Among national statistical agencies it has long been acknowledged that "confidence in the quality of the information it produces is a survival issue for a statistical agency" (Brackstone 1999). However, given the amount of data that a statistical agency handles each and every day, the risk of releasing incorrect statistics to the public and the subsequent loss of confidence in the organisation by users for this type of error is very real.
Errors in statistical outputs can be minimised by committing to quality management strategies, such as risk management. Risk management is concerned with identifying potential risks, analysing their consequences, and devising and implementing responses, ensuring that corporate and business objectives are achieved while upholding quality.
The Australian Bureau of Statistics (ABS) leads Australia's national statistical service, running hundreds of surveys and publishing thousands of pages of output every year. As with any large and complex organisation, problems with processes do arise and the ABS has suffered errors in their data in the past with varying degrees of impact on the public domain. Most errors are detected in-house before publication, however this has at times resulted in intense last-minute work to correct the problems leading to delays in the release of data. Other errors have only been discovered after release, resulting in re-issue of statistical output. As a result of these errors the ABS has endeavoured to instigate better quality management practices through the development and use of the risk mitigation strategy known as quality gates.
Quality gates are designed to improve the early detection of errors or flaws in production processes. Specifically, the principles that underpin the quality gates framework are:
- Quality of statistical processes should be managed in a holistic manner i.e. Total Quality Management;
- Quality management and assessment of fitness-for-purpose of statistical processes should be evidence based;
- Any problems arising in statistical processes should be detected as early as possible;
- Roles and responsibilities in the management of process quality should be clear and explicit;
- Knowledge and information about specific stages of a statistical process should be documented and shared; and
- Regular evaluation should capture lessons learnt and lead to continuous improvement of quality management of statistical processes.
The concept of quality gates is not a new one. It has been used in other fields for many years such as the automotive and information technology industries. In the ABS, quality gates consist of six components which distinguishes them from more general every day risk management strategies. The six components of a quality gate are Placement, Quality Measures, Roles, Tolerance, Actions and Evaluation. This paper describes in more detail the quality gate framework to enable its use as a statistical risk mitigation strategy primarily for statistical processes. In particular, the paper provides an explanation of each of the six components of a quality gate, followed by examples and templates to assist agencies to apply the framework.
SIX COMPONENTS OF QUALITY GATES
Quality gates can be used to improve the visibility of quality in the production process as well as being used to measure and monitor quality in real time at strategic points. Quality gates consist of a set of acceptance criteria imposed at predetermined points in a production process whereby each of the components (Placement, Quality Measures, Roles, Tolerance, Actions and Evaluation) play an important part in determining the fitness for purpose of the process.
Quality gates are designed to facilitate the detection, discussion and resolution of issues and problems through a collaborative effort to improve the quality of products.