Operational definitions: turning research concepts into measurable variables

You may have a clear research question and still be unsure what to put in your questionnaire. “Support,” “performance,” and “engagement” sound like recognisable ideas. They do not yet tell another researcher what you will observe or how you will interpret the responses.

An operational definition makes that next step explicit. It describes how a variable will be measured or recorded in your study. A useful definition lets someone trace the connection between the concept you discuss and the evidence you will actually collect.

That connection deserves attention before data collection. If the evidence captures a different idea from the one in your question, a more complicated analysis will not repair the mismatch.

Once the data file exists, a data dictionary can document how its fields store the evidence, including labels, coding and missing values. It complements an operational definition; it does not replace the study’s measurement decision.

Separate the concept from the procedure used to measure it

A conceptual definition explains what you mean by an idea. An operational definition specifies how you will represent it in the study. The BCcampus chapter on psychological measurement makes this distinction and explains why a construct can have more than one operational definition.

For example, “access to workplace support” could concern the availability of help, employees’ perceptions of that availability, or their actual use of support. Those are related questions, but they need different evidence.

An organisational policy might establish that a support service exists. It cannot, on its own, establish whether employees know about it or feel able to use it. A questionnaire about perceived access cannot establish how often the service was used unless it also asks about use, with appropriate limits on recall.

Start by completing two sentences in your own notes:

In this study, the concept means…

Evidence of that concept will be obtained by…

Then check whether the second sentence actually follows from the first. If you need several sentences to explain the procedure, use them. Precision matters more than fitting the definition into a neat phrase.

Distinguish constructs, indicators, items, and scores

These terms describe different parts of the measurement process:

On a narrow screen, scroll the table sideways to read every column.

Term	What it identifies	Illustrative use
Construct	The idea the researcher wants to investigate	Perceived access to support
Indicator	An observable response or record used to represent some aspect of that idea	A response about whether help is available when needed
Questionnaire item	The particular prompt presented to a respondent	A question with a defined timeframe and answer options
Score	A result produced by applying a stated rule to responses	A subscale score calculated according to the instrument’s instructions

The examples are conceptual illustrations, not a proposed validated instrument. An item does not become a sound measure simply because it uses the name of the construct.

Also avoid assuming that every variable requires a multi-item scale. Some questions concern counts, dates, categories, or recorded durations. Others concern concepts that need a carefully developed instrument. Let the research question and measurement evidence determine the choice.

Build a measurement table before writing the methods paragraph

The following example is synthetic. The organisation, proposed records, and measurement choices are invented to demonstrate planning; no data have been collected.

Suppose a researcher wants to study whether employees’ access to task guidance is associated with how promptly requests are completed in an administrative service team. Their initial plan is to ask staff whether they perform well and label the answer “service performance.”

That would measure a self-assessment. It would not directly measure completion time. A more transparent planning table separates the two:

On a narrow screen, scroll the table sideways to read every column.

Variable	Proposed observation or measure	Scope and interpretation
Perceived access to task guidance	An established questionnaire measure, selected only after checking its definition, items, population evidence, and permitted use	Employees’ perceptions of access; not an audit of whether guidance is objectively available
Recorded completion duration	Elapsed time from a request’s logged receipt to its recorded completion, using specified timestamps	Duration under the record system’s rules; not a complete measure of service quality
Employee self-rated task performance	A separate appropriate measure if self-assessment is part of the question	Employees’ views of their performance; must not be relabelled as recorded performance

The first row is deliberately not a finished operational definition. Until an actual instrument and scoring procedure are selected, it is a measurement requirement. Writing “an established scale will be used” does not complete that work.

The second row also raises decisions. Does the system count evenings and weekends? What happens when a request is paused because information is missing? Are reopened requests treated as new cases? A recorded value can be precise while answering the wrong version of the question.

A fuller operational definition for that outcome could read:

Recorded completion duration will be the elapsed calendar time between the first receipt timestamp and final completion timestamp for eligible requests received during the observation window. Requests without a final completion timestamp will be identified separately rather than assigned a duration of zero. The measure describes recorded processing duration and will not be interpreted as a comprehensive measure of service quality.

This wording is still a synthetic illustration. A real study would specify the observation window, eligibility rules, handling of incomplete requests, and rationale for calendar rather than working time. Those decisions should be made with knowledge of the actual process.

Keep the unit of analysis visible

In the example, the questionnaire produces information about employees, while the administrative system records requests. The rows are about different units.

Before combining them, ask how the relationship will be represented. Can requests be attributed to individual employees? Do several staff members work on each request? Is the relevant comparison between employees, teams, or time periods?

Do not repeat one employee’s questionnaire score beside every request and assume you now have that many independent employee observations. The structure needs to be recognised in the study design and analysis.

The immediate measurement task is to record the unit clearly. For each variable, state who or what receives a value. If aggregation is needed, explain the rule and what information it loses. An employee average and a team average answer different questions, even if both are called “performance.”

If this reveals that the question needs narrowing, revisit the focused research-question guide before finalising the instrument.

Check what supports the interpretation of the measure

Selecting a published measure is a starting point. Read the original development or evaluation study and the current instructions. Check whether the construct, respondents, language, administration method, and intended use resemble your project.

Reliability concerns consistency of measurement; validity concerns the support for the interpretation you want to make. A measure can be consistent without representing the intended construct adequately. BCcampus’s discussion of reliability and validity explains that distinction.

For the synthetic service-team study, a reliable questionnaire about satisfaction with supervisors would not automatically measure access to practical task guidance. The problem would remain even if its scores were easy to analyse.

The same scrutiny applies to administrative records. Ask who creates them, for what purpose, and whether recording practices are consistent. A completion field may be filled in when an employee closes a case, which could occur after the underlying work was finished. Document the meaning of the field rather than treating the label as self-explanatory.

Specify scoring and missingness without inventing a new instrument

If your chosen instrument produces a total or subscale score, record its actual instructions. These may address response coding, reverse-scored items, and how to handle unanswered questions. Preserve the version and cite the source of those rules.

Do not silently average items from different measures because they seem to discuss the same topic. The resulting score would require its own justification. Boateng and colleagues’ scale-development primer describes how item selection, structure, reliability, and validity form a wider development process.

For a direct record-based variable, write an equally clear rule. “Missing” is not always equivalent to “none.” An unanswered question about support use may mean that a respondent skipped it; it does not establish that they used no support.

You do not need to complete dataset cleaning at this stage. You do need to decide what the values are intended to mean, so the later codebook and analysis can preserve that meaning.

Use a final definition check

For each important variable, check that your notes identify:

the conceptual definition and its scholarly basis;
the actual source of the observation or response;
the unit receiving a value;
the relevant timeframe;
the instrument version or record-field definition;
the scoring or recording rule;
the treatment of unavailable or inapplicable observations;
the interpretation the evidence supports, and the claims it does not support.

Try handing the definition to someone who has not helped design the study. Could they explain what a high, low, absent, or missing value would mean? If their explanation differs from yours, resolve that ambiguity before collecting data.

For guidance on whether your planned measurements fit your question and available evidence, explore our research support. The aim is to make your methodological choices clear and defensible while keeping responsibility for the study with you.