Before starting any data quality or data governance initiative, an important first step is to establish the expectations for this program. Specifying the expectations of the data’s consumers provides a means for measuring and monitoring the conformance of data (and associated processes) within an operational data governance framework. These agreements can be formalized under a data quality service level agreement (DQ SLA), which specifies the roles and responsibilities associated with the management and assurance of data quality expectations.

SLAs are familiar to anyone with an IT background, but they are typically focused on issues of system availability, service turnaround and other issues. This paper discusses how implementing a DQ SLA via formalized processes can transform data quality management from a constant “fire-fighting” mode to a more consistent, proactive approach.

The objective of the data quality service level agreement is establishing data quality control. This relies on monitoring conformance to data quality rules define using agreed to dimensions of data quality, such as accuracy, completeness, consistency, reasonableness, and identifiability, among others. We will consider these dimensions of data quality, and the ways that data quality rules are defined. Despite the best efforts to ensure high data quality, there are always going to be issues requiring attention and remediation. As a result, identifying data errors as early as possible in the processing stream(s) supports the objective of the DQ SLA: notifying the right individuals to address emergent issues and resolving their root causes in a reasonable amount of time. We will look at the process of defining data quality rules, their different levels of granularity, approaches for introducing measurement processes, and choosing appropriate acceptability thresholds.

This paper will then consider the relevance of measurement and monitoring: defining inspection routines, inserting them into the end-to-end application processing, and reporting measurements. When the quality of data does not meet the level of acceptability, data quality issue events are generated, the issues are logged in a data quality incident tracking system, and the individuals specified in the data quality service level agreement are charged with diagnosis and remediation. Through this operational data governance, an organization can internalize the observance of the DQ SLAs, and consequently continuously monitor and control the quality of organizational data.