1) Identify key data

2) Identify existing rules and patterns

1) Identify and prioritize problems

2) Execution problem root cause analysis

1) Prioritize actions based on business impact

2) Develop preventive and corrective measures

3) Confirm planned actions

1) Develop data quality operating rules

2) Correct data quality defects

3) Measure and monitor data quality

4) Report data quality levels and findings

Data quality management should focus on the data that is most important to the business and its customers, and improvements should be prioritized based on the importance of the data and the level of risk if the data is incorrect.

Data quality management should cover the entire life cycle of data from creation or creation to disposal, including data management when it flows within and between systems (each link in the data chain should ensure that data has high-quality output) .

The focus of a data quality program should be on preventing data errors and situations that reduce data availability, not on simply correcting records.

Improving data quality is more than just correcting errors. Because data quality problems are often related to process or system design, improving data quality often requires changes to processes and the systems that support them, not just understanding and solving them.

Data governance activities must support the development of high-quality data, and data quality planning activities must support and sustain a governed data environment.

All stakeholders in the data lifecycle will have data quality requirements. Where possible, quantifiable data quality requirements should be defined in the form of measurable standards and expectations.

Data quality levels need to be measured objectively and consistently. Measurement processes and measurement methods should be discussed and shared with stakeholders as they are the arbiters of quality.

Business process owners are responsible for the quality of the data generated through their processes, and they must implement data quality standards into their processes.

System owners must enforce data quality requirements on the system.

Data quality reporting and issue management should be integrated into service level agreements (SLAs)

1. Regulatory reporting

2. Financial reporting

3. Business policy

4. Going concern

5. Business strategy, especially differentiated competitive strategy

The Strong-Wang framework focuses on data consumers’ perspectives on data

In his book "Data Quality in the Information Age", Thomas Redman developed a set of data quality dimensions based on data structure. Redman defines a data item as a "representable triple": a collection of entity attributes and values.

In his book "Improving Data Warehousing and Business Information Quality", Larry English proposed a set of comprehensive indicators, divided into two categories: inherent characteristics and practical characteristics. Intrinsic features have nothing to do with data usage. Practical features are dynamic and related to data expression. Their quality value depends on the use of data.

1) Completeness. Stored data volume as a percentage of potential data volume

2) Uniqueness. Entity instances (things) should not be recorded multiple times on the basis of satisfying object recognition.

3) Timeliness. The extent to which the data represents reality from the requested point in time.

4) Effectiveness. Data is valid if it conforms to its defined syntax (format, type, range).

5) Accuracy. The extent to which the data accurately describes the "real world" object or event being described.

6) Consistency. Compare the differences between various expressions and definitions of things.

The Deming Circle is a problem-solving model known as "plan-do-check-act"

P: plan

D: Execute

C: Check

A: Action/process

Confirm that data definitions are understood identically and are implemented and used correctly throughout the organization;

Rules defining whether missing values ​​are acceptable

A value assigned to a data element according to a specified pattern, such as a standard for formatting phone numbers

The assignment of the specified data element must be included in the enumeration value of a certain data value field. For example, the reasonable value of the state field is the 2-character U.S. zip code.

The data element assignment must be within a defined number, dictionary or time range, such as a number range greater than 0 and less than 100.

Represents a value assigned to a data element that must correspond to a selected value that maps to another equivalent corresponding value range.

Refers to the conditional determination of the relationship between two (or more) attributes based on the actual values ​​of these attributes.

Compare the data value to the corresponding value in the system of record or other verification source (such as marketing data purchased from the vendor) to verify that the values ​​match.

Rules that specify which entities must have unique representations and have exactly one record for each representation of a real-world object.

Rules indicating characteristics expected to be associated with data accessibility and usability.