Good survey design is the foundation of good analysis. If questions are unclear, sequencing is weak, or response options are poorly defined, no amount of post-hoc cleaning can fully recover data quality.

Step 1. Define Decisions Before Drafting Questions

Start with:

  • Primary research question
  • Indicators required for analysis
  • Target population
  • Planned outputs (tables, models, policy use)

Every survey question should map to a clear analytical purpose.

Step 2. Build a Variable Plan

Before finalizing wording, create a variable plan that includes:

  • variable name
  • concept definition
  • response type and coding
  • expected range
  • skip conditions

This reduces ambiguity between survey design and data processing.

Step 3. Write for Comprehension, Not Technical Precision

Use short, concrete language. Avoid double-barreled questions and abstract terms that respondents may interpret differently.

Better practice:

  • one idea per question
  • clear recall period (e.g., last 7 days, last month)
  • response options that are mutually exclusive and complete

Step 4. Design Logical Flow

A practical order:

  1. Intro and consent
  2. Easy factual questions
  3. Core modules
  4. Sensitive modules later
  5. Closing and verification

Flow affects respondent comfort and completion quality.

Step 5. Pilot, Then Revise Aggressively

Piloting should test comprehension, timing, and skip logic under real conditions. Remove weak questions early.

Track pilot findings systematically:

  • misunderstood items
  • frequent “other” responses
  • interviewer confusion points
  • timing bottlenecks

Step 6. Plan Data Quality During Collection

Daily checks should cover:

  • completeness
  • out-of-range values
  • duplicate IDs
  • unusual interview durations
  • enumerator-level error patterns

Quality control is most effective while fieldwork is still active.

Step 7. Prepare Cleaning Rules Before Analysis

Define cleaning logic in advance to avoid ad hoc decisions:

* Example checks
duplicates report household_id
misstable summarize
assert age >= 0 & age <= 120 if age < .

Predefined rules improve transparency and reproducibility.

Step 8. Keep Instruments Manageable

Long questionnaires increase fatigue and reduce response quality. If a module does not support the core research objective, remove it.

Short, focused instruments usually produce better data than broad, overloaded ones.

Final Standard

A well-designed survey is clear to respondents, easy for enumerators to implement, and directly usable for analysis.

Survey quality is not about asking more questions; it is about asking the right questions in the right way.