FIELD GUIDE / 05
Field research turns a protocol into real observations. Reliable data depends on questions people understand, tools that prevent avoidable errors, trained teams, ethical consent, active quality monitoring, and documentation of what actually happened—not only what was planned.
Follow the reasoning
Start in order, but do not treat research as a straight line. Later evidence often sends you back to fix an earlier decision—and that is part of doing the work properly.
Use simple language, defined reference periods, balanced response options, and validated instruments where suitable. Avoid double questions, hidden assumptions, leading wording, and unnecessary identifiers.
Quality check: Every question should map to a variable, objective, or essential operational need.Use skip logic, range checks, required fields, calculated values, repeat groups, unique IDs, and safe defaults in tools such as ODK. Keep version control and test the exact form that will go to the field.
Quality check: Validation rules should block impossible values without blocking unusual but valid realities.Test consent, wording, timing, translations, device workflow, connectivity, uploads, supervision, and respondent burden with people similar to the intended population.
Quality check: A pilot tests the system, not merely whether the form opens.Use a shared manual, role-play difficult situations, standardise measurements, practise consent, define escalation, and assess competency before independent collection.
Quality check: Daily debriefs reveal misunderstandings before they become a large batch of unusable records.Track completeness, duplicates, duration, missingness, unusual distributions, enumerator patterns, timestamps, locations where appropriate, and protocol deviations. Query records without pressuring teams to fabricate corrections.
Quality check: Quality control is faster and fairer when checks are planned before the first interview.Minimise identifiers, encrypt devices and transfers, restrict access, back up safely, separate contact information, document changes, and define retention and deletion procedures.
Quality check: A clean dataset without consent, privacy, and traceability is not good research data.See it in context
Create a short, pilot-tested digital questionnaire with unique anonymous IDs, standard definitions, programmed checks, trained enumerators, and a daily dashboard for completeness, duration, missingness, and unusual response patterns.
Clear reference periods, suitable validated scales, neutral wording, and translated versions where needed
Eligibility gate, consent, skips, ranges, required critical fields, and review screen
Role-play, measurement practice, privacy scenarios, troubleshooting, and competency checks
Daily counts, missingness, durations, duplicates, outliers, and interviewer-level patterns
Documented queries and audit trail; never silent overwriting or invented answers
Minimal identifiers, encrypted transfer, role-based access, backups, and a retention plan
Before you move on
Every question has a purpose and is understandable to the intended participant.
01The exact form, devices, consent process, and transfer workflow were piloted.
02Field staff are trained, supported, and assessed consistently.
03Quality indicators and escalation rules are monitored during collection.
04Privacy, access, backups, corrections, and deviations are documented.
05Field data quality is produced during collection; it cannot be fully manufactured later during cleaning.
Follow the next decision
A good question should shape the literature search; the evidence should sharpen the design; and the methods should make the result possible to understand and challenge.