Data Quality for Institutional Decision-Making

Status, question and scope

Research question

What must decision-makers know before relying on institutional data?

Scope

Data are useful when their origin, definition, completeness, limitations and relationship to a decision are understood.

ASPC application and limitations

Proposed application

• Maintain a data dictionary, source owner, collection period and transformation record.
• Assess completeness, timeliness, consistency and plausible bias before use.
• State what a dataset can and cannot support and record how it informed the decision.

Limitations

• A quality process reduces avoidable error but does not eliminate uncertainty or establish causality.

Documented findings

Governance connects data to mission decisions

GAO-21-152 describes data governance as organizational structures and supports intended to improve data quality and availability. In its review of four federal agencies, GAO found progress in creating governance frameworks but gaps in maturity assessments, staff data-literacy assessments and mechanisms for tracking progress.

GAO identified three questions for data leaders: whether governance strategies answer priority mission questions, what objectives governance serves, and how its value is communicated. The report also found that plain-English descriptions matter because users may otherwise misunderstand what reported activity or spending is intended to accomplish.

Implication for evidence use

A technically complete dataset can still be unfit for a decision if definitions, provenance, timeliness or decision relevance are unclear. Governance focused only on compliance may likewise fail to improve operational choices.

Data-quality controls

Decision-use record

• State the decision the data will inform and the consequence of error.
• Record the owner, originating system, collection method and transformation history.
• Define completeness, validity, accuracy, consistency, timeliness and uniqueness thresholds.
• Document missing values, exclusions, revisions and known sources of bias.
• Assign an approval date and a date after which the dataset must be reassessed.

Proportionate assurance

Higher-consequence decisions require stronger verification, external challenge and reproducibility. Exploratory analysis may use provisional data if uncertainty is visible and no irreversible conclusion is presented.

ASPC publications should link claims to stable source records and separate reported facts, analytical transformations and judgments. Corrections should preserve the original publication date, identify the changed statement and explain whether conclusions were affected.