Community Survey in India: How to Design One and Run It Well

A community based survey collects data from people who share a place, an institution or an identity. The shared thing is what makes the group worth studying, and it is also what makes the sample hard to draw.

The output is a description of that group, not an estimate for India. If you need a national estimate, use a probability panel. If you need to act inside one community, a community survey will give you more usable answers per rupee.

Write the intended use in one sentence before designing anything. A survey built to rank problems for a local plan needs different questions from one built to test a campaign message.

A boundary you cannot draw is a boundary you cannot sample. Decide whether the community is a place, a membership, a service user group or a language group, then write the inclusion rule in plain words.

For a place, name the streets, blocks or habitation list. For a membership, name the register. For a service group, name the period of use. Vague boundaries produce a sample nobody can replicate.

Check whether the boundary you chose matches the decision. A ward-level plan needs ward residents, not everyone who shops in the ward. A clinic study needs patients, and often the family member who decides whether to return.

Decide who answers for a household. One adult per household is a common rule, but the person who manages the kitchen, the school run or the clinic visit may not be the person listed as head of household.

A sampling frame is the list you draw from. In community work the best frame is often physical: a habitation list, a voter roll extract, a society register, a school enrolment list, or a map with household counts per lane.

Where no list exists, use a route-based approach. Walk a defined path, count structures, and apply a fixed selection rule such as every third dwelling. Record the path and the rule so another team can repeat it.

Count before you sample. A quick listing round that records structure type, estimated households and whether anyone is present will tell you whether your target number is achievable in the time you have.

Keep a replacement rule for locked homes and refusals, and log every replacement. Silent substitution is the most common way a community sample drifts away from the group it claims to describe.

In most Indian community settings, access runs through someone: a resident welfare association, a ward officer, a school principal, a mosque or temple committee, a self-help group, or an employer at a worker settlement.

Ask early, and be specific about what you need: dates, hours, number of interviewers, languages, and whether you will photograph or record anything. Vague requests get refused or delayed.

Explain the purpose in the language of the community, not in research vocabulary. Say who commissioned the study, how long an interview takes, what is voluntary, and how the data will be stored.

Getting permission is not the same as getting consent. Consent is collected from each respondent, separately, and can be withdrawn. A letter from a local body does not substitute for that.

Draft in the language of the interview, not in English and then translate. Translation after the fact tends to preserve the grammar of the source and lose the meaning.

Keep the questionnaire short. Community respondents are often interviewed at the door, in heat or in the middle of other work. A twenty-minute instrument needs a reason.

Use local terms for the things you ask about. A question about a ration card, a water connection or a school fee will be answered differently depending on which local word you use.

Pilot with people who look like the sample, including at least one low-literacy respondent. Read the pilot out loud and note every question that needs repeating.

Avoid asking for income directly if the decision does not need it. Ask about the things people are willing to discuss, such as work type, travel time, or what they did the last time a service failed.

Hire interviewers who speak the local language and, where it matters, who share the respondent's gender. A mismatched interviewer changes what people are willing to say about health, money and family decisions.

Set the day and time around the community's routine. In many settlements, the reliable window is early morning or after dusk, not the middle of the working day.

Back-check a sample of interviews by phone or in person. Ask two or three factual questions from the interview and confirm that the respondent remembers it.

Record the outcome for every contact: completed, refused, locked, not eligible, replaced. The refusal pattern is itself a finding, and it belongs in the report.

Follow the data protection rules that apply to the data you collect. Personal data needs a stated purpose, a retention period and a way to delete it on request.

Report the group, the frame and the response rate before you report percentages. A number without its denominator is not a finding.

For small communities, do not break results into many subgroups. A ward sample of 120 completes cannot support six age bands and two genders at once.

Pair every table with the decision it informs. A community needs assessment should end with a ranked list of problems, the evidence behind each rank, and who is responsible for acting on it.

Say what the study cannot tell you. If the sample excludes people who moved away, people without a phone, or people who refused, name them and describe how that could shift the result.

Use the platform for the questionnaire and the quotas, and keep the listing work with local field staff. Set quotas by lane, dwelling type or gatekeeper group so the achieved mix stays visible.

Test the instrument on the free 30 responses in the field language before the main round, then check drop-off and completion time before you expand.