Public Relations Research and Insight Development

Introduction

Public relations (PR) is often described as a communication function, but effective communication begins before messages are designed. Organisations must first understand their publics, contexts, expectations, concerns, media environments and patterns of behaviour. PR research provides the evidence for that understanding. It reduces reliance on assumptions, supports strategic decision-making and enables practitioners to evaluate whether communication has created meaningful change.

Research in corporate communication is not limited to measuring publicity after a campaign. Formative research is conducted before and during planning. It identifies problems, opportunities, stakeholder expectations, reputational risks, information gaps and competing interpretations of an issue. The purpose is not merely to collect data, but to develop defensible insights that can guide action.

An insight is an evidence-based interpretation that explains what is happening, why it may be happening and what the organisation should consider doing. For example, “customers dislike the company” is a broad assumption. A stronger insight might be: “Among urban customers aged 18–30, trust has declined primarily because delivery complaints receive automated responses rather than visible human resolution.” The second statement links evidence, an audience, a mechanism and a possible communication implication.

Learning Objectives

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

• explain the role of formative research in public relations and corporate communication;
• distinguish qualitative, quantitative and mixed-methods research;
• select appropriate methods, including surveys, interviews, focus groups, observation, content analysis, social listening, media analysis and secondary research;
• apply basic principles of sampling, validity and reliability;
• recognise ethical, legal and governance requirements in communication research;
• analyse evidence and convert findings into actionable insights; and
• assess the limitations of research designs and the risks of overclaiming.

The Role of Research in the PR Planning Process

A useful planning logic is:

Situation → publics → objectives → strategy → messages and channels → action → evaluation

Research contributes at every stage. Situation analysis identifies the issue and its history. Publics research identifies affected or influential groups. Channel and media research show where conversations occur. Formative research informs objectives and strategy. Evaluation research later assesses outputs, outtakes, outcomes and, where feasible, organisational effects.

Formative research commonly answers five questions:

1. What is happening?
2. Who is affected, involved or influential?
3. How do relevant publics understand the issue?
4. What factors shape their attitudes or behaviour?
5. What communication or organisational response is credible and useful?

Research does not replace professional judgement. Data can be incomplete, contradictory or biased. The practitioner’s responsibility is to make the reasoning from evidence to recommendation transparent.

A practical research-to-insight cycle

Figure: A cyclical research-to-insight process for public relations planning.

Research Paradigms and Method Selection

A research paradigm consists of assumptions about reality, knowledge and appropriate evidence. Quantitative research generally seeks patterned measurement. It may estimate the prevalence of attitudes, compare groups or test relationships. Qualitative research seeks depth, meaning and context. It may explore how stakeholders interpret an organisational action or why a message feels credible. Mixed-methods research combines both approaches when breadth and depth are needed.

No method is automatically superior. The research question should determine the design. “How many customers have seen the announcement?” may require a survey or platform analytics. “How do customers interpret the announcement?” may require interviews or focus groups. “How did news coverage frame the announcement?” calls for media or content analysis.

Formulating a Researchable Problem

A research brief should state the decision that research will inform, the population or publics of interest, the issue, the time period, constraints and intended use of findings. A vague brief such as “find out what people think” produces unfocused data. A more useful brief might ask: “What prevents first-year students from using the university’s mental-health service, and which sources would make information feel safe and credible?”

Research questions should avoid presuming an answer. “Why do students distrust us?” embeds a conclusion. “How do students assess the service’s credibility, accessibility and confidentiality?” is more open. A hypothesis may be appropriate for quantitative research, but exploratory research does not need one.

Operationalisation converts abstract concepts into observable indicators. “Trust” might be represented through perceived honesty, competence, transparency and willingness to rely on organisational information. A good operational definition does not claim to capture every aspect of a concept; it states what will be measured in this study.

Major Formative Research Methods

Surveys

A survey collects standardised responses from many participants. It is useful for estimating awareness, attitudes, behaviours, preferences and perceived credibility. Questions may be closed-ended, such as scales or multiple-choice items, or open-ended.

Good survey design uses clear, neutral and single-purpose questions. “How satisfied are you with our price and service?” is double-barrelled because price and service may be evaluated differently. Leading wording, unexplained jargon, socially desirable response options and excessively long questionnaires reduce data quality. Response options should be exhaustive where possible and mutually exclusive when only one answer is intended. A “don’t know” or “not applicable” option is often necessary.

A Likert-type item may ask respondents to indicate agreement from “strongly disagree” to “strongly agree.” Such responses are ordinal: the categories have an order, but the distance between categories cannot automatically be assumed equal. Researchers should therefore avoid treating every scale as a precise measurement without considering its construction and analysis.

Pilot testing with people resembling the target respondents can reveal confusing wording, technical problems and missing options. Surveys can be administered online, by telephone, face-to-face or through panels. Each mode affects coverage, accessibility and response behaviour.

Interviews

An interview is a purposeful conversation designed to elicit experiences, interpretations and explanations. Structured interviews use the same questions in the same order. Semi-structured interviews use a guide but allow probing. Unstructured interviews are more open and are useful in early exploration, though they are harder to compare systematically.

Interviewers should ask open questions, listen actively, avoid arguing and distinguish clarification from leading. Probes such as “Can you describe what happened next?” encourage detail. The interviewer should record consent, context and reflexive notes about their own influence. A small number of information-rich interviews can be valuable, but findings should not be presented as statistically representative unless the design supports that claim.

Focus groups

A focus group uses guided group discussion to explore shared language, disagreement, norms and reactions to concepts or messages. Interaction among participants is part of the data: one participant may challenge or stimulate another. Groups should be composed thoughtfully. Large power differences, such as supervisors and subordinates in one group, may suppress honest responses.

A skilled moderator establishes ground rules, balances participation and avoids signalling a preferred answer. Confidentiality cannot be guaranteed fully because participants hear one another. Sensitive topics may therefore be better explored through individual interviews. Focus groups reveal how people talk about an issue in a social setting, not how frequently a view exists in the entire population.

Observation

Observation records behaviour, environments and interactions rather than relying only on what people say. It may be participant or non-participant, overt or covert, structured or open. In PR, observation can examine customer-service encounters, event participation, workplace communication or how visitors navigate an information service.

An observation protocol should specify what will be observed, when, where and how behaviour will be recorded. Researchers must distinguish description from interpretation. “The participant looked at the sign for eight seconds and asked for help” is descriptive; “the sign was ineffective” is an interpretation requiring further evidence.

Content analysis

Content analysis systematically examines texts, images, audio or video. It may be quantitative, using a coding scheme to count features, or qualitative, interpreting meanings and frames. A study might code whether news stories mention safety, cost, responsibility, affected people and organisational response.

The researcher defines the unit of analysis, population, sample, variables and coding rules before coding. A codebook gives each category a definition and examples. Categories should be sufficiently clear that trained coders can apply them consistently. Inter-coder agreement should be assessed when more than one coder is used. Agreement is not proof that the categories measure the right concept, but low agreement signals ambiguity or inadequate training.

Social listening

Social listening examines publicly available digital conversations, including posts, comments, forums and reviews, to identify topics, emerging concerns, language and networked reactions. It can support issue detection and rapid situational awareness. Keyword counts alone are not equivalent to public opinion. Automated sentiment tools can misread sarcasm, slang, multilingual content, images and context. Bots, coordinated activity, duplicate posts and platform-specific user populations can distort apparent volume.

A responsible listening study documents search terms, platforms, time period, inclusion and exclusion rules, language handling, duplicates and the limits of access. Public availability does not automatically make every use ethically harmless. Researchers should minimise collection of personal identifiers and avoid exposing individuals unnecessarily.

Media analysis

Media analysis considers how an organisation or issue is represented in news and other mediated environments. It may assess prominence, source visibility, tone, key messages, frames, inaccuracies, geographic distribution and the presence of competing voices. Tone scores require explicit coding rules; a headline that mentions criticism is not necessarily negative overall.

Media analysis should distinguish media content from audience impact. A large number of stories does not demonstrate changed attitudes or behaviour. Where possible, findings should be combined with audience research, web analytics or stakeholder feedback.

Secondary data

Secondary data were collected previously for another purpose. Sources include census data, regulatory filings, customer records, prior surveys, industry research, academic studies, complaint logs, web analytics and public policy documents. Secondary research is efficient and provides historical context, but definitions, sampling, missing data and collection motives may differ from the current question.

A source evaluation asks: Who collected the data? For what purpose? When and where? From whom? Using what definitions and method? What is missing? Is the source independent, transparent and comparable? Published does not mean valid, and a precise-looking statistic can still be unsuitable.

Sampling and Representativeness

A population is the full group to which a study seeks to refer. A sampling frame is the accessible list or procedure used to identify members of that population. A sample is the selected subset. Coverage error occurs when the frame excludes relevant members. Nonresponse error occurs when people who do not respond differ systematically from respondents.

Probability sampling gives known, non-zero selection probabilities and supports statistical generalisation when implemented appropriately. Simple random sampling selects units by chance. Systematic sampling selects every kth unit after a random start. Stratified sampling divides a population into relevant groups and samples within them, which can improve representation. Cluster sampling selects groups such as schools or branches, often for practical reasons, but observations within clusters may be similar.

Non-probability methods include convenience, purposive, quota and snowball sampling. They may be appropriate for exploratory qualitative work or hard-to-reach publics, but they do not normally justify population estimates with a known sampling error. A purposive sample is selected because participants possess relevant experience, not because they mirror the population statistically.

For a proportion estimated from a simple random sample, a commonly used approximate margin of error at a chosen confidence level is:

Here, E is the margin of error, z is the standard-normal critical value for the selected confidence level, p is the estimated population proportion and n is the sample size. The approximation assumes a probability sample, independent observations and a sufficiently large sample for the normal approximation. For a 95% confidence level, z is approximately 1.96. When p is unknown, p = 0.5 gives the largest variance and therefore a conservative estimate. Finite-population corrections, weighting, clustering and nonresponse can alter the calculation. A margin of error does not correct coverage or nonresponse bias.

Worked example. Suppose a simple random sample has n = 400 and the anticipated proportion is p = 0.50. At 95% confidence:

E = 1.96 × √(0.50 × 0.50 ÷ 400)

The result is approximately E = 0.049, or 4.9 percentage points. Thus, under the stated assumptions, an estimate of 50% would have an approximate interval of 45.1% to 54.9%. This is not a guarantee that the true value lies in that interval, and it says nothing about non-sampling bias.

Qualitative sampling uses concepts such as relevance, variation, information power and saturation. Saturation is not a mechanical number: it refers to the point at which additional data produce little new insight for the research question. Researchers should explain how they judged adequacy rather than treating a small sample as automatically sufficient.

Validity, Reliability and Research Quality

Validity concerns whether a study supports the interpretation being made. Face validity asks whether a measure appears sensible. Content validity asks whether it covers the relevant dimensions of a concept. Construct validity considers whether the measure behaves as expected in relation to the theoretical concept. Criterion-related validity considers association with an appropriate external criterion. Internal validity concerns whether a study supports a causal explanation; external validity concerns transfer or generalisation beyond the study setting.

Reliability concerns consistency. Test–retest reliability examines stability over time. Internal consistency examines whether items intended to measure a construct produce coherent responses. Inter-coder reliability examines consistency between coders. Reliability is necessary but not sufficient: a measure can consistently measure the wrong thing.

Quantitative quality improves through clear operational definitions, piloting, standardised procedures, appropriate sampling and transparent analysis. Qualitative quality may be strengthened through triangulation, thick contextual description, reflexivity, negative-case analysis, an audit trail and participant checking where appropriate. Triangulation means comparing different sources, methods, researchers or theoretical perspectives; it does not mean forcing all sources to agree.

Researchers should separate data, interpretation and recommendation. A useful evidence table records the finding, source, strength, alternative explanations, implication and confidence level. This makes uncertainty visible and prevents a striking quotation from being treated as proof of prevalence.

From Findings to Insight

The movement from raw data to insight can be organised into five steps:

1. Prepare: clean, transcribe, anonymise and document the data.
2. Describe: identify distributions, recurring terms, events, differences and notable absences.
3. Compare: examine differences by relevant public, time, channel, location or experience.
4. Explain: consider mechanisms, context and alternative interpretations.
5. Act: state what the organisation should learn, change, test or communicate.

A strong insight normally contains four elements: an observed pattern, the public or context concerned, a plausible explanation and a strategic implication. For example: “New customers understand the sustainability claim but doubt its credibility because the evidence is difficult to find; trust may improve if the claim is supported by accessible, independently verifiable proof rather than more promotional language.” This is still a proposition to test, not an unquestionable fact.

A finding is not an insight merely because it is surprising. Insights should be relevant to a decision, grounded in more than one weak signal where possible, sensitive to context and expressed at an appropriate level of certainty.

Ethics, Privacy and Governance

Research ethics begins with respect for persons, informed consent, minimisation of harm, fairness and responsible use of data. Participants should understand the purpose, procedures, risks, benefits, voluntary nature of participation and withdrawal arrangements. Consent must be meaningful, not hidden in confusing terms. Incentives should compensate time without exerting undue pressure.

Confidentiality means the researcher protects information from unauthorised disclosure; anonymity means identity is not collected or cannot reasonably be linked to the data. These are different conditions. Small samples and distinctive quotations can make re-identification possible even after names are removed.

Organisations should collect only data necessary for the stated purpose, restrict access, define retention periods, secure files and dispose of data safely. Applicable privacy and data-protection laws must be followed, including requirements concerning lawful processing, sensitive information, children, cross-border transfer and data-subject rights. Institutional review or organisational ethics approval may be required.

Digital research raises additional issues. Public posts may still be contextually private. Researchers should consider user expectations, platform terms, vulnerability, quotation traceability and the consequences of reproducing content. AI tools used for transcription, coding or sentiment analysis require human oversight, documentation of prompts and models where feasible, protection of confidential data, and checks for systematic error. AI-generated summaries are not evidence by themselves.

PR professionals also face conflicts of interest. Commissioned research should not be designed to manufacture favourable results. Sponsors should not suppress inconvenient findings, and reports should disclose material limitations. Research participants must not be deceived unless a strong ethical justification exists and appropriate safeguards and debriefing are provided.

Illustrative Case: Rebuilding Trust in a Food-Delivery Platform

Background. This fictional case illustrates a formative research design. A food-delivery platform experienced a rise in complaints about late deliveries and customer-service responses. Management initially proposed a campaign emphasising speed and convenience.

Problem. Existing analytics showed complaint volume, but management did not know whether dissatisfaction arose mainly from delivery performance, refund policies, communication tone or expectations created by advertising. The research question was: “What drives trust and continued use among customers who experienced a late delivery?”

Analysis and approach. The team used mixed methods. First, it reviewed twelve months of complaint records and segmented them by issue, location and resolution time. Second, it conducted a survey of recent users, using screening questions to distinguish customers with and without a late-delivery experience. Third, it held semi-structured interviews with customers who had complained. Fourth, researchers observed the complaint-resolution journey using test accounts. Finally, they analysed news and public social conversations, carefully removing duplicated posts and not publishing user identities.

The team piloted the survey and revised ambiguous questions. Interview participants received consent information and could withdraw. The report distinguished the probability sample used for the survey from the purposive interview sample. Coding rules for complaint categories were documented, and two analysts independently coded a subset to identify disagreements.

Findings and insight. The data showed that late delivery was important, but the strongest trust decline occurred when customers received an automated message that gave no specific explanation and when refund decisions were unclear. Social posts amplified unusual cases, but they could not establish their prevalence. Interviews indicated that customers tolerated occasional delays when the platform communicated early, explained responsibility and offered a predictable remedy.

The resulting insight was: “For customers affected by delay, perceived fairness and explanation are more important to continued trust than a general promise of speed.” The recommended response was therefore not simply a faster campaign. The organisation redesigned delay alerts, clarified refund rules, empowered service agents to resolve cases and tested a message explaining the operational changes.

Outcome. In a subsequent evaluation, the organisation tracked resolution time, repeat use among affected customers, comprehension of the refund policy and trust ratings. It did not claim that the communication campaign alone caused all improvement because operational changes occurred simultaneously.

Learning points. Research changed the problem definition, connected communication with organisational behaviour, combined breadth and depth, and prevented overclaiming. It also demonstrated that listening data should stimulate investigation rather than dictate conclusions.

Challenges and Limitations

Research can be distorted by weak problem definition, convenient samples, low response rates, poorly worded questions, interviewer effects, group conformity, selective platform visibility, coding ambiguity and confirmation bias. Organisational pressure may encourage researchers to highlight favourable findings. Time and budget constraints can make a perfect design impossible.

The answer is not to claim certainty. Researchers should prioritise the decision, use the strongest feasible design, record limitations, compare sources and recommend further testing where uncertainty matters. Correlation should not be presented as causation. Positive media coverage does not prove reputation improvement, and online engagement does not necessarily indicate behavioural change.

Best-Practice Checklist

Before conducting research, clarify the decision, define the public, formulate neutral questions and select methods that fit the information need. During collection, pilot instruments, obtain consent, train researchers, protect data and document deviations. During analysis, preserve an audit trail, examine disconfirming evidence, distinguish samples from populations and use appropriate claims. During reporting, present methods, limitations, uncertainty, evidence strength and practical implications. After action, evaluate what changed and whether the original insight remains valid.

Chapter Summary

Public relations research provides the evidence base for understanding situations, publics, issues and communication environments. Formative research is especially important because it informs problem definition, objectives, strategy and message development before resources are committed. Surveys provide structured breadth; interviews and focus groups provide interpretive depth; observation examines behaviour in context; content and media analysis examine messages and representations; social listening detects digital conversations; and secondary data supply historical and environmental context.

Sound research depends on a clear question, appropriate sampling, credible measurement and ethical practice. Probability samples may support statistical generalisation under stated assumptions, whereas purposive and convenience samples serve different exploratory purposes. Validity concerns the defensibility of an interpretation, while reliability concerns consistency. Neither eliminates bias by itself.

The professional task is to move carefully from evidence to pattern, explanation, insight and action. Researchers must disclose limitations, avoid confusing visibility with prevalence or correlation with causation, protect participants and use technology—including AI—with human oversight. The strongest PR research connects communication with organisational conduct and treats insight as a testable basis for decision-making rather than a decorative conclusion.

Key Terms

Formative research: Research conducted to understand a situation, public or issue before or during programme planning.

Insight: An evidence-based interpretation linking a pattern to an explanation and a possible action.

Population: The complete group to which a study seeks to refer.

Sampling frame: The accessible list or procedure from which a sample is selected.

Probability sampling: Sampling in which members have known, non-zero selection probabilities.

Purposive sampling: Deliberate selection of participants because they possess relevant characteristics or experience.

Coverage error: Bias caused when the sampling frame excludes parts of the target population.

Nonresponse error: Bias arising when respondents differ systematically from nonrespondents.

Operationalisation: Converting an abstract concept into observable indicators or measures.

Validity: The extent to which evidence supports the intended interpretation or use of a measure.

Reliability: The consistency or stability of a measurement or coding procedure.

Triangulation: Comparing multiple data sources, methods, researchers or perspectives.

Content analysis: Systematic examination and coding of communication content.

Media analysis: Study of how issues or organisations are represented in media channels.

Social listening: Analysis of digital conversations to identify topics, reactions and emerging issues.

Secondary data: Data previously collected for a different purpose.

Inter-coder reliability: Consistency among people applying a coding scheme.

Informed consent: Voluntary agreement based on adequate information about participation.

Confidentiality: Protection of information from unauthorised disclosure.

Operational definition: A specification of how a concept will be observed or measured in a study.

Review Questions

Short Answer Questions

1. What is formative research, and why is it important in public relations planning?
2. Distinguish validity from reliability.
3. State two advantages and two limitations of social listening.
4. What is the difference between a population, a sampling frame and a sample?
5. Why should a researcher distinguish a finding from an insight?

Long Answer Questions

1. Compare surveys, interviews and focus groups as methods for researching stakeholder trust.
2. Explain how a content-analysis codebook improves the quality of a media study.
3. Discuss the ethical issues involved in using publicly available social-media posts in PR research.
4. Explain probability and non-probability sampling and identify situations in which each may be appropriate.
5. Describe how a PR researcher can move from raw data to an actionable insight without overclaiming.

Analytical / Application Questions

1. A university receives many online complaints about its admissions process. Design a mixed-methods formative research plan, including questions, samples, methods and ethical safeguards.
2. A survey reports that 68% of respondents trust a company, but the sample was drawn from the company’s email subscribers and had a low response rate. Analyse what can and cannot be concluded.
3. News coverage of a product recall is predominantly negative, while customer-service records show improving satisfaction. Propose an analysis that explains the apparent contradiction and recommends next steps.

Multiple Choice Questions with Answers

Question 1

Which statement best describes formative research in public relations?

A. Research conducted only after a campaign to calculate publicity value

B. Research used to understand a situation and publics before or during planning

C. Research that guarantees a campaign will change behaviour

D. Research limited to counting media mentions

Correct Answer: B. Research used to understand a situation and publics before or during planning

Question 2

Which method is most suitable for exploring how stakeholders interpret an organisational apology?

A. A semi-structured interview

B. A simple frequency count of website visits

C. A random number generator

D. A financial audit

Correct Answer: A. A semi-structured interview

Question 3

What is a sampling frame?

A. The complete theory explaining a public’s behaviour

B. The final report delivered to management

C. The accessible list or procedure from which a sample is selected

D. The number of questions in a questionnaire

Correct Answer: C. The accessible list or procedure from which a sample is selected

Question 4

A measure gives very consistent results but does not capture the intended concept. Which conclusion is most accurate?

A. It is reliable but may lack validity

B. It is valid because it is consistent

C. It is representative by definition

D. It has no measurement properties

Correct Answer: A. It is reliable but may lack validity

Question 5

At 95% confidence, with n = 400 and p = 0.50, the approximate margin of error for a simple random sample is:

A. 0.49 percentage points

B. 2.5 percentage points

C. 4.9 percentage points

D. 9.8 percentage points

Correct Answer: C. 4.9 percentage points

Question 6

Which claim about social listening is most defensible?

A. The number of posts always equals the number of people holding a view

B. Automated sentiment classification is never affected by sarcasm

C. Public posts automatically provide a representative sample of public opinion

D. Social listening can identify emerging conversations but requires attention to bias and context

Correct Answer: D. Social listening can identify emerging conversations but requires attention to bias and context

Question 7

What is the main purpose of a codebook in content analysis?

A. To define coding categories and guide their consistent application

B. To increase the number of media stories about an organisation

C. To replace the need for a research question

D. To guarantee that findings are causal

Correct Answer: A. To define coding categories and guide their consistent application

Question 8

Which practice best protects research participants?

A. Collecting every available personal identifier

B. Explaining the study and obtaining voluntary informed consent

C. Publishing identifiable quotations without notice

D. Hiding the purpose of the research from all participants

Correct Answer: B. Explaining the study and obtaining voluntary informed consent

Question 9

A purposive interview sample is primarily selected because participants:

A. Are selected by a random-number table

B. Represent exactly the national population

C. Possess relevant experience or characteristics

D. Responded first to an advertisement

Correct Answer: C. Possess relevant experience or characteristics

Question 10

Which statement correctly distinguishes media analysis from audience research?

A. Media analysis examines representation; it does not by itself prove audience impact

B. Media analysis always measures behavioural change directly

C. Audience research can only use social-media data

D. Media analysis eliminates the need to define a sample

Correct Answer: A. Media analysis examines representation; it does not by itself prove audience impact