Quality assurance processes
Quality assurance is applied throughout the data-matching cycle.
These assurance processes include:
- registering the intention to undertake a data-matching program on an internal register
- risk assessment and approval from the data steward and relevant senior executive service (SES) officers before any data-matching program starts
- conducting program pilots or getting sample data to ensure the data-matching program will achieve its objectives before full datasets are collected
- notifying the OAIC of our intention to undertake the data-matching program and to seek permission to vary from the data-matching guidelines (where applicable)
- restricting access to the data to approved users and using access management logs to record the details of those who have accessed the data
- embedding quality assurance processes into compliance activities, including the
- review of risk assessments, taxpayer profiles and case plans by senior officers before client contact
- review of cases by subject matter experts
- regular review of samples of case work by independent panels to ensure our case work is accurate and consistent.
These processes ensure data is collected and used in accordance with our data-management policies and principles and complies with the OAIC's Guidelines on data matching in Australian Government administrationExternal Link.
How we ensure data quality
Data quality is a measure to determine how suitable data is for its intended use. It helps us to understand the data asset and what it can be used for.
Data quality management:
- allows us to use data with greater confidence
- assists us meet data governance requirements
- ensures we have a greater understanding of the data we hold.
The ATO Enterprise Data Quality (DQ) framework guides the effective and sound use of data.
This framework outlines 6 core DQ dimensions:
- Accuracy – the degree to which the data correctly represents the actual value.
- Completeness – if all expected data in a data set is present.
- Consistency – whether data values in a data set are consistent with values elsewhere within the data set or in another data set.
- Validity – data values are presented in the correct format and fall within a predefined set of values.
- Uniqueness – if duplicated files or records are in the data set.
- Timeliness – how quickly the data is available for use from the time of collection.
To assure data is fit for consumption and the intended use throughout our data-matching programs, the following data quality elements may also be applied:
- Currency – how recent the time period is that the data set covers.
- Precision – the level of detail of a data element.
- Privacy – access control and usage monitoring.
- Reasonableness – reasonable data is within the bounds of common sense or specific operational context.
- Referential integrity – when all intended references within a data set or with other data sets are valid.
Data is sourced from providers' systems and may not be available in a format that can be readily processed by our own systems. We apply extra levels of scrutiny and analytics to verify the quality of these datasets.
This includes but is not limited to:
- meeting with data providers to understand their data holdings, including their data use, data currency, formats, compatibility and natural systems
- sampling data to ensure it is fit for purpose before fully engaging providers on task
- verification practices at receipt of data to check against confirming documentation; we then use algorithms and other analytical methods to refine the data
- transforming data into a standardised format and validating to ensure that it contains the required data elements before loading to our computer systems
- evaluating program effectiveness before determining whether to continue to collect future years of the data or to discontinue the program.