Questionnaire Design
Turn a research objective into a question people can understand, answer, and place within a clear set of options. Then test the route from their experience to the data you will analyse.
What you will learn
- Build a measurement plan linking objectives, concepts, items, and analysis.
- Write clear questions and construct useful response categories.
- Choose formats and scales suited to what you want to measure.
- Plan question order, filter routes, language, and accessible presentation.
- Pretest, revise, code, and document a questionnaire.
Plan: Use Sections 1–3 to define what each item should contribute.
Build: Work through wording, options, scales, routing, and presentation in Sections 4–8.
Test: Apply Sections 9–11 to a draft. Watch the lecture and complete the workshop in Section 12.
Begin with the research objective
Questionnaire design translates a research purpose into a sequence of answerable items. Each item should collect information that contributes to a planned description, comparison, explanation, or decision.
The lecture makes a useful starting point: write the objectives and an analysis plan before drafting questions. “Understand students’ media literacy” names a broad aim. A more workable objective might be to describe whether students checked a news claim before their most recent act of sharing it. That objective specifies a behaviour, an occasion, and a possible descriptive result.
- Define the objectiveDescribe checking before the most recent news-sharing occasion within a specified period.
- Specify the indicatorA respondent’s report of attempting to check the claim before that occasion.
- Draft the item and routeIdentify recent sharers, then ask them about that particular message.
- Plan the resulting claimReport the relevant responses among the eligible respondents, with the denominator and missing information explained.
Create a measurement blueprint
A construct is the concept you want to investigate, such as confidence in checking news. An indicator specifies an observable sign of it. An item is the question, statement, or task through which the response is obtained.
On a narrow screen, scroll the table sideways.
| Objective | Information needed | Planned use |
|---|---|---|
| Describe checking before recent sharing | Whether a respondent shared news in the reference period; checking before the most recent occasion. | A distribution among eligible recent sharers. |
| Describe perceived checking ability | A clearly labelled confidence rating. | The distribution of perceived confidence, considered separately from demonstrated skill. |
| Explore barriers to checking | An open account of difficulties, if any. | A documented coding or qualitative analysis of the responses. |
For every proposed item, complete: “I need this answer because ______, and I will use it to ______.” This test helps you remove interesting questions that add burden without serving the study.
Distinguish behaviour, beliefs, knowledge, and attitudes
A questionnaire can capture different kinds of information. Naming the kind helps you choose the wording, response format, and interpretation.
Behaviour
What people report doing: “In the seven days ending yesterday, did you share a news message through a messaging app?”
Belief
What people regard as true: “How likely do you think it is that this message came from an official source?” Specify the message.
Knowledge
What people can answer accurately against a defensible key: for example, identifying the publisher named in a supplied news report.
Attitude
A person’s evaluation: “To what extent do you support or oppose the university offering a session on checking news claims?”
Attribute questions describe relevant characteristics, such as programme level or years of professional experience. Collect each characteristic for a stated purpose and use categories that suit the intended respondents.
Follow the respondent’s answer process
The CDC’s cognitive interviewing guidance draws attention to comprehension, recall, judgement, and response. These steps help you diagnose why an apparently simple question may produce different answers.
Choose an appropriate time window and level of detail. Counting every news message over a year may exceed memory. Asking about the most recent occasion can make recall more manageable, while changing the finding to one about that occasion. State which you need.
Match the response format to the task
The response format determines what people can express and how their answers enter the dataset. Choose it through the information needed and the effort required.
On a narrow screen, scroll the table sideways.
| Format | Useful contribution | What to plan |
|---|---|---|
| Open-ended | Respondents explain an experience in their own words and may introduce unanticipated issues. | A focused prompt, manageable writing demand, and a coding or qualitative analysis plan. |
| Single choice | One category represents the answer. | Clear instructions and options that cover relevant answers without overlap. |
| Multiple response | Several activities or experiences can be selected. | “Select all that apply”; exclusive handling of “none”, uncertainty, or refusal where offered. |
| Partly closed | A list provides structure while “Another reason: please specify” allows an addition. | A plan for reviewing and coding write-ins. |
| Numerical entry | A count or quantity in specified units. | A meaningful reference period, plausible range, and a way to express uncertainty. |
| Rating or ranking | A rating gives each object a level; a ranking orders objects relative to one another. | Clear anchors for ratings; a short, manageable set and explicit tie rules for rankings. |
A rating of two podcasts can give both the same score. Ranking them asks which comes before the other. Select the task through the comparison you actually want.
Choose open questions deliberately
“What, if anything, makes checking a news claim difficult for you?” can reveal barriers that a fixed list misses. A long series of such questions may increase writing effort and leave a complex analysis task. Place open items where the additional detail serves a clear purpose.
In an interviewer-administered survey, an open answer may be recorded using a hidden coding list. The respondent still generates the answer without hearing the options. Train interviewers and document how unmatched answers are captured. This differs from reading a closed list aloud.
Write one clear, neutral question at a time
A good question tells respondents what to consider and asks for an answer they can reasonably provide. Familiar words, a defined reference period, and a single focus carry much of the work.
The revisions below show how a change in wording changes the response task. Treat each revision as a draft to test with the intended respondents.
On a narrow screen, scroll the table sideways.
| Initial wording | Design issue | A stronger draft |
|---|---|---|
| “How regularly do you consume digital news?” | “Regularly”, “consume”, and “digital news” may invite different interpretations. | Define the material, then ask: “On how many of the seven days ending yesterday did you read, watch, or listen to news online?” |
| “How satisfied are you with the accuracy and speed of this news app?” | Two judgements may receive different answers. | Ask separately about satisfaction with the accuracy of its reports and the speed of its updates. |
| “Do you agree that responsible students should always verify news?” | The wording signals a socially approved answer and embeds an absolute standard. | For behaviour: “Before sharing that message, did you try to find out whether its main claim was accurate?” |
| “Do you disagree that students should not share unverified news?” | Negation makes the response direction difficult to track. | For an attitude: “How acceptable or unacceptable is sharing a news claim before checking it?” |
| “Why have you stopped reading newspapers?” | The question assumes both previous readership and stopping. | Establish previous and current readership first. Ask about reasons for stopping only where applicable. |
| “How many hours did social media make you waste last year?” | A loaded judgement, a demanding recall task, and an assumed causal interpretation. | If the objective concerns time use: “About how many minutes did you spend using social media yesterday?” Define included activities and allow an estimate. |
Specify the reference frame
A question may need a person, activity, place, period, or occasion. “Your main news app” requires a rule for choosing one if the respondent uses several. “Last month” may mean the previous calendar month; “the past 30 days” describes a rolling period. Choose the frame that fits the objective and use it consistently.
Definitions should settle real ambiguities. In the worked example, “share” includes sending or forwarding a message, link, image, or clip to another person or group. Keep the definition close to the relevant item and test whether people apply it as intended.
Give respondents a fair task
People can often report their own experiences more directly than estimate what everyone else does. If you need their perception of others, identify it as such. Hypothetical scenarios can measure expectations or intended responses when the scenario is precise and relevant. Interpret those responses within that frame.
Construct response options that fit the question
Response options translate a person’s answer into a recorded category. They should cover the relevant possibilities, preserve meaningful distinctions, and make the response task clear.
Build categories without gaps or overlap
For a single-choice age band, 18–24, 25–34, and 35 or older are distinct for an adult sample. “18–25” followed by “25–35” leaves a 25-year-old with two eligible categories. The complete list must also fit the study’s actual age range.
Match categories to the unit. For “On how many days…?”, the integers 0 through 7 provide an exact bounded response. For an attitude, use verbal labels that describe the dimension. Check whether “Other”, “None”, or “Not applicable” represents a real possibility.
Strongly oppose · Somewhat oppose · Neither support nor oppose · Somewhat support · Strongly support
Separate different reasons for a non-substantive answer
Neutral position
“Neither support nor oppose” locates an attitude on the stated continuum.
Don’t know or cannot remember
These identify uncertainty about an opinion, fact, or recalled event. Choose the label that fits.
Not applicable
The question does not fit the respondent’s circumstances. A filter may handle this more clearly.
Prefer not to answer
The respondent declines to provide the information. Preserve that distinction in the data.
Choose a midpoint through the construct and intended task. An even-numbered scale asks respondents to choose a direction; it may push someone with a neutral position into an inaccurate category. An explicit uncertainty option can increase its use, while revealing that some respondents lack the information needed to answer.
Handle multiple-response items explicitly
Write “Select all that apply” when several choices can be true. Keep “None of these” mutually exclusive with substantive choices. In an online instrument, program and test that rule. If an interviewer records answers, provide equivalent instructions.
Choose and evaluate rating scales
A rating scale asks respondents to locate an evaluation, experience, or perception along a defined dimension. Its labels determine what the recorded levels mean.
Likert-type items and multi-item scales
A Likert-type agreement item presents a statement and asks for a level of agreement. For example: “I feel confident checking the main claim in a news message.” A five-category response might run from strongly disagree to strongly agree, with a neutral midpoint.
A Likert scale combines several related items intended to measure a construct. Begin with a definition of that construct and a considered item pool, then evaluate content, response processes, structure, and score reliability. Several items sharing the same response labels do not automatically form a coherent scale.
Question: “How confident are you that you could check whether a news message’s main claim is accurate?”
Options: Not at all confident · Slightly confident · Moderately confident · Very confident · Completely confident
These labels address confidence directly. A separate “Cannot judge” option can capture difficulty assessing one’s ability.
Semantic differential
A semantic differential places an object between opposing descriptors on a specified dimension. After showing a particular news explainer, you might ask respondents to rate its language from “Very unclear” through a neutral point to “Very clear”. Choose genuinely opposed anchors and keep each item focused on one dimension.
Guttman or cumulative scaling
A Guttman scale seeks an ordered pattern in which endorsing a higher-level item tends to imply endorsing lower-level items. Establishing that pattern requires empirical evaluation. A set of yes/no items about increasingly demanding checking tasks may suggest a candidate ordering, but respondents can follow different patterns.
Plan scoring and interpretation
A single agreement item supplies ordered categories. Combining items into a score requires evidence and a stated rule. Describe which items enter the score, their direction, the treatment of missing responses, and what higher scores mean.
For an item scored 1–5, reverse scoring uses 6 − original score. A response of 2 becomes 4. Apply this only where the item’s meaning requires the opposite scoring direction. Reversed wording can itself confuse respondents, so test it rather than adding it mechanically to counter agreement tendencies.
Plan question order and filter routes
Question order builds a path through the topic. A clear path introduces the task, establishes relevance, groups related items, and guides each respondent to the questions they can answer.
Start with an understandable introduction and any necessary consent or eligibility steps. Then use easy, relevant questions to establish the topic. Move from concrete experiences to more demanding judgements where this serves the design. Place sensitive characteristics later when possible; ask an essential eligibility characteristic early when the route requires it.
Use filters to establish applicability
In the worked questionnaire, a question about checking before a recent share applies to students who actually shared news in that period. The route below connects Q1–Q3 with the common questions Q4–Q7.
No, cannot remember, or prefer not to answer → Q4.
Check context and order effects
A detailed block about misinformation may influence a later general judgement of trust in news. Decide whether an unprompted general judgement belongs before that block. Preserve wording and context when comparing waves; if a change is necessary, consider testing the old and new versions through a randomised split-ballot design.
List-position effects can also vary with presentation mode. Visual lists may favour earlier options; orally presented lists may favour recent ones. The direction and size depend on the task. Pew Research Center’s guidance offers useful examples.
Adapt language, presentation, and administration
A questionnaire works through a particular language, device, setting, and mode of administration. Design and test that encounter as carefully as the individual item.
Self-completion online
Use clear labels, visible instructions, large selection targets, and a layout that works on a phone. Avoid grids that force repeated sideways scrolling.
Interviewer administration
Specify what is read aloud, which prompts are permitted, how uncertainty is recorded, and how routing is followed.
Paper questionnaires
Provide readable spacing, sufficient writing room, obvious selection marks, and skip instructions beside the relevant answer.
Accessible completion
Check keyboard operation, screen-reader labels, contrast, text size, and whether meaning depends on colour or a visual arrangement alone.
Translate the response task
For a Bengali–English questionnaire, examine whether both versions ask the same substantive question with comparable response labels. A literal translation of “checking a claim” may leave respondents thinking only about reading a message carefully. Ask people in each language what action they understood.
The TRAPD approach organises translation, review, adjudication, pretesting, and documentation. Team discussion and respondent testing help identify differences in meaning. Preserve a record of difficult terms and the reasons for the chosen wording.
Create conditions for a candid response
Explain who is conducting the study, what participation involves, how long it is expected to take, how information will be used, and how privacy will be handled. Estimate completion time through testing. Make voluntary participation practical, especially where a teacher or employer recruits respondents.
Plan any data-quality checks before fieldwork. Examine response patterns alongside timing, item content, routing, and other relevant evidence. Repeated identical ratings or midpoint responses may be meaningful; a pattern alone provides an incomplete basis for exclusion.
Pretest, revise, and test again
Pretesting examines whether the questionnaire performs as intended. Use complementary methods to investigate meaning, navigation, timing, recruitment, recording, and the resulting data.
Expert review
Check construct coverage, wording, options, mode, ethics, routing, and the analysis plan.
Cognitive interviews
Explore how people similar to the intended respondents understand and answer selected items.
Pilot administration
Try the instrument and procedures in conditions resembling the planned survey, including data export and cleaning.
Use probes that reveal the response process
Ask a pretest participant to explain their thinking while answering or immediately afterwards. Keep the approach consistent with the purpose of the test. Useful probes include:
- Comprehension: “What did ‘share’ include for you in this question?”
- Recall: “How did you decide which message was the most recent?”
- Judgement: “What made that action count as checking?”
- Response selection: “Was there an option that matched the answer you wanted to give?”
Record the difficulty and its likely source before revising. If “checking” means merely opening a message for some respondents, decide whether the intended construct requires a clearer definition, examples, or a different question. Then test the revision.
Issue: Participants included personal greetings when answering about “messages”.
Revision: Define a news message through its claim about a current event or public issue.
Retest question: Can participants explain which messages they would include and exclude?
Make the pilot inspect the whole workflow
Check completion and break-off points, item nonresponse, response distributions, write-ins, route errors, display problems, and exported variable labels. Test the shortest and longest routes. Enter cases deliberately to confirm that valid combinations survive and contradictory combinations receive appropriate handling.
Choose the pretest participants and the pilot size through the problems you need to investigate. A small pilot can expose usability or routing faults while providing too little evidence for stable factor analysis or population estimates. Document which questions each testing stage can answer.
The US Census Bureau’s questionnaire testing guidance and Cross-Cultural Survey Guidelines describe complementary evaluation methods. Finalise a dated version after revisions, with a record of what changed.
Examine validity and reliability
Measurement quality concerns the interpretation you want to make from an answer or score. Assemble evidence suited to that interpretation, population, language, and use.
Validity: support the meaning of the result
A report of checking before sharing may be influenced by memory, social approval, and different meanings of “checking”. Confidence ratings may reflect self-belief more closely than skill. Validation investigates such possibilities through theory and evidence.
- Content: Do the items represent the intended construct and its important dimensions? Review coverage with relevant experts and potential respondents.
- Response processes: Do people understand and answer the items in ways consistent with the intended meaning?
- Internal structure: For a multi-item score, do relationships among items support the proposed structure? Exploratory or confirmatory factor analysis can contribute appropriate evidence.
- Relationships with other variables: Do associations follow defensible expectations? Examine convergence with related measures, distinction from different constructs, or relationships with relevant external criteria.
Predictive evidence concerns a justified later outcome; concurrent criterion evidence concerns a relevant criterion measured at a similar time. Comparing with another questionnaire measuring a related concept may contribute convergent evidence. Explain why the comparator is useful.
The Standards for Educational and Psychological Testing frame validity around the evidence and theory supporting a specified interpretation and use. This keeps the claim tied to the score and its purpose.
Reliability: examine consistency under relevant conditions
A reliability study specifies what should remain consistent and which sources of measurement error matter. Choose the method through that purpose.
Test–retest
Examine stability over an appropriate interval when the underlying attribute is expected to remain stable. Account for memory and genuine change.
Internal consistency
Evaluate a justified multi-item score using an appropriate coefficient and model. Cronbach’s alpha and omega have assumptions that require attention.
Split-half
Examine consistency between parts of a scale, with an appropriate correction for the reduced length and attention to how the items are divided.
When adapting an existing measure, review its permissions, intended use, population, language, wording, and scoring instructions. Preserve comparability where needed and evaluate meaningful changes. Generalising survey results also depends on the population, selection, participation, and wider design.
A worked questionnaire: news sharing and checking
The sample below connects a defined behaviour with a reference period, filter route, response format, and coding plan. It shows how the parts of an instrument work together.
Intended respondents: Adult university students. Purpose: Describe recent news sharing, checking before the most recent share, checking actions, perceived confidence, and support for a university session. An optional open item explores barriers; programme level serves a planned comparison if the study requires it.
In the seven days ending yesterday, did you share at least one news message through a messaging app?
Here, a news message presents a claim about a current event or public issue. Sharing includes sending or forwarding a message, link, image, or clip to another person or group.
If unanswered: Go to Q4 and retain an item-missing status for Q1.
Think of the most recent news message you shared during those seven days. Before sharing it, did you try to find out whether its main claim was accurate?
If unanswered: Go to Q4 and retain an item-missing status for Q2. Test what respondents understand as an attempt to check.
What did you do to check that claim before sharing the message?
Then: Go to Q4. The two uncertainty/refusal choices are exclusive with every other option. Record an unanswered item separately.
How confident are you that you could check whether a news message’s main claim is accurate?
To what extent do you support or oppose your university offering a session on checking news claims?
What, if anything, makes checking a news claim difficult for you?
What is the level of your main university programme?
If enrolled in several programmes, identify the one you regard as your main programme. Test whether these categories fit the institutions included.
Build the codebook before fieldwork
Give variables stable names and store the full wording, response labels, eligible universe, route, and missing-data rules. The simplified plan below separates recorded values from response status.
On a narrow screen, scroll the table sideways.
| Variable | Substantive values | Status or processing rule |
|---|---|---|
shared7 | 0 = No; 1 = Yes. | Separate status for answered, cannot remember, declined, or item missing. |
checked_last | 0 = No; 1 = Yes. | Q1 must be Yes; otherwise record routed/not asked, with the Q1 reason retained. |
check_web, check_source, check_person, check_other | One binary variable per action: 1 = selected; 0 = not selected within a substantive Q3 answer. | For skipped, uncertain, declined, or unanswered Q3, retain the status and leave the action values missing. |
confidence and support | 1–5 in the displayed substantive order. | Store uncertainty, refusal, and item missing separately; keep the two constructs distinct. |
barrier_text and programme | Preserve open text; use labelled nominal codes for programme level. | Document text coding and write-ins. A blank open answer leaves the reason unknown. |
Write the denominator into the analysis plan
A result from Q2 concerns checking before the most recent share among eligible recent sharers. Report the number eligible, substantive responses, and uncertainty or missing responses. Q3 concerns actions reported by those who said they attempted checking. A multiple-response percentage may total more than 100% because one person can report several actions.
Preserve the questionnaire version, codebook, routing specification, and analysis decisions together. A later reader should be able to reconstruct how a recorded value arose.
Learn with the lecture and design workshop
Use the lecture as a guided review. Pause at each passage and make one concrete improvement to your own draft questionnaire.
Open Dr. USP’s lecture on YouTube. These approximate starting points follow the supplied transcript. The chapter adds worked examples and develops distinctions for practical use.
Try each task before opening the response
Activity 1 · Repair a double-barrelled item
Prompt: Revise “How useful and trustworthy is this podcast?”
Suggested response: Ask “How useful is this podcast for [specified purpose]?” and separately “How trustworthy do you consider the information in this podcast?” Supply labels matching each dimension and establish whether the respondent has enough experience to judge.
Activity 2 · Separate neutrality and uncertainty
Prompt: A respondent says, “I have never heard of the proposed campus channel.” Which response design would help?
Suggested response: Establish familiarity or provide an appropriately neutral description if the purpose permits an informed response. Keep a genuine neutral evaluation separate from insufficient information or “Don’t know”. The choice depends on whether you want an existing opinion or an evaluation after a description.
Activity 3 · Interpret a skipped question
Prompt: Q1 says the student shared no news in the reference period. What value belongs in Q2 about checking before the most recent share?
Suggested response: Record Q2 as not asked through the route and retain the Q1 answer. Entering “No checking” would create a behavioural answer for an occasion that did not occur.
Activity 4 · Evaluate a reliability claim
Prompt: A researcher calculates alpha for age, programme, news-sharing, confidence, and support, then calls the questionnaire validated. How would you improve the evaluation?
Suggested response: Begin with each intended interpretation. Evaluate the individual items and any justified multi-item scores using appropriate evidence. These diverse variables serve different purposes; a single internal-consistency coefficient across them does not evaluate the instrument’s quality.
Your independent design workshop
Develop an 8–12-item draft about a low-stakes communication topic, such as podcast choice, campus news use, or study-related messaging.
- State the population, objective, concepts, and planned descriptive results.
- Create an objective–item–variable map and justify each item.
- Include a clear reference period, one appropriate filter, and a planned open response.
- Specify response categories, routing, coding, and treatment of missing answers.
- Arrange appropriate review and voluntary pretesting with people similar to the intended respondents, following the relevant institutional requirements.
- Record comprehension or usability problems, revise the instrument, and retest key changes.
- Write a brief account of what the instrument can measure and what evidence remains to be collected.
Completion check: Every item has a purpose; every eligible response has a place; every route reaches a clear next step; every recorded value has a documented meaning.
A final instrument audit
- Are the population, reference periods, definitions, and intended interpretations clear?
- Does each item ask one manageable question with a suitable response format?
- Are options sufficiently complete, distinct, balanced where appropriate, and consistently labelled?
- Can respondents navigate the shortest and longest routes on the intended device or mode?
- Have language, consent, privacy, burden, and accessibility been checked?
- Have pretesting findings, revisions, coding rules, and the final version been documented?
Glossary and further reading
These terms help you explain a questionnaire’s design. The readings develop the reasoning behind wording, response formats, evaluation, and score interpretation.
- Construct
- The concept the study seeks to investigate, such as perceived confidence or trust.
- Operationalisation
- Specifying how a concept will be represented through indicators and recorded observations.
- Item
- A question, statement, or task designed to obtain a response.
- Double-barrelled item
- An item asking about two or more things that may require different answers.
- Acquiescence
- A tendency to agree with statements regardless of their particular content.
- Social desirability
- The influence of perceived social approval on how someone answers.
- Filter and routing
- A question and instructions that determine which later items apply to the respondent.
- Cognitive pretesting
- Investigating how respondents understand, recall, judge, and select answers to draft items.
- Split-ballot test
- Random assignment of questionnaire versions to examine effects such as wording or order.
- Validity evidence
- Evidence supporting a specified interpretation and use of an answer or score.
- Reliability
- The consistency or precision of a score under specified conditions and sources of error.
- Codebook
- A record of variables, wording, labels, eligible respondents, codes, and processing rules.
Sources and guided reading
- Dr. USP. Questionnaire Design [Video]. YouTube.The chapter’s accompanying lecture and transcript introduce the principal design decisions.
- American Association for Public Opinion Research. Best practices for survey research.Place questionnaire development, pretesting, fieldwork, and reporting within the wider survey process.
- Pew Research Center. Writing survey questions. See also Methods 101: Survey question wording [Video].Read and watch for examples of wording, response categories, and question context.
- Australian Bureau of Statistics. (2023). General forms design principles: Question structure.Practical guidance on response tasks, structure, and the demands placed on respondents.
- Centers for Disease Control and Prevention, National Center for Health Statistics. (2024). Cognitive interviewing.Use the response-process probes to plan a cognitive pretest.
- US Census Bureau. Appendix A2: Questionnaire testing and evaluation methods for censuses and surveys.Compare cognitive interviews, usability testing, debriefing, and other evaluation methods.
- Survey Research Center, Institute for Social Research, University of Michigan. Cognitive interviewing. Cross-Cultural Survey Guidelines.Develop the pretest design, analysis of problems, and revision process.
- European Values Study. The TRAPD method for survey translation.Translation, review, adjudication, pretesting, and documentation through a team approach.
- American Educational Research Association, American Psychological Association, & National Council on Measurement in Education. (2014). Standards for educational and psychological testing. American Educational Research Association.Consult for validity evidence, reliability, fairness, and the interpretation and use of scores.
- Boateng, G. O., Neilands, T. B., Frongillo, E. A., Melgar-Quiñonez, H. R., & Young, S. L. (2018). Best practices for developing and validating scales for health, social, and behavioral research: A primer. Frontiers in Public Health, 6, Article 149. https://doi.org/10.3389/fpubh.2018.00149An extended guide for studies that develop or adapt a multi-item scale.
- Sijtsma, K. (2009). On the use, the misuse, and the very limited usefulness of Cronbach’s alpha. Psychometrika, 74(1), 107–120. https://doi.org/10.1007/s11336-008-9101-0Read when interpreting internal-consistency claims and the structure of a proposed score.
Carry the draft into the next stage
Take your revised questionnaire and its codebook back to the study plan. Decide who needs to answer, how they will be reached, and what the resulting evidence can support. The Research Methodology section helps connect these decisions. If pretesting reveals meanings you had overlooked, use Ethnography to deepen your attention to context, or Content Analysis to develop a systematic approach to the open responses.