Databases Reference
In-Depth Information
the old NXX to frame relationship. As a result, the planner is presented with
a distorted projection that has the appearance of correct data, but the re-
sulting decision could be tragic!
Other data “land mines” arise as a result of changing data relationships
that have not been anticipated in the data manipulation/transformation
code. Even the simple addition of a new product can produce unexpected re-
sults if the existing transformation code was not designed to dynamically ad-
just to new codes (or someone forgets to update a product-matching list).
In addition to the problems that stem from hidden or invisible business
rules, there are subtle (and not-so-subtle) problems that crop up all the time
when users or programmers violate an existing, but hidden rule. Senior
managers are particularly adept at this as they will tend to ignore all of the
subtle exceptions to normal business flow and relationships as they search
for the big picture. This results in views which portray conflicting informa-
tion. This leads senior management to no end of headaches, and leaves a
bad taste about the “reliability” of systems. IT usually gets the blame.
Query “land mines” are not limited to incorrect data relationships. Prob-
lems frequently arise from assumptions made in the query code. Clients of-
ten wrongly assume that all attributes are data-filled, and then make wrong
decisions on what they believe to be a complete answer. For example, when
the planner needs to know how many customers are eligible for a new lo-
cation-sensitive service, they must not rely entirely on postal codes that
may or may not be data-filled. Often, this problem can be mitigated by
weighing factors computed from the percent of populated data.
In another example, a client assumed that a join between the Circuit ta-
ble and the Customer table would give them a list of all customers affected
by the rearrangement of a high-capacity fiber optics cable. Unfortunately,
the query was coded as a conventional (inside) join, resulting in the
omission of switched circuits from the planning (i.e., circuits without a ded-
icated customer).
CONCLUSION
Data may just be strings of 1s and 0s, but the value of that data is the
combination of the data and the meaning hidden in the data, data relation-
ships, queries, interpretive programs, and the hidden assumptions of the
people who use the data. Changes and mismatches occur over time as
rules change, as the business rules change, as entities merge and separate
and as regulations and generally accepted principles change. This article
has identified the places where the
of the data is hidden, and the
structure of how changes and mismatches come into being.
meaning
There are three types of mismatch: missing attributes, non-atomic data,
and business rule changes. Data practitioners need to work aggressively
Search WWH ::




Custom Search