# Survey Questionnaire Design: A Step-by-Step Guide for India

Design a survey questionnaire that avoids bias: map questions to decisions, order blocks, fix double-barreled questions, pilot it, and a 12-point checklist.

Published 2026-10-16 · Updated 2026-11-16
Canonical: https://hercules.works/blogs/survey-questionnaire-design-india

Write the decision first, then one question per piece of evidence the decision needs. Order the survey screener, warm-up, core, profile. Ask one thing at a time, keep answer lists short, randomise unordered options and pilot before you launch. Pew Research notes that small wording changes can substantially shift answers. Use the checklist below before you field. Hercules Works can draft a questionnaire from a one-paragraph brief and you edit it by chat.

## Every Question Must Earn Its Place

Most survey problems are written into the questionnaire, not the analysis. Leading wording, two questions in one and answer lists that miss obvious options cannot be fixed afterwards. A short, well-ordered questionnaire beats a long one.

This guide follows a fixed sequence and ends with a checklist. The principles on wording, order, answer options and double-barreled questions come from Pew Research Center's page on [writing survey questions](https://www.pewresearch.org/writing-survey-questions/). The rewrites and the checklist are illustrative.

## Start From the Decision, Not the Questions

For each question, write the decision it feeds. If you cannot, delete the question. Use the [brief template](/blogs/research-for-business-decisions-india) to fix the decision, audience and study type before you write a line.

Then list the evidence the decision needs, and write one question per item. Put it in a table so nothing is missed.

| Decision | Evidence needed | Question |
|---|---|---|
| Which pack to launch | Preference between two packs | Which of these two packs would you choose? |
| Whether to enter a city | Share who buy the category there | Which of these have you bought in the last month? |
| Price to set | Acceptable price range | Four price-sensitivity questions |

## Order the Blocks

1. **Introduction.** Purpose in one line, length, privacy. Do not name the brand being tested.
2. **Screener.** Confirms the person is in your audience. Screen-outs end the survey politely.
3. **Warm-up.** Easy behavioural questions about the category.
4. **Core.** Unaided questions before aided ones, general before specific. Pew notes that earlier questions can unintentionally provide context for the questions that follow.
5. **Profile.** Age band, gender, city. Put sensitive items last.
6. **Close.** Thank and, if needed, ask consent for follow-up.

## Wording Rules and Rewrites

| Problem | Weak question | Better question |
|---|---|---|
| Double-barreled | How satisfied are you with the price and quality? | Two questions: one on price, one on quality |
| Leading | How much do you love our fast delivery? | How would you rate the delivery? |
| Loaded term | Do you support the government's pro-farmer reforms? | How much do you support or oppose [the specific reform]? |
| Vague | Do you shop online often? | In the last 30 days, how many times did you buy online? |
| Jargon | How do you rate the UX? | How easy was it to find what you wanted? |

Pew's advice also covers answer options: keep the number of choices small, randomise unordered lists, and note that forced-choice questions tend to yield more accurate responses than select-all lists, especially for sensitive questions. Use select-all sparingly. For rating scales, see the [Likert guide](/blogs/likert-scale-survey-creator).

## Design for Indian Respondents on Phones

- **Mobile first.** Most answers arrive on a phone, so avoid wide grids and long scrolls.
- **Language.** Write each question in the respondent's language and have a native speaker check it. Test that a translated scale still reads as balanced.
- **Length.** Cut anything that does not feed a decision. Check the length on a phone, not on a laptop.
- **Consent.** Respondents under 18 need parental consent, see the [teen survey guide](/blogs/top-survey-platforms-for-market-research-on-teenagers-2026).

## A 12-Point Checklist Before You Launch

1. Each question feeds a named decision.
2. One idea per question.
3. No leading or loaded words.
4. Answer lists cover the real options, with "other" and "none" where needed.
5. Unordered options are randomised.
6. Scales are balanced and labelled the same way throughout.
7. Unaided questions come before aided ones.
8. The screener does not give the answer away.
9. Sensitive questions are last and optional.
10. Routing sends each person to the right questions.
11. The survey is tested on a phone and in each language.
12. A pilot of a small group is read, not only counted.

## A Sample Question Sequence

This is an illustrative sequence for a packaged-snack concept test. Adapt the wording to your own category.

| # | Block | Question wording | Type |
|---|---|---|---|
| 1 | Screener | Which of these did you buy in the last 7 days? (list with "none of these") | Multi-select |
| 2 | Screener | In which city do you live? | Single |
| 3 | Warm-up | How often do you buy packaged snacks? | Single |
| 4 | Core, unaided | Which snack brands can you name? | Open text |
| 5 | Core, aided | Which of these brands have you heard of? | Multi-select |
| 6 | Core | Here is a new snack. How likely are you to buy it? | 5-point scale |
| 7 | Core | What is the main reason for your answer? | Open text |
| 8 | Core | Which of these two packs would you choose? | Single |
| 9 | Profile | Which age band are you in? | Single |
| 10 | Close | May we contact you again about this study? | Yes/No |

Notice that question 4 comes before 5. If you list brands first, you cannot learn which ones people remember on their own. Question 1 includes "none of these" so the screener does not force a false answer.

## Designing Answer Options

Options must not overlap and must cover every case. Compare two lists for a spending question.

| Faulty list | Problem | Better list |
|---|---|---|
| Up to ₹500, ₹500 to ₹1,000, ₹1,000 to ₹2,000 | Someone who spent ₹500 fits two options, and nothing covers above ₹2,000 | Less than ₹500, ₹500 to ₹999, ₹1,000 to ₹1,999, ₹2,000 or more |
| Never, sometimes, often | No middle for "every week", and "often" means different things to different people | Not in the last month, 1 to 3 times, 4 to 8 times, more than 8 times |
| Satisfied, Very satisfied, Dissatisfied | Unbalanced and has no neutral point | Very satisfied, Satisfied, Neither, Dissatisfied, Very dissatisfied |

Put numeric ranges on real behaviour, with a time window. Add "don't know" only where a person may honestly not know, such as the price of an item they did not buy. Add "other, please specify" where your list may be incomplete. Pew also suggests offering a choice between alternative statements as a better practice than agree-disagree statements ([Pew](https://www.pewresearch.org/writing-survey-questions/)).

## Count Items, Not Questions

Respondent burden depends on the number of things people must answer. A grid of 8 rows is 8 items. Take a survey of 15 questions where two are grids of 8 rows. It has 13 + 16 = 29 items, not 15.

Audit your draft in a table: question number, type, number of items, decision fed. Cut any row with no decision. Break long grids into smaller questions or cut rows. Then take the survey on a phone, in each language, and note where you got tired.

We do not set a standard length. The right length is the one where people still read the last question. A pilot will show where they stop.

## Mistakes That Bias the Answers

| Mistake | Why it biases results |
|---|---|
| Showing the brand name in the introduction | People answer for the brand, and later unaided questions are contaminated |
| Starting with demographics | Feels like a form, so people skim |
| Using "agree" statements for most questions | Invites agreement whatever the content; Pew describes this as acquiescence |
| A select-all list that is long | People tick the first few and stop |
| Asking about the future ("Would you buy?") with no context | Stated intent is easy to overstate, so treat it as a ranking and not a forecast |
| Translating word for word | A translated scale can read as stronger or weaker |

For translation, ask one person to translate and another to translate back, then compare the two English versions. Differences show where meaning moved. Then test the final version with a few native speakers.

## Fill-In Revision Log

Keep this log for every draft. It forces you to say why each change was made.

| Question | Problem found | Source (read-through, pilot, interview) | New wording |
|---|---|---|---|
| Q__ | | | |
| Q__ | | | |

Have one person who did not write the survey and does not agree with you read it aloud. Mark every place where they hesitate.

## Running This Study on Hercules

If your job is a consumer survey in India, the AI survey builder drafts the questionnaire, screener and routing from your brief, then checks it against a quality rubric. You edit by conversation. This is the pilot and main study at list price.

Pricing is per response, and the free plan covers a 30-response pilot: see [pricing](/pricing). How respondents are verified and sampled: [panel methodology](/panel-methodology).

The limit: the app reaches smartphone users only, so offline households need face-to-face fieldwork.

## Frequently asked questions

### How do I design a survey questionnaire?

Start with the decision, list the evidence it needs, write one question per item, order the survey as screener, warm-up, core and profile, pilot it, then launch.

### What is a double-barreled question?

A question that asks about two things at once, such as price and quality. Split it into two questions so each answer means one thing.

### How do I avoid biased survey questions?

Use neutral wording, balanced scales, a short list of options, randomised order for unordered lists and a pilot. Have someone who disagrees with you read the draft.

### How long should a survey questionnaire be on a phone?

As short as the decision allows. Cut every question that does not feed a decision and check the length on a phone.

### Can AI design a questionnaire for me?

It can draft one. Review every question for leading wording, check routing and pilot it before the main study.

### Should the first question be about demographics?

No. Start with the screener, which decides eligibility, then easy behaviour questions. Put age band, gender and city near the end, with sensitive items last.

### When should I add a "don't know" option?

Add it when people may honestly not know, such as the price of something they did not buy. Leave it out when the question asks about the respondent's own experience, since it invites skipping.

### How do I test my questionnaire before launch?

Read it aloud, walk through every route in the preview, take it on a phone, then run a pilot and read the open answers. See the [pilot survey guide](/blogs/pilot-survey-india).
