Free template

Survey Questionnaire Design Template

An example questionnaire, pilot plan and ethics note, plus sheets that tie each question to a planned table and check answer formats and skip logic.

Edit with AI Download (.docx + .csv)

Free download · No account needed

What you get

Pack of 7 files: 4 documents, 3 spreadsheets

Survey Questionnaire Design Template

How This Instrument Was BuiltDocument
InstrumentDocument
Pilot ProtocolDocument
Ethics NoteDocument
Construct to Item MapSpreadsheet
Response Option RegisterSpreadsheet
Routing LogicSpreadsheet

Survey Questionnaire Design Template

Construct to Item Map

[Study name], [wave or round]

Expected completes [frame size times the response rate you expect]
Draft item count [before this sheet]
Kept item count [after this sheet]
Read the sheet right to left tables first, then analyses, then items

Construct              Items  Planned analysis          Table    Verdict
[primary outcome]      [n]    [the specific analysis]   [Tbl n]  keep
[primary exposure]     [n]    [the specific analysis]   [Tbl n]  keep
[covariate]            [n]    [the specific analysis]   [Tbl n]  keep
[construct]            [n]    none                      none     cut
[construct]            [n]    none                      none     cut
                       ----                                       
KEPT SUBTOTAL          [n]
CUT SUBTOTAL           [n]
DRAFT TOTAL            [n]    kept plus cut must close

The last two columns are the whole sheet. "Descriptive statistics" is not an analysis and "the results section" is not a table. An item that cannot fill both does not enter the instrument.

Cut list, grouped by reason

Duplicates a construct already measured / no research question behind the construct / construct named with no planned analysis / demographic with neither an analysis nor a benchmark to weight against.

Page 1 of 1165 words

How to design a survey questionnaire

A survey questionnaire is the fixed set of questions, answer options and skip logic that every respondent works through. Design it backward from your analysis. List the tables your report must contain, keep only the questions those tables need, and choose answer formats that support each planned analysis. Then check how many people reach each branch, and pretest anything new before you launch.

Wording problems usually surface in pretesting, while you can still fix them. A badly chosen answer format often goes unnoticed until analysis, when it's too late to ask the same people again.

A person types on a laptop showing a survey form with answer buttons and an interviewer note about recording flight numbers — Each answer option on the screen decides what you can analyze later. Photo: Corey Martin for ODISSEI on Unsplash

Design the questionnaire backward from your tables

Writing a questionnaire forward, one question at a time, is how surveys end up with answers nobody analyzes. US federal statistical surveys start at the other end. The Office of Management and Budget requires agencies starting a new survey to write a plan setting out the tabulations and analytic results that will inform decisions¹.

Write the research questions
Make each one specific enough to decide what to ask: "Is unpaid overtime linked to lower wellbeing?" rather than "What is staff experience like?"
Sketch the table shells
Number each table and figure the report needs, and note its rows, columns and cells.
Name the constructs
These are what each table measures, such as wellbeing or contract type. Give each one a planned analysis, like a mean compared with last year.
Map every question to a table
Cut any question that can't name its construct, its analysis and its table.

ExampleTwo rows from a construct-to-item mapAn illustrative example written for this guide, from a made-up workforce study.

Unpaid overtime hours. Research question: is unpaid overtime linked to lower wellbeing? Items: 4. Analysis: continuous exposure in the main model. Table: Table 4. Verdict: keep.
Manager relationship quality. Research question: none. Items: 3. Analysis: "good to know." Table: none. Verdict: cut.
Whole draft: 68 questions, 41 kept, 27 cut. Median completion time: 21 minutes before, 13 after.

Read it right to left. The table justifies the analysis, and the analysis justifies the question.

Cutting helps with privacy too. In the UK, the Information Commissioner's Office says organizations must not collect personal data on the off-chance that it might be useful in the future². Where answers count as personal data, a question kept "in case it's useful" is exactly what that rules out.

Borrow tested questions before writing new ones

Survey researchers Jon Krosnick and Stanley Presser advise designers to review questions from earlier surveys before writing their own³. Reusing a question lets you compare results across times and populations. That is why, they add, a borrowed question can be worth using even when it departs from best practice.

Comparison only works if the question stays put. To measure change, Pew Research Center advises the same question wording and a similar context⁴, including where it sits in the questionnaire. If one must change, OMB's guidelines suggest bridge studies that preserve trend analyses¹, running old and new versions side by side.

Leave published scales whole

Dropping items, changing the answer scale or rewording a stem turns a validated scale into a new measure without its published reliability. Use a validated short form if one exists, and record every change.

Write questions everyone reads the same way

Krosnick and Presser distill the best advice from hundreds of methodology textbooks into a short list. Use simple, familiar words and simple syntax. Ask about one thing at a time³, make answer options exhaustive and mutually exclusive, and avoid leading questions and double negatives.

Four common faults and their fixes

Fault	Draft	Better
Two questions in one	"How satisfied are you with your pay and benefits?"	Ask about pay and benefits separately
No time frame	"How often do you work late?"	"In the last four weeks, how many extra hours did you work?"
Options that leave people out	"Full-time or part-time?"	Permanent full-time, permanent part-time, fixed term, agency, or other (please specify)
Agree or disagree	"I have enough time for my work." Agree or disagree	"Do you usually have enough time for your work, or not enough?"

The last fault matters because some people tend to agree whatever a statement says, a habit called acquiescence. Across eight studies that Krosnick and Presser review, 14% more people, on average, agreed with an assertion³ than chose the same view when offered a choice between views.

Order questions for the person answering: eligibility and consent first, easy factual questions next, the main outcome while attention is fresh, sensitive topics later and demographics last.

Choose answer formats that fit the analysis

An answer format decides what you can calculate, and it's the hardest thing to fix after launch. Check each one against the analysis it must support.

Formats that block a planned analysis

Planned analysis	Format that blocks it	Ask instead
Mean age, or age groups that match another source	Age ranges	Year of birth, grouped later
Hours as a number	Hour ranges, such as 1 to 5	A number, with a stated maximum
Change since the last round	A 4-point scale after a 5-point one	The earlier scale, word for word
Share using each option	Select all that apply	Yes or no for each option
Comparing groups	Free text	A coded list plus "Other (please specify)"

In a 2018 randomized experiment, Pew Research Center asked 4,581 US adults about undesirable experiences, such as losing a home to foreclosure. Separate yes-or-no questions gave estimates 8 percentage points higher on average⁵ than a select-all list. Pew judged those estimates more accurate and now uses that format online whenever possible.

Give every number field a stated maximum, so impossible values stand out. Record zero, "not applicable" and "skipped" as different answers; once merged, they can't be told apart.

Work out who reaches each branch

Skip logic, also called routing, sends people past questions that don't apply to them. Each condition multiplies, so a block several conditions deep can reach a tiny share of those who finish. Before launch, work out each block's expected base: the number of people you expect to reach it.

ExampleA block four conditions deepAn illustrative example written for this guide, using made-up rates.

Expected completes: 3,400.
Contract includes overtime (59.0%): 2,006.
Reports any overtime (93.9%): 1,884.
Works more than 10 extra hours (33.4%): 629.
Raised a formal complaint (4.1%): 26.
Planned table: 26 people across four contract types, about 6 per cell.

Multiply from the parent. Each base is the previous base times the rate, not all completes times the rate.

Compare each cell with the smallest base you would publish. The US National Center for Health Statistics, for example, suppresses percentages based on a sample of fewer than 30⁶. If a block can't reach your floor, cut it, pool its groups, or report one overall percentage.

Pretest what's new, then launch

OMB's pretesting standard accepts two kinds of evidence that a survey component works: a pretest, or a successful earlier fielding¹. So spend your pretest on what is new or changed, including every answer format you moved.

Start with cognitive interviews. In a typical one, OMB's glossary says, respondents report aloud everything they are thinking¹ as they answer. Then run a small pilot that people complete alone, so the timing is real, and check that someone enters every branch.

Write the decision rules before the pilot runs. For example, any routing error stops the launch, and a median time over your limit means cutting the least valuable questions.

Check whether you need ethics review

In the US, survey-only research with adults can be exempt from the Common Rule, for example when answers are recorded so respondents can't readily be identified⁷. Ask your IRB to confirm. If the study needs review, start from an IRB application template.

Before you launch

Every question names its construct, analysis and table
Trend questions match the earlier round word for word, in the same place
Every answer format supports its planned analysis
Every routed block can fill its planned table
Every branch, including "don't know", leads somewhere
New and changed questions have been pretested

When responses arrive, the open-ended response coding tool turns free-text answers into counted themes, and the survey results report tool writes up the findings. To build the map, format checks and branch sizes from your own research questions, open this template in River.

How to use this template

Use the files as they are, or open them in River. Every file comes filled in for a made-up workforce study, so you can see the method at work before you swap in your own. Choose Edit with AI and River asks for your research questions, any earlier version of the survey and who you plan to invite. Then it builds the Construct to Item Map and, if you send a draft, shows which questions can't name a table.

Download it or open it in River
Take the four Word documents and three CSV sheets as a free zip, or choose Edit with AI to start in River.
Start from your tables
List your research questions and the tables your report needs, then fill in the Construct to Item Map, or let River build it.
Check formats and branches
Test each answer format against its planned analysis, and work out how many people will reach each routed block.
Pretest, then launch
Pilot what is new or changed, apply the rules you wrote in advance, and record each change in the sheets.

What's in the template

Construct	Research question it answers	Items	Planned analysis	Table
Wellbeing (primary outcome)	Did wellbeing change between wave 3 and wave 4?	8	Mean scale score; paired change against wave 3	Table 3, Figure 2
Unpaid overtime hours (exposure)	Is unpaid overtime associated with lower wellbeing?	4	Continuous exposure in the main model	Table 4 cols 1–3
Job demand	Does demand account for the association?	5	Covariate, mean of five items	Table 4
Supervisor support	Does support moderate it?	4	Covariate and interaction term	Table 4, Table 5
Contract and working hours	Which contract groups carry the exposure?	6	Stratifier; drives the overtime routing	Table 1, Table 2
Sickness absence	Does absence move with wellbeing?	3	Secondary outcome, count model	Table 6
Carer responsibilities	Do employees with caring duties differ?	4	Subgroup definition	Table 7
Tenure and role	Standard adjustment	4	Covariate	Table 1, Table 4
Consent and administration	Eligibility and record linkage	3	Defines exclusion rules E2 and E5	Exclusion cascade
Kept subtotal		41

Reason it was cut	Items
Duplicates a construct already measured, with no reliability comparison planned	11
Construct named, no planned analysis ("good to know", "a future paper")	8
Demographic with neither a planned analysis nor a benchmark to weight against	5
No research question behind the construct at all	3
Cut subtotal	27
Draft total: 41 kept plus 27 cut	68

Item	Format as drafted	Planned analysis	Supports it	What it forecloses	Fixable after fielding
Age	Six age bands	Continuous covariate; weighting against a benchmark published in single years	No	A mean age, a continuous term, and any weight built on age groups that cut across the six bands	No
Wellbeing items (8)	Four-point forced choice	Mean scale score compared against wave 3	No	The wave 3 comparison, which is why wave 4 is being fielded at all	No
Unpaid overtime hours	Select one of five bands	Continuous exposure; hours compared against wave 3	No	A mean, a per-hour coefficient, and the wave 3 comparison	No
Benefits used	Select all that apply	Percentage using each benefit	No	A stated denominator. The percentages sum past 100	No
Contract type	Free text	Four-level stratifier in Table 2	No	A clean stratifier. Free text produces spellings, not levels	Partly
Overtime provision	Yes / No	Eligibility gate for the overtime block	Yes	Nothing	n/a
Dependants needing care	Yes / No	Gate and subgroup definition	Yes	Nothing	n/a
Tenure	Numeric months	Covariate	Yes	Nothing	n/a
Sickness absence days	Numeric days in 12 months	Secondary outcome, count model	Yes	Nothing	n/a
Employer	Coded list of 11 employers	Cluster identifier for the variance estimator	Yes	Nothing	n/a
5 of 10 formats cannot support their planned analysis	4 of those 5 are unrepairable after fielding

ID	Block	Shown when	Rate	Applies to	Expected base	Cells	Expected cell	Verdict
R1	Overtime	Contract carries an overtime provision	59.0%	3,400	2,006	4	501	Field
R2	Hours detail	R1 and any overtime reported	93.9%	2,006	1,884	4	471	Field
R3	Reasons	R2 and more than 10 hours	33.4%	1,884	629	4	157	Field
R4	Formal complaint	R3 and a complaint was raised	4.1%	629	26	4	6	Cut
R5	Carer block	Reports dependants needing care	28.0%	3,400	952	2	476	Field
R6	Night-shift carer	R5 and works night shifts	11.2%	952	107	11	10	Restructure
R7	Absence detail	Reported any absence	63.1%	3,400	2,145	4	536	Field
R8	Return to work	R7 and absence over 7 days	21.4%	2,145	459	4	115	Field
R9	OH referral	R8 and referred	12.6%	459	58	4	14	Restructure
9 routed blocks	3 cannot support the table they were written for: 1 cut, 2 restructured

InstrumentThe example questionnaire, block by block, with each question's wording, answer options and the reason for its form.
How This Instrument Was BuiltThe four decisions you can't undo after launch, and the order the files are worked in.
Pilot ProtocolWhat the pilot tests, who takes it, and which results would change the questionnaire.
Ethics NoteWhat respondents are told, why each question is collected, and how consent and sensitive items are handled.
Construct to Item Map sheetEach construct with its research question, item count, planned analysis, table and a keep or cut verdict.
Response Option Register sheetEach answer format checked against its planned analysis, with a fix and whether it can be repaired after launch.
Routing Logic sheetEach routed block with its condition, expected rate and base, smallest table cell and a verdict.

Common questions

What is the difference between a survey and a questionnaire?

A questionnaire is the instrument: the questions, answer options and skip logic people see. A survey is the whole study around it, including who you sample, how you contact them and how you analyze the answers. This template covers the questionnaire, from planning the tables to the pretest, and stops at launch.

How many questions should a survey have?

As many as your planned tables need, and no more. No standard count exists, so judge length by completion time and measure it in a pilot. In the template's made-up workforce study, cutting a 68-question draft to the 41 that fed a table brought the median time down from 21 minutes to 13.

Should I ask age as a number or in ranges?

Ask for year of birth or exact age, then group it at analysis time. Ranges can't give you a mean age, can't enter a model as a number, and can't be regrouped to match population figures or an earlier round grouped differently. A number can always be grouped later, but a range can never be split.

How do I pretest a questionnaire?

Begin with cognitive interviews, where a few people think aloud while answering so you hear how they read each question. Fix what confused them. Then run a small unassisted pilot to get a real completion time, make sure someone qualifies for every branch, and agree beforehand which results will change the questionnaire.

What is a construct in survey research?

A construct is the thing a set of questions measures, such as wellbeing, job demand or contract type. Some, like wellbeing, can't be observed directly, so you measure them with several questions. Mapping questions to constructs shows where two questions measure the same thing and where a construct has no question at all.

Is this survey questionnaire template free?

Yes. The four Word documents and three CSV sheets download as a zip with no signup. They come filled in with a made-up workforce study, which shows how each file works before you replace it with your own. Edit with AI opens the same files in a free River workspace.

What does Edit with AI do?

It opens the template in a free River workspace and asks for your research questions, any earlier version of the survey to stay comparable with, and who you'll invite. River builds the Construct to Item Map first. Send a draft and it places every question, then groups any question without a table by reason, with completion time before and after.

Sources

Office of Management and Budget. “Standards and Guidelines for Statistical Surveys.” obamawhitehouse.archives.gov (opens in a new tab)
Information Commissioner's Office. “Principle (c): Data minimisation.” ico.org.uk (opens in a new tab)
Handbook of Survey Research. “Question and Questionnaire Design.” web.stanford.edu (opens in a new tab)
Pew Research Center. “Writing Survey Questions.” pewresearch.org (opens in a new tab)
Pew Research Center. “When Online Survey Respondents Only ‘Select Some That Apply’.” pewresearch.org (opens in a new tab)
National Center for Health Statistics. “National Center for Health Statistics Data Presentation Standards for Proportions.” cdc.gov (opens in a new tab)
eCFR. “45 CFR 46.104: Exempt research.” ecfr.gov (opens in a new tab)

Written by

Chandler Supple

Co-founder & CEO, River

Chandler has spent the last six years working in AI, building tools for millions of users and Fortune 100 companies. In 2025, he co-founded River to build a new kind of operating system for AI agents.

LinkedIn GitHub X

Reviewed by

Brian Joseff, Co-founder, River

LinkedIn X

Survey Questionnaire Design Template

[Study name], [wave or round]

Cut list, grouped by reason

How to design a survey questionnaire

Design the questionnaire backward from your tables

Borrow tested questions before writing new ones

Write questions everyone reads the same way

Choose answer formats that fit the analysis

Work out who reaches each branch

Pretest what's new, then launch

How to use this template

Download it or open it in River

Start from your tables

Check formats and branches

Pretest, then launch

What's in the template

Construct to Item Map

Response Option Register

Routing Logic

Common questions

Sources

Chandler Supple

Related tools and templates

More free templates