Resource
Data Quality Assurance Protocol
A field-to-desk QA protocol covering enumerator briefing standards, daily supervision checks, back-check sampling rates, outlier review, and documentation requirements.
Donor context: M&E teams
Context tags indicate reporting alignment or funder familiarity; they do not imply donor endorsement.
Purpose
This protocol establishes minimum standards for data quality assurance across all phases of a CAPI-based field survey — from enumerator briefing through to final dataset handover. Adapt thresholds to your specific study design and field context.
Phase 1: Pre-Field Briefing Standards
1.1 Enumerator Briefing (minimum 30 minutes per module)
- Instrument walkthrough: each section read aloud by supervisor, enumerators follow on device
- Common errors review: share anonymised error log from a comparable previous study
- Skip logic verification: enumerators trace at least 3 complete paths through the instrument
- Consent script delivery: role-play with supervisor as respondent; minimum 2 rounds
- Data submission demo: enumerator submits a test form and confirms server receipt
1.2 Supervisor Briefing
- Daily quota and quality targets communicated in writing
- Escalation contacts confirmed (lead researcher, field coordinator)
- Back-check assignment methodology explained
- Daily monitoring dashboard access demonstrated
Phase 2: Daily Supervision Checks
2.1 Morning Checks (before field departure)
| Check | Standard |
|---|---|
| Device charge | ≥ 80% |
| App version | Must match current version number |
| Sync status | All previous day's submissions received |
| Assignment review | Cluster assignments for the day confirmed |
2.2 Real-Time Monitoring (during field hours)
- Supervisor reviews live submissions every 2 hours via monitoring dashboard
- Flag any interview duration < 60% of median pilot duration for immediate callback
- Flag any interview with > 5 constraint violations for supervisor review
- GPS coordinates plotted daily against expected cluster boundaries
2.3 Evening Debrief (mandatory)
- Supervisor reviews enumerator daily report (forms submitted, refusals, replacements)
- Any incomplete forms identified and reason documented
- Issues escalated to lead researcher if affecting > 5% of daily quota
Phase 3: Back-Check Sampling
3.1 Sampling Rate
| Enumerator status | Back-check rate |
|---|---|
| Standard | 10% of completed interviews |
| Flagged in Phase 2 | 25% minimum |
Back-checks must be conducted within 48 hours of the original interview.
3.2 Back-Check Procedure
- Supervisor contacts original respondent independently (phone or in-person)
- Asks 6–10 verification questions (demographic anchors + 2–3 substantive items)
- Records responses on back-check form (separate from CAPI)
- Compares responses; discrepancies > 2 items flagged as non-compliant
3.3 Non-Compliance Response
| Discrepancy count | Action |
|---|---|
| 0–1 items | No action required |
| 2–3 items | Enumerator counselled; supervisor increases monitoring frequency |
| 4+ items | Enumerator suspended pending investigation; affected forms quarantined |
Phase 4: Outlier Review
4.1 Daily Flagging Rules
- Interview duration < 40% of median → automatic flag
- Interview duration > 200% of median → automatic flag
- Income / expenditure values > 3 SD from cluster mean → flag for review
- GPS coordinates outside expected cluster boundary → flag for review
4.2 Outlier Resolution
Each flagged record receives one of the following dispositions:
| Disposition | Documentation required |
|---|---|
| Confirmed valid | Supervisor note explaining the outlier |
| Corrected | Correction documented with source |
| Excluded | Reason documented; record moved to quarantine dataset |
All dispositions recorded in the Outlier Decision Log (maintained by field supervisor).
Phase 5: Documentation Requirements
During field phase
- Daily field log (supervisor): forms submitted, refusals, replacements, issues
- Back-check register: all back-checks completed, results, dispositions
- Outlier decision log: all flagged records and resolutions
- Constraint violation log: enumerator, date, question, violation type
Post-field / cleaning phase
- Cleaning protocol document (all rules applied, in order)
- Variable construction notes (derived variables, calculation method)
- Exclusion log (records excluded, reasons)
- Final variable dictionary
- Dataset version register (filename, date, changes)
Handover package
- Cleaned dataset (Stata .dta or .csv with UTF-8 encoding preferred)
- Variable dictionary (.xlsx)
- Field report (sample achieved, response rates, non-response patterns, QA issues)
- Outlier and exclusion logs
For bespoke QA system design or field supervision advisory, contact NM Research and Advisory.