MKT5612 Chap.3 Measurement Scales and Construct Quality
Measurement Scales and Construct Quality
Measurement assigns observable responses to attributes according to rules. Nominal measures distinguish categories, ordinal measures add rank, interval measures add interpretable equal differences, and ratio measures also have a meaningful zero.
The scale level governs which comparisons and summaries are defensible; a numerical code does not automatically create numerical quantity.
Construct quality is a separate question. Reliability concerns consistency, while validity concerns whether an instrument captures the intended construct. Random error makes measurements unstable; systematic error can produce consistent but misleading scores.
Good operationalisation therefore begins with a construct definition, maps dimensions to items, chooses a suitable response scale, pretests interpretation and gathers evidence for both reliability and validity before treating a score as decision-ready.
What this chapter covers
- 01
Constructs, variables and operational definitions
- 02
Nominal and ordinal measurement
- 03
Interval and ratio measurement
- 04
Permitted comparisons and summaries
- 05
Random and systematic measurement error
- 06
Reliability, validity and pretesting
Worked example · free
Audit four customer measures
- 3Classify tier and format with a reason.
- 3Classify spend and the rating with a reason.
- 4State valid summaries and one caution about coding.
Key terms
- Nominal Scale
- A measurement level that distinguishes categories without imposing an order among them.
- Ordinal Scale
- A measurement level that preserves rank but does not establish equal distances between positions.
- Interval Scale
- A measurement level with ordered values and meaningful equal differences, but without a true zero.
- Ratio Scale
- A measurement level with equal intervals and a meaningful zero, permitting ratio comparisons.
- Reliability
- The consistency or stability of a measurement across items, occasions or raters under comparable conditions.
- Validity
- The extent to which evidence supports interpreting a measure as the construct it is intended to represent.
- Systematic Error
- A directional distortion that can make repeated measurements consistently differ from the intended value.
Measurement Scales and Construct Quality FAQ
Does a numerical response automatically have an interval scale?
No. Numbers may be category codes or ordered labels. Interval status requires the differences between adjacent values to represent equal empirical differences, which cannot be established merely by printing digits beside response options.
Can a measure be reliable but invalid?
Yes. An item can produce highly consistent responses while repeatedly capturing the wrong construct or containing directional bias. Reliability supports precision, but it does not by itself establish that the intended attribute was measured.
Why define a construct before writing items?
A definition sets the domain the measure must represent and prevents attractive but irrelevant questions from entering the instrument. It also provides a basis for evaluating content coverage, interpreting dimensions and explaining validity evidence.
What happens when a ratio variable is grouped into bands?
The grouped version usually preserves order but discards exact distances and ratios between observations. Analysis must follow the information retained in the bands, and averaging the category codes does not reconstruct the original continuous values that were removed.
Exam move
Build a permission ladder rather than memorising labels alone. For each scale, state what relation is preserved, what summary becomes possible and what tempting operation remains unjustified. Then separate scale-level questions from construct-quality questions: a ratio measure can still be biased, and a reliable scale can still miss its target.
Practise diagnosing cases in four moves: define the construct, identify the observable indicator, classify its empirical structure and name the dominant error risk. In written answers, a one-line justification is essential because the same digits can encode very different measurement levels.
Assemble a table of ordinary marketing variables and classify both the original form and a recoded form, such as exact spending converted into bands. State what information was lost and which analysis permission disappeared. Next, take one broad construct such as trust and write a definition, two intended dimensions and one item per dimension.
Challenge each item for content omission, systematic wording bias and random instability. Compare the exercise with a behavioural proxy and explain why convenient observability does not guarantee construct validity. Finish every answer by distinguishing what the numerical structure permits from what the construct evidence supports; these are separate warrants.
Working through Measurement Scales and Construct Quality in MKT5612? Sia is AskSia’s AI Marketing tutor — ask any MKT5612 Measurement Scales and Construct Quality question and get a clear, step-by-step explanation grounded in how MKT5612 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.