How Do I Implement Business Intelligence? (Part Two - Get to Know the Kitchen)

Characterize the Environmentplace governing naming conventions, data type usage,
The following list outlines the items and issues to coverdata structure designs, etc?
in attempting to characterize the environment withinData is not always sitting in a database, nor is it
which your challenge is encapsulated.always easily accessible. Sometimes it is necessary to
- Characterize the Environmentaccess data that has already been extracted from its
- Organizational Scopeoriginal source and stored somewhere else. It may be
- Project Sponsorshipnecessary to reuse the data feed of such a transfer
- Business Contextof data from the destination. This can introduce serious
- Business Objectivesissues of data reliability and stability. if the second
- Current Data Landscapedestination has itself transformed the data or if the
- Solution Objectivesdata has been summarized to a higher level of
- Technical Environmentgranularity or even stored as a periodic snapshot
- Resourcesinstead of at an atomic level, there will be issues.
- FundingResources
The topics outlined above may not all be necessaryBesides the need for resources on the BI project,
for some smaller organizations and in any case, it isthere is often the need to call on the help of others
not recommended to take a heavy approach to mostoutside the project. The lack of such resources, or
of these items. All that is required is that they be givenlimitations on their availability can cause problems for
some thought and be documented if there is anythe BI Initiative.
useful information to be captured.Funding
Organizational ScopeLarge companies tend to be able to fund large
It may be that you are launching a BI initiative that isprojects. Conversely, small companies cannot.
intended to eventually serve the entire organization.However, outside of this fact, there is a need to
Titles like 'Enterprise Data Warehouse' (EDW) andconsider whether a large project is actually the most
'Corporate Information Factory' (CIF) have frequentlyeffective approach to BI. Many initiatives have come
been given to such initiatives.to grief because the challenges involved, compounded
On the other hand, it may be that only a subset ofby the inherent difficulty of coordinating large scale
your organization is the intended user and customer ofprojects, leading to massive over-runs and wasted
the product. A division of your company, or possiblytime.
even a single department.One of the key areas for devastating misjudgments is
An all-too-common practice has been the creation ofthat in which organizations inexperienced in DW and BI
Data Marts an a virtually ad-hoc basis, resulting in antake on projects that are too large and too complex
unknown and uncontrolled proliferation of such entities.for their relatively inexperienced team.
It was recently published that Hewlett PackardRegardless of whether funding is easily obtained or
launched an offensive against such a proliferationnot, it is strongly recommended that some kind of 'pilot
when it was discovered that there were more thanproject' be initiated first, to gain experience and
750 Data Marts in existence across the organization.confidence in dealing with the new technologies
This is when Business Intelligence could start to lose itsinvolved. Then after an initial smaller success, the team
true intended purpose.(or individual, as is often the case) can move on to
Capturing the details of the organizational scope oftake more territory.
your BI Initiative will serve to define the limits of theKey to success is knowing how much to attempt
project.rather than how much to fund.
Project SponsorshipEstablish a Roadmap
Any project needs to have its sponsors well defined. ItThe mode under which the BI Initiative will be
is a good strategy to encourage a vibrant relationshipconducted is the most important consideration at the
between the sponsors and the delivery team. Recordstart of the project.
who they are and work on that relationship. If you areHistorically, most software application development
the sponsor then be prepared to maintain closeprojects followed a phased plan that was essentially
contact and involvement with the person or teamserial in nature. Often referred to as 'waterfall model'
implementing your BI solution.processes, these were so named because the
Business Contextproduct of the first phase 'poured' into the second
Within the organization that is sponsoring or requestingphase, the second into the third and so on, regardless
a BI facility, there may any number of businessof the name or purpose of each phase.
processes, databases, application systems andEventually, it became obvious that this approach could
sub-organizations. Take time to make a record ofbe responsible for many of the ills that plagued such
those for which the solution is to relevant, either asprojects. The fact that any given phase is dependent
data sources or expected areas of enhancement.entirely on the quality, appropriateness and timeliness
Business Objectivesof the products of its predecessor meant that any
Why do we need a BI solution? Someone in thesingle phase had the potential of setting the limit on the
organization came to the conclusion that this would beoverall project success, of being the weakest link.
beneficial. Someone is aware that the solution will meetFurthermore, the very act of feeding one phase with
one or more needs. What are those needs? Whatthe documented results of another could introduce
does that person or group expect to achieve? In whatdefects due to the misunderstandings, ambiguities, lack
way will their aspirations be met? They must have aof clarity, etc., resulting from this form of
reason, or many reasons, perhaps for going ahead.communication. Surprisingly, the creators of one
Those reasons will indicate and help crystallize thephase's deliverable were often not available for
objectives of the BI Initiative.consultation with the recipients of their efforts in the
Current Data Landscapenext phase. Many weaker processes did not mandate
Now that we have organizational scope, projecta review (or quality gate) at the end of each phase.
sponsors, a business context and objectives, weYet another issue with Waterfall models is that the
should try to define the data sources that need to beproject is not finished until all phases are complete. As
leveraged in order to provide the candidate informationthis could extend into years, it would be necessary to
from which some intelligent business decisions can be'freeze' requirements until the project was complete.
derived.Business changes would have to go unattended to
At this level it is sufficient to speak of this in terms ofbecause interrupting this process was too disruptive to
databases or applications or functional areas coveredbe tolerated. Also, any error in delivering against the
by an application.original requirements may not come to light until the
Additional information about the data sources shouldvery end of the string of phases, when it is then the
be collected. How easy will it be to learn about themost expensive to repair.
data source? Is it well documented? Are there stillEvolving the Solution
people available who developed, maintained orAvoiding all the above issues is not a trivial task.
supported the database that can contribute insideSplitting the project into 'waves' is sometimes
knowledge? Is the source 'out-of-bounds'? Can weattempted to reduce the work going through the
only obtain data from it under some constraint, such assequence of phases and therefore reducing the time
time, frequency, security policy, etc.?frame of each wave is one, often used approach for
Solution Objectivesvery large projects. However, this does not remove
The information gathered so far will help direct yourthe problem-causing imperfections of waterfall models,
efforts and maintain control over scope, cost andonly 'divides' in order to 'conquer' them.
delivery aspects of the project. They may not,The approach that often produces the most
however, define everything we might be able to learnsuccessful result is that of 'Iterative' development, or
about the nature of the target solution. It is possible thatprototyping. The main aim in this method is to work as
corporate or IT policy may also impact the size andquickly as possible toward a reduced end-product but
scope of the project.one which can still be used, or at least, demonstrated.
It is possible that technical limitations exist or resourceIts advantages are: a quicker end point at which the
availability limitations exist. There may even be limits onproduct can be evaluated; an opportunity to learn from
funding that prevent all the business objectives frommistakes, misunderstandings or unseen challenges and
being pursued all at once.the availability of a version of the solution that the
Sometimes, it makes sense to extend the scope a littlesponsors can try out to see if their ideas were sound
because the effort of visiting one or more dataor in need of refinement. Indeed, the all-round learning
sources just to obtain the requested data me beprocess afforded by the first iteration is one of the
wasting an opportunity to pull more useful (but not yetstrongest arguments for the approach. This takes us
requested) data, This is a very important considerationaway from the need to 'get everything absolutely
because returning to the same source later is farcorrect, at the first and only attempt.
more expensive than picking up data now that mayThe virtual impossibility of such an outcome with the
even be in adjacent columns to that which we havewaterfall model is certainly what leads to the creation
come looking for. This is one area where anof a 'Phase II' on most projects. Phase II may include
experienced Dimensional Modeler is able to point tonew features but it usually includes a lot more in the
'easy gains' that could soon satisfy a growing appetiteway of 'rework' of the initial release.
for analytic data.Establishing a road-map for the project involves the
Technical Environmentdecision to use the appropriate style of process and
It will be much easier to estimate the size of thedetermining what to deliver by way of a prototype.
project if we know something about the challenges itDividing the project into two or more iterations offers
may present. One such challenge is dealing with themany advantages if that option is available.
issues surrounding the way the data is currently..to be continued in Part Three of 'How do I Implement
hosted. Is it all in the same kind of database? Is it all inBusiness Intelligence?'.
the same location? Are there common standards in