample or complete data set
Platform Completely conducted on an Either similar to RBA or implemented on the
independent isolated platform such platform of choice to test things such as batch
as a laptop cycle time network connections CPU
performance elapsed time performance etc
Testing Goal Applying explicit business rules to A POC takes the results of a business rules
sample source data in order to build a audit and scales the testing to address specific
target table s production level issues
You certainly can implement either technique to mitigate your risk But complete risk
mitigation is achieved by conducting both
Rule Based Audit and Proof of Concept Steps
The steps to conduct an RBA POC are straight forward and similar to the steps this
author outlined in the article Conducting A Data Quality Audit4
The five steps for a Rule Based Audit and Proof of Concept are as follows
1 Always perform the RBA POC before committing to a project Doing so uncovers
potential problem areas in regards to data quality and target data requirements as
well as quantifies performance and scale issues in order to accurately size the
scope of your project effort
2 Select your RBA POC tool s The tool s must be capable of applying a wide
variety of explicit business rules and yet be simple to install modify and execute
There are three core criteria to consider during your selection process
The tools must be robust enough to apply a wide variety of explicit business
rules to source data in order to create a target table s This includes complex
joins sorting and filtering
The tools must not interfere with the objective of the RBA POC Use
something that doesn t require specialized skills or training
The tools must install completely on a single laptop and scale to the most
powerful platforms The RBA is always conducted on a laptop using sample
data but a POC must be able to test full data volumes and target production
3 Gather source data definitions This step provides the initial scope of the tables
and their attributes relevant to the effort
4 Run initial audit The RBA is based on sample data in an attempt to determine if
we can build target data given known business rules and existing source data
There are three sub steps to running an initial audit
Apply known business rules These are explicit business rules that must be
applied to transform the source data into whatever target is necessary for
subsequent analysis
Create sample target Once all rules are defined you now can attempt to build
the target structure s
Test results Assuming the target table s can be built the final step is to
actually test the results That doesn t mean building pie charts but to test the
Gonzales Michael L Conducting A Data Quality Audit Intelligent Enterprise July 10 2004
data itself In other words can I aggregate sales grouped by sales orders
products and reversing entries made over the last quarter to get an accurate
5 Run Full Scale POC Once you have proven that you can create target tables
based on your RBA you must scale up to address production risks There are four
sub steps to this effort as follows
Choose a production data set s Testing current and future data volumes
requires sufficient data
Establish a testing environment If you can t use the real production
environment for your POC you must emulate it as much as possible This
means that if production uses X amount of disk space and is assigned Y
processors and Z memory then run your POC under similar conditions
Create metrics that can be verified and repeatable You must be able to
measure elapsed time performance and platform resources consumed
including CPU memory disk space etc This is critical as justification proof
for management
Synchronize POC target results with RBA results Even though you are
running a scaled up test the target data results must represent the results of
your audit
Once you know that target table s can be created with an RBA then you will want to
prove what can be achieved at production scale using a Proof of Concept Refer to the
notes section of this article for more information regarding these risk mitigating
techniques
Technology That Supports Your Risk Mitigation
Rule Based Audits Using BRE
A Rule Based Audit is specifically designed for audits based on business rules Its sole
purpose is to allow subject matter experts to import source data apply explicit business
rules and attempt to create a target table to meet the user requirements For the purpose
of illustration Figure 3 shows a data flow for a sample audit
Figure 3 Sample data flow
The goal of the process shown in Figure 3 is to use business rules to determine the total
deposits by account for the current month from the current month s transactions To do
so we first define three new tables by filtering data from the main Current Transactions
table as follows
Debit corrections are transactions with an OPR TYP operation type of 305
Credit corrections are transactions with an OPR TYP operation type of 003
Deposit corrections are transactions with an OPR TYP operation type which is
listed in the Deposits list
The next step is to create a list of credit corrections that have not been zeroed out by a
corresponding debit correction Credit corrections and debit corrections are compared by
TRANSIT ACCOUNT and TRX AMT with any matches being filtered out of credit
corrections A similar process is then used to filter the deposit transactions table with
deposit transactions by comparing them to the filtered credit corrections Once the
deposit transactions have been filtered for off setting corrections they are filtered to
make sure only debit transactions are used in the calculations The month s deposits per
account are then calculated by grouping transactions by TRANSIT and ACCOUNT and
then a sum of TRX AMT
It sounds confusing I know But that is exactly why we want to conduct this RBA in the
first place Complex filtering and integration increases risks Instead of assuming we can
do the work the audit allows us to confidently determine if we can actually build the type
of target necessary
My firm HandsOn BI LLC has designed developed and maintains a product called
Business Rule Engine BRE BRE is an effective tool used for our audits based on
sample data sets and completely isolated from a technical architecture It is made
available to our members only and is unsupported Most advanced integration rules or
complex calculations can be applied in the BRE as shown in Figure 4
Figure 4 BRE rules definition
Although BRE is an effective tool for RBAs it is homegrown There are better tools
available to perform an audit Tools that are more robust stable and actually come with
technical support DMExpress is one
Proof of Concepts Using DMExpress
Once you have proven that you can create target tables using explicit business rules
against sample source data you must scale up to address production risks For example
can you transform the data volume necessary on a nightly basis given target production

---