Routine Data Quality Assessment

Estimated time: 35 minutes

This lesson connects Excel, digital collection and multi-source analysis with the question of whether programme figures are fit to use.

6.5.1 What is data quality?

Data quality is how well data meets the requirements of its intended use. A district manager may need timely and sufficiently complete information to plan services. A researcher may need accurate and consistently defined data. Different uses need different checks.

6.5.2 Important dimensions of data quality

Common dimensions include accuracy, completeness, timeliness, consistency and validity. Some frameworks also use uniqueness, reliability or integrity. There is no single list that applies identically to every dataset.

Accuracy asks how closely the report reflects the event. Completeness asks whether required records, fields or reports are present. Timeliness asks whether the data arrived by the deadline. Consistency asks whether definitions and rules match across records, places and periods. Validity asks whether values follow the required formats, ranges and rules.

Five checks used in a routine data quality assessment

A standard routine data quality assessment looks at these five dimensions together. A pass on one does not excuse a failure on another.

Dimension	Question to ask	Example
Availability	Can an authorised reviewer find the source?	The monthly summary exists, but the immunisation register for the last Friday is missing.
Completeness	Are the expected reports and required fields present?	Eight of ten facility reports arrived. Two facilities sent nothing.
Timeliness	Did the report arrive by the deadline?	The summary is correct and arrives after the planning meeting.
Integrity	Can the figure be traced, without silent changes, from the source to the report?	The register says 28 and the submitted total says 82, with no correction note.
Confidentiality	Is access limited to people who are allowed to see it?	A line list of names is pasted into a general dashboard.

6.5.3 A fictional data quality problem

An NGO expects ten facility reports and receives eight. Completeness is 8 ÷ 10 × 100 = 80%. Completeness does not establish accuracy. The eight reports can still have missing indicator values, duplicates, incorrect totals, inconsistent periods and data-entry errors.

6.5.4 Common data quality checks

Completeness: expected reports and required fields are present.
Timeliness: submission dates against the deadline.
Consistency: periods, definitions and units match.
Validity: values follow documented rules. A percentage outside 0 to 100, where that is the rule, should be investigated.
Accuracy: where authorised, compare the reported total with the register or approved summary.

6.5.5 Verification versus correction

A facility that usually reports about 100 antenatal visits now reports 950. Possible explanations include a typing error, a different period, a changed definition, a genuine increase, or a duplicate. Investigate and document. Do not replace 950 with 100 because 100 looks familiar.

6.5.6 What should a data quality assessment document?

Field	Purpose
Indicator or dataset	What is being reviewed
Reporting period	The period assessed
Expected reports or records	The reference requirement
Reports or records available	What was received
Checks performed	The methods used
Issues identified	Discrepancies or gaps
Corrective action	What needs to be addressed
Responsible person	Who will follow up
Follow-up status	Whether the issue was resolved

6.5.7 Why data quality matters for M&E

Poor quality can produce wrong performance estimates, misleading dashboards, poor allocation, inappropriate conclusions, less confidence in reports, and late recognition of implementation problems. Build the checks into routine work. Do not leave them for a rare exercise.