-
Development
time for reports in greatly increased (usually disproportional to mart
development effort)
-
Business
Rules can be implemented differently. Even
if great effort is taken to avoid this, it still happens.
-
Changes
to a single business rule need to be made in many different places
(reports).
-
It
becomes difficult to integrate multiple systems in reports.
-
It
becomes cumbersome on the live OLTP system resources during business hours.
-
Report
performance is drastically decreased.
-
Slicing,
and especially drilling, data in reports becomes so difficult that it can
easily become prohibitive.
-
High-level
analysis, especially "Ad-Hoc", is not usually possible without
taking extreme measures. Basically,
this means that "data mining" will not be possible, or at lease
not practical. Data mining is
very powerful for finding business value in the data that would otherwise
not be apparent.
Example:
One
famous example is when Wal-Mart analysis used a data warehouse to perform a
"market basket analysis" and found that beer and diapers had a high
correlation of sales in the same orders. By applying this knowledge to increase
the sales floor proximity of these items, beer sales doubled. No one would have
thought to even look into this, but by running some general data mining queries
against the data, this correlation was identified.
The
point:
The
most important question is often the one left unasked.
Data mining is basically a way to get answers to questions that no one
thinks to ask. This is something
that OLTP system reporting cannot provide.
-
If
a business user needs s simple answer to a question, either he/she must be
very knowledgeable of the data, or a report must be developed by someone who
is very knowledgeable of the data (i.e.: a developer).
In other words, true "Ad Hoc" querying is not possible, or
at lease not practical.
-
When
reports are developed in a certain technology, like COGNOS, one is
essentially tied to that technology forever.
Changing reporting tools becomes very expensive and difficult.
In the case of a mart, the front-end reporting side is not doing as
much sophisticated activity because the complex business rules and data
relationships are all implemented in the periodic ETL process.
-
In
the future, if data becomes sourced by a different system, the all reports
using this data will need to be rewritten.
In a mart / warehouse, only the ETL process would need to change and
the reports would remain untouched.
-
Perhaps
the most important requirement is the need to report historical and "as
of" data. This requirement can only be met by storing this data
offline. In many other cases, special files loaded by special programs have
been created in order to meet this "as of" and historical
reporting requirement. This causes a swarm of hot files, custom tables, and
one-off programs. The bottom line is that there will certainly be off line
data storage in order to meet the businesses requirements.
-
Users
need to be able to run ad-hoc analaysis that can not be done with canned
reports.
![[Company Logo Image]](images/logo.gif){kind=logo}
Jump to: DW University,
Employment
![[Under Construction]](images/undercon.gif){kind=small}
