Measurement, Scales, and Questionnaire Design

Introduction

Market research converts managerial questions into evidence about customers, markets, products, and behaviour. That conversion is not automatic. Concepts such as satisfaction, trust, perceived value, adoption intention, brand attachment, and usability are abstract constructs. They cannot usually be observed directly. Researchers must therefore operationalize them: define what a construct means, identify observable indicators, select an appropriate measurement scale, and collect responses through a carefully designed instrument.

Poor measurement can produce precise-looking but misleading results. A biased question may overstate demand; an inappropriate scale may make two groups appear different when they are not; and a reliable questionnaire may consistently measure the wrong construct. Product managers consequently need more than the ability to write survey questions. They must understand the logic of measurement, the properties of scales, questionnaire cognition, reliability, validity, and ethical data collection.

Learning Objectives

By the end of this chapter, the reader should be able to:

1. Explain constructs, variables, indicators, operational definitions, reliability, and validity.
2. Distinguish nominal, ordinal, interval, and ratio scales and select appropriate analytical procedures.
3. Design Likert, semantic differential, behavioural, ranking, and constant-sum measures.
4. Write clear, neutral, balanced, and respondent-centred questionnaire items.
5. Identify and reduce coverage, sampling, non-response, acquiescence, social-desirability, recall, and order biases.
6. Pretest a questionnaire and assess its measurement quality.
7. Apply measurement principles to product and market research decisions.

1. From Managerial Concepts to Measurable Variables

A construct is an abstract concept that a researcher wishes to study, such as customer loyalty or perceived ease of use. A variable is a measurable characteristic that can take different values across people, products, occasions, or markets. An indicator is an observable item used to represent a construct. An operational definition specifies exactly how a construct will be measured in a study.

For example, a product team may define perceived onboarding simplicity as the customer’s evaluation of how easy it is to complete the first-use process without assistance. Possible indicators include:

• “The onboarding steps were easy to understand.”
• “I could complete setup without help.”
• “The amount of information required during setup was reasonable.”

The construct is not identical to any one item. It is inferred from a pattern of responses across suitable indicators. This distinction matters because a single question may omit important dimensions or be overly influenced by wording.

1.1 Conceptual and operational definitions

A conceptual definition explains what a construct means theoretically. An operational definition explains how it will be observed. For example:

Operationalization should be driven by the decision the research must support. If the decision concerns feature prioritization, broad satisfaction alone is insufficient; the questionnaire should measure feature importance, current performance, unmet needs, and behavioural consequences.

1.2 Reflective and formative measurement

In reflective measurement, the construct is assumed to cause the observed indicators. If trust declines, several trust items may decline together. Indicators are manifestations of the underlying construct and are often expected to correlate.

In formative measurement, indicators combine to form the construct. For example, service burden might be formed by waiting time, number of contacts, documentation effort, and resolution effort. These components need not correlate strongly. Removing one may change the meaning of the construct.

Confusing these models creates design errors. A formative index should not automatically be judged by the same internal-consistency logic used for a reflective scale. The researcher must first specify the construct model, based on theory and the intended decision.

2. Levels and Types of Measurement Scales

The level of measurement determines what comparisons are meaningful and which statistical operations are defensible. The four classical levels are nominal, ordinal, interval, and ratio. In applied research, the wording and use of a scale can be more important than its label. For example, a five-point agreement item is formally ordinal, although a multi-item Likert scale is often treated as approximately interval for some analyses when assumptions are reasonable.

2.1 Nominal scales

A nominal scale classifies observations into mutually exclusive categories without a meaningful order. Examples include subscription plan, operating system, region, or whether a customer has used a feature.

Appropriate summaries include frequencies, percentages, mode, cross-tabulations, and tests designed for categorical data. Assigning codes such as 1 = Android and 2 = iOS does not make the variable quantitative; the numbers are labels, not amounts.

Categories should be exhaustive for the target population or provide a legitimate “Other” or “Not applicable” option. They should also be mutually exclusive unless multiple responses are explicitly permitted.

2.2 Ordinal scales

An ordinal scale provides rank order, but the distance between adjacent categories is not known to be equal. Examples include:

• beginner, intermediate, advanced;
• low, medium, high;
• first, second, and third preference;
• strongly disagree to strongly agree.

The median and percentiles are generally meaningful. The mean can sometimes be used as a pragmatic summary of a multi-item scale, but this is an analytical convention rather than proof that the response categories have equal intervals.

2.3 Interval scales

An interval scale has ordered values with equal intervals, but zero does not represent the complete absence of the measured attribute. Temperature in Celsius is a classic example. In marketing research, attitude scales are often treated as approximately interval when several well-designed items are combined and the distribution is suitable. This assumption should be stated rather than hidden.

Addition and subtraction are meaningful on an interval scale; ratio statements such as “twice as satisfied” are not necessarily meaningful.

2.4 Ratio scales

A ratio scale has equal intervals and a meaningful, non-arbitrary zero. Examples include number of purchases, time spent using an application, revenue, quantity purchased, and age in years. Ratio comparisons are meaningful: ten purchases is twice five purchases, subject to accurate measurement and a valid zero.

A common error is to treat all numerical-looking responses as ratio data. A coded satisfaction response of 1 to 5 is not a ratio scale merely because it contains numbers.

3. Common Measurement Formats

3.1 Likert-type items and Likert scales

A Likert-type item presents a statement and asks respondents to indicate their degree of agreement, commonly on five or seven ordered categories. For example:

“The product makes it easy for me to monitor my monthly spending.”

Response options might range from “Strongly disagree” to “Strongly agree.” A Likert scale usually refers to a composite score formed from several related Likert-type items. Items should represent the same construct, use consistent direction, and be reviewed for redundancy.

Five-point scales are easy to use and reduce cognitive burden. Seven-point scales can provide greater discrimination when respondents can meaningfully distinguish the categories. More categories do not automatically create better measurement. Labels should be clear, and the scale should include a neutral or midpoint only when genuine neutrality is plausible.

Agreement is not always the best response frame. For factual or evaluative questions, direct alternatives may reduce acquiescence. Instead of asking whether a respondent agrees that delivery was fast, ask, “How would you rate the delivery speed?”

3.2 Semantic differential scales

A semantic differential asks respondents to position an object between bipolar adjectives, such as:

The adjectives must be genuinely opposite, relevant to the object, and understandable to the target population. Randomly reversing scale direction across items can increase errors; if reversed direction is used to control response patterns, it should be applied sparingly and checked carefully.

3.3 Rating, ranking, and constant-sum measures

A rating asks respondents to evaluate each object independently, such as rating ten features from 1 to 5. A ranking asks them to order alternatives. Ranking reveals relative preference but becomes burdensome with many alternatives and does not show how far apart preferences are.

A constant-sum question allocates a fixed total, often 100 points, across attributes. For example, respondents might allocate 100 points among speed, price, reliability, and design according to importance. The method captures trade-offs but requires arithmetic and can be difficult on mobile devices. The instructions must state whether zero is allowed and whether the total must equal 100.

3.4 Behavioural and factual measures

Behavioural measures concern actions rather than attitudes: number of purchases in the last 90 days, date of last use, renewal status, feature activation, support contacts, or channel used. When available and ethically permissible, observed or verified behavioural data may be more accurate than self-reported intentions.

Self-reports remain valuable because many behaviours are unavailable to the researcher and motives or future plans cannot be directly observed. They are affected by memory, interpretation, telescoping, and social desirability. Improve accuracy by specifying a reference period, using salient events, offering ranges where exact recall is unrealistic, and providing “Don’t know” or “Prefer not to say” where appropriate.

3.5 Intention and probability measures

Purchase intention is not purchase behaviour. It can be useful as an early signal, but it is affected by price, availability, competing priorities, and the gap between stated and actual behaviour. A high-quality intention item defines the object, context, and time frame: “How likely are you to subscribe to Plan X at ₹499 per month within the next three months, if it remains available?”

4. Designing a Questionnaire

Questionnaire design is a form of measurement engineering. The instrument should make the desired interpretation easy and alternative interpretations unlikely.

4.1 Begin with the information plan

Before drafting items, specify:

1. the decision to be made;
2. the target population and eligibility criteria;
3. the constructs and behaviours required;
4. the analysis planned for each variable;
5. the minimum data needed to answer the decision question;
6. privacy, consent, and data-retention requirements.

A measurement blueprint can link research questions to constructs, indicators, item wording, response format, and analysis. This prevents the common failure of collecting interesting but unusable data.

4.2 Recommended questionnaire flow

A practical sequence is:

1. introduction, purpose, estimated time, consent, and confidentiality statement;
2. screening and eligibility questions;
3. recent behaviour or experience questions;
4. core evaluations and construct measures;
5. preferences, trade-offs, or reactions to concepts;
6. sensitive questions, where necessary;
7. demographic or classification questions;
8. open-ended feedback and closing message.

The order is not universal. Screening must occur early, while sensitive questions may sometimes be placed earlier if essential for eligibility. Randomization or rotation can reduce primacy and order effects in lists, but fixed logical order is preferable when respondents need a natural sequence.

4.3 Writing good items

A strong item is clear, specific, relevant, answerable, and neutral. Apply the following principles:

• Ask one thing at a time. “The app is fast and easy to use” is double-barrelled because speed and ease may differ.
• Avoid undefined jargon, technical terms, and acronyms.
• Specify the object, context, and time period.
• Prefer concrete language to vague terms such as “often,” “regularly,” or “good.”
• Avoid leading wording, emotional language, and assumptions that respondents have knowledge they may not possess.
• Use balanced response options and include “Not applicable” when a legitimate non-substantive response exists.
• Make all response options parallel in grammar and mutually exclusive.
• Avoid unnecessary negatives and especially double negatives.
• Do not ask respondents to calculate difficult quantities if records or ranges can be used.

Poor: “How satisfied are you with our affordable and reliable service?”

Better: “Overall, how satisfied are you with the service you received during your most recent interaction?”

The improved question avoids presupposing affordability and reliability and anchors the evaluation to a defined experience.

4.4 Open-ended questions

Open-ended questions permit respondents to express unanticipated issues and provide diagnostic detail. They are useful after a rating, for example: “What is the main reason for your rating?” Too many open-ended questions increase burden and coding cost. Use them selectively, provide enough space, and define a coding framework before analysis.

4.5 Sensitive and demographic questions

Collect sensitive information only when it serves a clear research purpose. Explain why it is requested, make response optional where possible, and use categories appropriate to the population. Income, health, identity, and location data require particular care. Avoid “prefer not to say” being treated as a substantive category during analysis.

5. Sources of Survey Error and Bias

Total survey error includes several interacting components:

• Coverage error: members of the target population lack a reasonable chance of inclusion.
• Sampling error: a sample differs from the population because only some members are observed.
• Non-response error: people who do not respond differ systematically from respondents.
• Measurement error: answers differ from the intended construct because of wording, scale, context, respondent interpretation, or interviewer effects.
• Processing error: coding, cleaning, weighting, or data-entry mistakes alter responses.

Questionnaire design primarily addresses measurement and respondent burden, but researchers must examine the whole evidence process.

Important response biases include:

• Acquiescence: tendency to agree with statements.
• Extreme response style: preference for scale endpoints.
• Central tendency: avoidance of endpoints.
• Social desirability: presenting oneself favourably.
• Recall bias: inaccurate memory of past events.
• Telescoping: placing an event outside its actual time period.
• Satisficing: choosing a convenient answer rather than carefully evaluating the question.
• Priming and order effects: earlier questions influence later responses.

Mitigation includes neutral wording, clear reference periods, randomized lists where appropriate, balanced scales, anonymity assurances, short instruments, attention to device usability, and analysis of completion patterns. No technique eliminates bias completely; transparency about residual limitations is essential.

6. Measurement Quality: Reliability and Validity

6.1 Reliability

Reliability concerns consistency of measurement. A reliable instrument produces similar results under conditions in which the construct has not changed. It does not guarantee validity.

Common forms include:

• Test–retest reliability: stability across time.
• Inter-rater reliability: agreement among observers or coders.
• Internal consistency: coherence among items intended to measure the same construct.

Cronbach’s alpha is a commonly reported estimate of internal consistency. Under the assumptions of essentially tau-equivalent reflective items, locally uncorrelated errors, and a suitable unidimensional construct, it may be written as:

Here, α is alpha, k is the number of items, σᵢ² is the variance of item i, and σₜ² is the variance of the total score formed by summing the items. Alpha is dimensionless. It tends to increase with the number of items and their intercorrelation, so a high value can reflect redundant items. It should not be used as proof of unidimensionality or validity. For multidimensional or formative measures, alpha may be inappropriate; factor analysis, theory, and alternative reliability estimates should be considered.

6.2 Validity

Validity concerns the degree to which evidence supports the intended interpretation and use of scores. It is not simply a permanent property of a questionnaire; it relates to a particular construct, population, context, and decision.

• Content validity: indicators adequately cover the construct domain. Expert review and a construct specification are useful.
• Face validity: items appear appropriate to respondents or experts. It is helpful for acceptance but is weak evidence by itself.
• Construct validity: scores behave as theory predicts. Convergent evidence occurs when related measures correlate; discriminant evidence occurs when distinct constructs are not indistinguishable.
• Criterion-related validity: scores relate to an external criterion, such as future retention or verified purchase, when that criterion is appropriate and measured adequately.
• Known-groups validity: a measure differentiates groups expected to differ.

Reliability is often necessary for validity, but it is not sufficient. A broken clock can be consistent; similarly, a leading item can produce highly consistent responses while measuring agreement with the wording rather than the intended construct.

6.3 Dimensionality and construct assessment

Before combining items, examine whether they represent one dimension or several. Exploratory factor analysis can help identify patterns in an early-stage scale, while confirmatory factor analysis tests a prespecified structure. These methods require adequate sample size, appropriate correlation assumptions, and substantive interpretation; factor extraction is not a substitute for theory.

A practical validation sequence is:

1. define the construct and its boundaries;
2. generate items from theory, qualitative research, and expert knowledge;
3. review content and cognitive interpretation;
4. pilot the instrument;
5. inspect item distributions, missingness, correlations, and dimensionality;
6. estimate reliability using a method appropriate to the measurement model;
7. test relationships with related outcomes;
8. revise cautiously and validate on new data where possible.

7. A Worked Measurement Example

A product team wants to assess “perceived control over personal finances” in a budgeting application. It defines the construct as the user’s perceived ability to understand, monitor, and influence their financial decisions through the application.

The team drafts six reflective items, each rated from 1 = strongly disagree to 5 = strongly agree. After cognitive interviews, one item is removed because users interpret “control” as account security rather than financial decision-making. The remaining items are checked for coverage of understanding, monitoring, and action.

Suppose a pilot produces an item-level variance sum of 4.80, total-score variance of 12.00, and k = 5. Applying the alpha formula:

The estimate is 0.75. This is not a universal pass/fail result. The team should inspect the construct’s dimensionality, item wording, sample, and intended use. If one item concerns security and the others concern budgeting, deleting it may improve consistency but could also reveal a construct-definition problem. The team should not delete items solely to maximize alpha. It must determine whether the revised five-item measure adequately represents the construct and predicts relevant outcomes such as budgeting-feature use or retention.

8. Applied Case: Redesigning a Subscription-Upgrade Survey

The following is an illustrative case.

Background

A streaming-product team plans to introduce a premium tier. Its initial survey reports that 68% of users are “likely to upgrade.” However, pilot sales are weak. The product manager suspects that the questionnaire encouraged optimistic answers.

Problem

The survey used the item: “Given the amazing value and exclusive benefits of our premium plan, how likely are you to upgrade soon?” It did not state the price, define “soon,” distinguish current subscribers from non-subscribers, or provide a “not familiar enough to answer” option. The item also measured intention without measuring current behaviour, perceived need, or trade-offs.

Analysis

The wording is leading and contains a positive presupposition. “Soon” has different meanings across respondents. The survey also suffers from hypothetical bias: respondents are not making a real payment decision. In addition, respondents who had not seen the plan were forced to evaluate it.

Approach and solution

The team creates a measurement blueprint with four components:

1. Eligibility and familiarity: current plan, tenure, and exposure to the concept.
2. Behavioural context: recent viewing frequency and use of features relevant to premium benefits.
3. Evaluation: perceived usefulness of each benefit, perceived price fairness, and trust.
4. Defined intention: “If the premium plan costs ₹499 per month and is available next month, how likely are you to subscribe within the following three months?”

The team uses a balanced likelihood scale, includes “Not familiar enough to answer,” randomizes the benefit list, and follows the intention item with an open-ended reason. It pretests the survey through cognitive interviews and a small pilot. It then compares stated intention with a controlled offer and tracks actual conversion, while clearly informing users about data use and avoiding deceptive pricing.

Outcome

The revised survey produces a lower but more differentiated intention estimate. High intention is concentrated among frequent viewers who value specific benefits and accept the stated price. The team uses these results to target a limited experiment rather than forecasting total adoption directly from the survey percentage.

Learning points

The case demonstrates that a lower estimate can be more useful than a flattering estimate. Good measurement defines the object and time frame, distinguishes attitudes from behaviour, uses eligibility filters, tests interpretation, and validates intentions against observed outcomes.

9. Ethics, Governance, Technology, and Contemporary Practice

Researchers must obtain informed consent appropriate to the study, explain purpose and expected burden, protect confidentiality, minimize collection of personal data, and specify retention and access controls. Participation should be voluntary, and refusal should not create unfair disadvantage. Sensitive segmentation should not be used to exploit vulnerability or enable discriminatory product decisions.

Digital survey platforms create additional risks: device fingerprinting, hidden metadata, panel fraud, bot responses, location inference, and unauthorized linkage to customer records. Governance should define data minimization, role-based access, encryption, quality checks, and deletion procedures. Regulatory requirements vary by jurisdiction, so organizations should consult applicable privacy and research-ethics rules rather than assuming that consent alone resolves all obligations.

AI can assist with item generation, translation, coding of open-ended responses, and detection of inconsistent patterns. It should not replace construct definition, expert review, cognitive testing, or accountability for decisions. Automated translation may alter cultural meaning; automated sentiment or demographic inference may reproduce bias. Maintain version control, document prompts and transformations, audit outputs, and obtain human review before deploying AI-assisted measurement.

Contemporary research increasingly combines survey responses with behavioural telemetry, experiments, interviews, and administrative data. Triangulation can strengthen decision quality, but linked datasets may increase privacy risk and introduce inconsistent definitions. The product manager should ask whether each data source measures the same construct, at the same time, for the same population, and with comparable error.

10. Best-Practice Checklist

Before fielding a questionnaire, confirm that:

• every item maps to a research decision or analytic need;
• constructs have conceptual and operational definitions;
• the population, eligibility, and reference periods are explicit;
• scale types and response options fit the information required;
• wording is neutral, single-purpose, specific, and understandable;
• “Not applicable,” “Don’t know,” or refusal options are used when legitimate;
• question order and randomization have been considered;
• the instrument works on the intended devices and languages;
• sensitive data are necessary and governed;
• cognitive interviews and a pilot have been completed;
• reliability and validity evidence will be assessed appropriately;
• data-quality rules are defined before inspecting results;
• limitations and uncertainty will be reported honestly.

Chapter Summary

Measurement translates abstract market and consumer constructs into observable evidence. Sound research begins with conceptual and operational definitions and distinguishes reflective from formative measurement. Nominal scales classify, ordinal scales rank, interval scales assume equal distances without a meaningful zero, and ratio scales add a meaningful zero and support ratio comparisons. Likert, semantic differential, ranking, constant-sum, behavioural, and intention measures each answer different research needs.

Effective questionnaires use a decision-focused blueprint, logical flow, neutral and specific wording, suitable response options, and careful treatment of sensitive data. Researchers must anticipate coverage, non-response, recall, social-desirability, acquiescence, order, and processing errors. Reliability addresses consistency; validity concerns whether the intended interpretation is supported. Cronbach’s alpha is useful only under appropriate assumptions and cannot substitute for theory or validity evidence. Pretesting, triangulation, behavioural validation, ethical governance, and transparent reporting are essential to professional product research.

Key Terms

Construct: An abstract concept intended for study.

Variable: A measurable characteristic that can take different values.

Indicator: An observable measure representing a construct.

Operational definition: A specification of how a construct will be measured.

Reflective measurement: A model in which a construct is expected to cause its indicators.

Formative measurement: A model in which indicators combine to define a construct.

Nominal scale: A categorical scale with no inherent order.

Ordinal scale: A scale with order but unknown or unequal intervals.

Interval scale: A scale with equal intervals but no meaningful zero.

Ratio scale: A scale with equal intervals and a meaningful zero.

Likert-type item: A statement rated on ordered agreement categories.

Semantic differential: A bipolar adjective rating scale.

Reliability: Consistency or stability of measurement.

Validity: Evidence supporting the intended interpretation and use of scores.

Measurement error: Difference between the observed response and the intended construct measurement.

Acquiescence bias: A tendency to agree with statements.

Recall bias: Systematic inaccuracy in remembering past events.

Cognitive pretest: An interview or test used to examine how respondents interpret items.

Questionnaire burden: The effort, time, and fatigue imposed on respondents.

Triangulation: Combining multiple methods or data sources to improve understanding.

Review Questions

Short Answer Questions

1. Distinguish a construct, variable, indicator, and operational definition.
2. Why is a numerical code for a nominal category not a quantitative measurement?
3. State two differences between a Likert-type item and a semantic differential scale.
4. Why does reliability not guarantee validity?
5. Identify two methods for reducing recall bias in a questionnaire.

Long Answer Questions

1. Explain nominal, ordinal, interval, and ratio measurement scales, using product research examples and appropriate analytical implications.
2. Discuss the principles of writing unbiased questionnaire items, including examples of double-barrelled, leading, vague, and improved questions.
3. Explain reflective and formative measurement and show why the distinction affects reliability assessment.
4. Describe a complete process for developing and validating a multi-item consumer-attitude scale.
5. Discuss the ethical and governance issues involved in combining survey data with behavioural telemetry and AI-assisted analysis.

Analytical / Application Questions

1. A food-delivery app asks, “How satisfied are you with our fast and affordable delivery?” Diagnose the measurement problems and rewrite the item and response format.
2. A product team obtains high Cronbach’s alpha for a six-item scale, but the items mainly repeat the same wording and do not predict renewal. Analyse what this implies about reliability, validity, and next steps.
3. Design a short questionnaire blueprint to evaluate willingness to adopt a new paid productivity feature. Specify the target construct, at least three indicators, scale formats, behavioural measures, screening questions, and one likely source of bias.

Multiple Choice Questions with Answers

Question 1

Which measurement scale classifies respondents without implying order or distance?

A. Nominal

B. Ordinal

C. Interval

D. Ratio

Correct Answer: A. Nominal

Question 2

What is the main distinction between a Likert-type item and a Likert scale?

A. A Likert item is always nominal, whereas a Likert scale is ratio

B. A Likert item is one statement; a Likert scale commonly combines several related items

C. A Likert item measures behaviour, whereas a Likert scale measures only demographics

D. A Likert item has no response options, whereas a Likert scale does

Correct Answer: B. A Likert item is one statement; a Likert scale commonly combines several related items

Question 3

Which is the clearest example of a double-barrelled question?

A. How many orders did you place last month?

B. Which plan do you currently use?

C. How satisfied are you with the app’s speed and ease of use?

D. How likely are you to renew within three months?

Correct Answer: C. How satisfied are you with the app’s speed and ease of use?

Question 4

In reflective measurement, indicators are generally understood as:

A. Components that collectively create the construct

B. Random demographic classifications

C. Manifestations caused by the underlying construct

D. External criteria that replace the construct

Correct Answer: C. Manifestations caused by the underlying construct

Question 5

In the alpha example, with k = 5, sum of item variances = 4.80, and total-score variance = 12.00, what is α?

A. 0.40

B. 0.60

C. 0.75

D. 1.25

Correct Answer: C. 0.75

Question 6

Which statement about Cronbach’s alpha is most accurate?

A. A high alpha proves that a scale is valid

B. Alpha is appropriate without qualification for every formative index

C. Alpha can increase through item redundancy and does not prove unidimensionality

D. Alpha measures sampling error only

Correct Answer: C. Alpha can increase through item redundancy and does not prove unidimensionality

Question 7

Which practice best reduces ambiguity in a behavioural recall question?

A. Use the word “regularly” without defining it

B. Ask about a clearly specified reference period

C. Require an exact answer for events from many years ago

D. Remove all “don’t know” options

Correct Answer: B. Ask about a clearly specified reference period

Question 8

What is a meaningful zero characteristic of which scale?

A. Nominal

B. Ordinal

C. Interval

D. Ratio

Correct Answer: D. Ratio

Question 9

Which is a validity issue rather than merely a reliability issue?

A. Whether scores are consistent across repeated administrations

B. Whether coders agree when classifying responses

C. Whether the measure captures the intended construct and predicts relevant outcomes

D. Whether items have similar variances

Correct Answer: C. Whether the measure captures the intended construct and predicts relevant outcomes

Question 10

A survey asks about upgrading “soon” without stating price or time period. What is the principal design weakness?

A. The question defines the decision context too precisely

B. The question leaves the object and reference period ambiguous

C. The question uses a ratio scale

D. The question is necessarily a behavioural measure

Correct Answer: B. The question leaves the object and reference period ambiguous