How We Ensure Data Quality
We apply multiple, concrete controls at each stage of data collection to prevent errors rather than correct them later.
01
Tool & questionnaire controls
-
Mandatory fields, skip logic, and range constraints built directly into KoboToolbox forms
-
Hard validations to prevent impossible values (e.g. age, dates, household size)
-
Soft warnings to flag unusual but possible responses for later review
-
Full piloting of tools before rollout, with revisions based on pilot results
02
Enumerator controls
-
Enumerator-specific IDs linked to every submission
-
Standardized training on questionnaires, probing techniques, and ethics
-
Clear daily targets and acceptable interview-duration ranges
-
Immediate suspension or retraining when abnormal patterns are detected
03
Real-time monitoring
-
GPS capture for each interview to verify location and coverage
-
Interview duration tracking to detect rushed or fabricated surveys
-
Daily monitoring of submissions by supervisors using Kobo dashboards
-
Coverage tracking to ensure correct distribution across districts and locality types
04
Supervision & verification
-
Daily supervisor review of completed interviews
-
When possible, random back-checks on a sample of interviews (phone or in-person where feasible)
-
Cross-checking responses for internal consistency (e.g. demographics vs. outcomes)
-
Immediate corrective action when flags recur at enumerator or location level
05
Post-collection validation
-
Duplicate detection and removal
-
Outlier analysis for extreme or patterned responses
-
Final dataset approval only after all flags are resolved
-
Full anonymization before analysis and reporting