Databases Reference
In-Depth Information
Do internal information collection, dissemination, and verification
procedures measure up to quality requirements?
ยท
Data quality training and awareness programs must be carried out jointly
with the data quality policy. Personnel involvement is a prerequisite to qual-
ity program success.
In addition, an information quality assessment process must be imple-
mented. English [49] puts forward a methodology called TQdM (Total
Quality data Management), which allows the assessment of an organizations
information quality. The methodology consists of the following steps:
1.
Identify an information group that has a significant impact in order
to give more added value.
2.
Establish objectives and measures for information quality, for
example, assess the information timeliness and measure the span
that passes from when a datum is known until it is available for a
specific process.
3.
Identify the information value and cost chain, which is an
extended business value chain focused on a data group. This chain
covers all the files, documents, DBs, business processes, programs,
and roles related to the data group.
4.
Determine the files or processes to assess.
5.
Identify the data validation sources to assess data accuracy.
6.
Extract random samples of data, applying appropriate statistical
techniques.
7.
Measure information quality to determine its reliability level and
discover its defaults.
8.
Interpret and inform others about information quality.
A crucial aspect for carrying out this process is the definition of significant
metrics that allow for the analysis and improvement of quality. In [45], three
kinds of metrics are given: subjective (based on user opinion about
data); objective, application-independent (e.g., accuracy); and objective,
application-dependent (specific to a particular domain).
Companies must also measure the value of the information, both infor-
mation produced by operational systems and information produced by
decision-support systems. The way of measuring both kinds of information
varies considerably. In Due [50], three different approaches (normative,
