Audience Management and Research Methods in Communication
Reasons for Audience Management
- Ensure product/service/company meets audience needs
- Cater to the audience
- Gain feedback
- Identify market gaps
Focus Groups (Case Study)
Engage in guided conversations with 6-10 people and a moderator on a specific topic for 45-90 minutes. Participants may or may not know each other. Convenient sampling is used, leading to systematic planning for a more scientific approach.
Types of Focus Groups:
- All members belong to the same pre-existing group
- Participants share the same role or job
- Volunteers recruited through flyers, banners, etc.
Focus groups utilize verbal, open-ended questions, allowing for follow-up comments and building upon each other’s statements. This fosters an exploratory environment.
Types of Questions:
- Engagement questions (e.g., “How often do you travel?”)
- Exploration questions (e.g., “What factors influence your travel decisions?”)
- Exit questions (e.g., “Do you have any additional comments?”)
Social Desirability
Social desirability refers to how people’s behavior changes when they are aware of being observed and judged by others. It encompasses how a company makes us feel, view ourselves, and how others perceive us. The process involves creation, obtaining, recruitment, analysis, and reporting.
Synthesize
Identify common themes across responses for each question. Assign a number to each category and label each entry with the number of the category that best fits.
Sampling
Focus on a homogeneous group to maximize disclosure among participants. This can be achieved through nomination or random selection.
Population
The population represents the entire set of people, cases, units, or events with specific characteristics that are being investigated. It is the group about which conclusions are drawn. The process involves moving from the population to the sampling process and finally to the sample.
Sample
A sample allows us to learn about a large population by observing a smaller representative group. The goal is to convince people that the small group accurately represents the larger population.
Research Methods
Top-Down Approach
This approach uses close-ended questions to gather quantitative data.
Loaded Questions
Loaded questions force respondents into answers that do not accurately reflect their opinions.
Internal Validity (Non-Spuriousness & Threats to Validity)
Internal validity ensures that the independent variable is the only factor influencing the dependent variable. Randomization, where participants have an equal chance of being placed in any group, helps achieve this. First-come-first-serve recruitment can also be used to avoid systematic biases.
Random Sampling
Random sampling involves:
- Random selection of participants from the population
- Equal likelihood of selection for each subject
- Representation of the population
Non-Spuriousness
Non-spuriousness involves ruling out alternative explanations after establishing time order (effect occurring after the cause) and co-variance (effect varying with the cause).
External Validity (Ecology, Population, Across Time)
External validity ensures that research findings can be generalized to other similar real-life situations.
Social Monitoring vs. Social Listening
Social monitoring focuses on tracking mentions and conversations related to your brand, particularly in the context of customer care (micro-level). Social listening aims to gain insights into customer sentiment regarding your brand and industry as a whole (macro-level).
Text Analytics
Text analytics involves extracting meaning and insights from unstructured content through extraction, categorization, clustering, and sentiment analysis.
Choosing Keywords
Effective keyword selection involves considering:
- False positives (keywords that match but are not relevant to the topic)
- False negatives (relevant keywords that do not contain tracked words)
- Context-specific keywords, which require gradual fine-tuning
Social Media Sentiment
Social media sentiment analysis identifies the polarity (positive, negative, neutral), subjectivity, irony, and emotional agreement expressed in online content.
Sentiment Analysis
Sentiment analysis utilizes aspect-based analysis, object/subject discrimination, and irony detection (currently in beta). Its purpose is to understand the author’s emotions and attitudes within a specific text, reflecting personal emotions.
Engagement (Increase Sentiment)
Engagement strategies to enhance sentiment include:
- Reactive engagement: Responding to comments, mentions, and direct messages on social media
- Proactive engagement: Initiating interactions with other users
Effective engagement aims to maximize positive interactions while promptly addressing negative mentions. This provides valuable insights for customer service, brand strategy, product development, competitor analysis, and crisis management.
Search Listening
Search listening involves analyzing anonymized, categorized, and aggregated search data. Anonymity ensures that individuals are not personally identifiable. Categorization involves determining the topic of a search query. Aggregation involves grouping similar searches together.
Google Search (Trends, Suggest/Autocomplete)
Google Search data, including trends, suggestions, and autocomplete, can be used for:
- Brand monitoring and reputation management
- Marketing strategy development
- Campaign effectiveness evaluation
- Competitor analysis
This data informs search engine optimization (SEO) and content creation strategies, leading to higher search engine rankings.
Relative Popularity
Relative popularity tracks how interest and the number of searches for a specific keyword change over time.
Cost Per Click (CPC)
CPC determines the cost advertisers pay for ads placed on websites or social media based on the number of clicks the ad receives.
A/B Testing and Framing
A/B testing involves comparing two versions of something (e.g., a webpage, an email) to see which performs better. Framing involves selecting specific aspects of a perceived reality and making them more salient in a communication text. This influences the audience’s perception and consumption of information. A/B testing allows for testing small variations one at a time to determine what works best for the current audience.
Elements to Test
Elements to test in A/B testing include:
- Post text (length, style, emojis, digits or numbered lists, punctuation, tone of voice (casual vs. formal, passive vs. active))
It is crucial to have clear social media goals and a specific question in mind when conducting A/B tests.
Outcome Variables (Dependent Variables)
Outcome variables in communication research often include attitudes towards the product/brand and intentions (e.g., buy, try, share, like, follow).
Information-Community-Action Framework
This framework categorizes tweets into three types:
- Information tweets: Provide factual information
- Community-building tweets: Foster social engagement
- Action tweets: Encourage specific actions
Data
Data analysis involves:
- Inclusion: Determining which tweets are relevant to the research question
- Exclusion: Identifying tweets that are not relevant, such as non-directed tweets, retweets, and quoted tweets
Latent Dirichlet Allocation (LDA)
LDA is a statistical model used for topic modeling. It assumes a set number of topics within a collection of documents. A topic is represented as a distribution of words. Words that are important for a topic tend to co-occur in documents more frequently than expected by chance. These co-occurrences are crucial for assigning words to specific topics.
Inductive Approach
An inductive approach to LDA involves selecting only interpretable and mutually exclusive topics. This selection is based on the coders’ background knowledge of the crisis and expertise in public relations. The chosen topics include items with similar meanings based on prominent words and exemplary documents. This approach is particularly useful when there is no published research on the specific topic being investigated.
Media Coverage Sharing
Media coverage sharing refers to how the crisis unfolds based on news reports.
Opinion Expressions
Opinion expressions reflect the public’s organizational identifications.
Causation
Establishing causation requires:
- Temporal sequence: The cause must precede the effect
- Covariance: Changes in the cause must lead to changes in the effect
- Non-spuriousness: All changes in the effect are solely due to the cause
Spurious (Fake) Relationship
A spurious relationship occurs when a third variable (a lurking variable) is responsible for the observed relationship between two other variables. This can happen when two variables share the same cause or are influenced by a common underlying factor.
Types of Research Questions
Descriptive Research Questions
Descriptive research questions do not involve hypothesis testing. They aim to describe phenomena, such as job satisfaction or opinions on a particular topic. They often focus on means, standard deviations, percentages, and frequencies.
Comparative Research Questions
Comparative research questions compare two or more groups, often using A/B testing. They seek to determine if there is a significant difference between the groups being compared.
Relational Research Questions
Relational research questions assess the relationship between two or more variables or groups. They often involve testing correlations and making predictions. For example, a relational research question might explore whether there is a correlation between the number of students in a class and their academic performance.
Correlation
Correlation indicates a relationship between two or more variables. It can be positive (as one variable increases, so does the other) or negative (as one variable increases, the other decreases). Correlation does not imply causation.
Essentials of Experimentation
Key elements of experimental research include:
- Random assignment of subjects to treatment and control groups to minimize systematic error
- Manipulation of at least one stimulus or treatment
- Presence of a control condition
- Observation of the effects of the manipulation
Levels of Measurement
There are four levels of measurement:
- Categorical/Discrete/Nominal: Variables with distinct categories (e.g., occupation). Comparisons can only be made between two nominal values.
- Ordinal: Variables with ordered categories (e.g., Likert scale). The intervals between orders are unknown, so it is not possible to determine how much larger or stronger one category is compared to another.
- Interval: Variables with equal intervals between measurements but without an absolute zero point (e.g., TOEFL scores). Only addition and subtraction are possible. It is not valid to say that one score is twice as large as another.
- Ratio: Variables with an absolute zero point, representing the absence of the attribute being measured (e.g., income). Ratio variables have all the properties of interval variables, allowing for comparisons of distances between values and relative magnitudes.
The Experimental Paradigm
The experimental paradigm is a systematic and scientific approach to research where the researcher manipulates one or more variables while controlling and measuring any changes in other variables. This approach aims to establish causal relationships.
Consistency Effects
Reciprocity Effects
Reciprocity effects occur when respondents feel pressured to provide answers in the same direction, potentially leading to biased results.
Priming Effects
Priming effects occur when certain ideas or events are brought into consciousness, influencing subsequent responses. For example, asking about current events before asking about family relationships might prime respondents to focus on negative aspects of their family life.
Contrast Effects
Hypocrisy Effects
Hypocrisy effects arise when respondents are asked general questions followed by more specific questions on the same topic. This can lead to inconsistencies in responses as respondents try to appear consistent.
Fatigue Effect
Fatigue effect occurs when respondents become tired or bored due to the length or tediousness of a questionnaire, potentially affecting the quality of their responses.
Informed Consent
Informed consent requires providing participants with information about the research, including its purpose, procedures, expected duration, foreseeable risks or discomforts, safeguards to minimize risks, and any benefits to the subject or others. It also includes explaining the extent of the research’s contribution to the body of knowledge.
Anonymity vs. Confidentiality
Anonymity means that the researcher cannot identify the participant. Confidentiality means that the researcher knows the participant’s identity but keeps it secret.
Deception
Deception in research involves concealing the researcher’s identity (affiliation, purpose). Justification for deception must be argued and approved by institutional review boards. Debriefing participants after the study is essential to explain the deception and its rationale.
Before Data Collection
Before collecting data, researchers should:
- Conduct a SWOT analysis to identify strengths, weaknesses, opportunities, and threats related to the research topic
- Gather up-to-date media insights and apply common sense
- Define clear research objectives
- Design research methods relevant to both objectives and background research
- Develop questionnaire items that align with objectives and background research
- Generate evidence-based insights and strategies
- Engage in a back-and-forth process of refinement and improvement
Hypothesis Testing
Null Hypothesis
The null hypothesis is the claim being challenged in a research study. It states that there is no relationship among the variables under study. Researchers aim to reject the null hypothesis, providing evidence for a relationship between the variables. The nature of science emphasizes rejecting rather than confirming hypotheses.
Alternative Hypothesis
The alternative hypothesis challenges the null hypothesis, suggesting that there is a relationship between the variables.
Significance (p > .05)
Significance refers to the degree of likelihood that an observed relationship could be attributed to sampling error alone. A p-value less than .05 indicates that the observed relationship is statistically significant, meaning it is unlikely to be due to chance.
Data Analysis
Data analysis steps include:
- Data cleaning: Preparing the data for analysis by addressing missing values and inconsistencies
- Defining experimental conditions: Specifying the different groups or treatments being compared
- Computing variables: Creating new variables based on existing data
- Correlation analysis: Examining the relationship between two or more variables
- Regression analysis: Modeling the relationship between a dependent variable and one or more independent variables
- T-test: Comparing the means of two groups
- Calculating p-values: Determining the statistical significance of the findings
Control
Control in experimental research involves ensuring that the only difference between the groups being compared is the treatment or manipulation being studied. This is essential for establishing causal relationships.
