Business Problems, Research Questions and Conceptual Frameworks

Introduction

Business analytics begins with a decision, but rigorous research begins by transforming that decision into a researchable problem. Managers commonly express concerns in practical language: sales are falling, customers are leaving, delivery costs are rising, or a new digital service is not being adopted. Such statements are important, but they are not yet research problems. A research problem identifies a gap between what is known and what decision-makers need to know, specifies the phenomenon to be studied, and establishes boundaries for systematic investigation.

This chapter explains how to move from a managerial problem to research objectives, research questions, hypotheses, and a conceptual or theoretical framework. It also explains scope, units of analysis, variables, operational definitions, causal reasoning, and the ethical implications of framing a business study. These skills are central to business analytics because the quality of an analytical result depends not only on the method used, but also on whether the right problem was investigated.

A poorly framed problem can produce technically impressive but strategically irrelevant analysis. Conversely, a well-framed problem allows the researcher to select appropriate data, methods, measures, and decision criteria.

Learning Objectives

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

1. Distinguish a managerial problem from a research problem.
2. Formulate a clear problem statement and research purpose.
3. Write aligned research objectives, questions, and hypotheses.
4. Define the scope, context, population, sample, and unit of analysis.
5. Distinguish constructs, variables, indicators, moderators, mediators, and controls.
6. Build and explain a conceptual or theoretical framework.
7. Assess whether proposed relationships are logically and ethically defensible.
8. Translate a framework into an actionable analytics plan.

1. From a Managerial Concern to a Research Problem

1.1 The managerial problem

A managerial problem is a situation requiring a decision or intervention. It is usually expressed in terms of undesirable outcomes or uncertainty. Examples include:

• A subscription service has experienced a rise in customer cancellations.
• A retailer wants to know whether personalized promotions increase repeat purchases.
• A bank is considering an artificial-intelligence credit-screening system.
• A manufacturer is experiencing inconsistent production quality.

The managerial problem is action-oriented. It asks, “What should the organization do?”

1.2 The research problem

A research problem is a precise statement of what must be investigated to support a decision. It asks, “What evidence is needed, and how can it be obtained systematically?” For example, the concern “customers are leaving” may become:

The company does not know whether cancellations are primarily associated with service failures, price dissatisfaction, low usage, or competitor switching, and it lacks evidence about which factors are most predictive after customer tenure and plan type are considered.

This statement identifies uncertainty, possible explanatory factors, and relevant controls. It does not prematurely assume that price is the cause merely because managers suspect it.

1.3 Symptoms, causes, and evidence gaps

A visible business outcome is often a symptom rather than a cause. Declining sales may result from price, availability, product quality, competition, seasonality, distribution, or changes in customer composition. A useful diagnostic sequence is:

1. What exactly has changed?
2. Compared with what baseline, period, or benchmark?
3. For which customers, products, locations, or time periods?
4. What explanations are plausible?
5. What evidence would distinguish among those explanations?
6. What decision will the evidence inform?

Researchers should separate three elements:

• Observed condition: what is happening, such as a 12% decline in monthly renewals.
• Possible explanation: why it may be happening, such as service dissatisfaction.
• Information gap: what is not yet known, such as whether dissatisfaction predicts renewal after controlling for usage.

1.4 Exploratory, descriptive, explanatory, and evaluative purposes

The purpose of a study influences the research question and design.

These purposes can coexist, but they should not be confused. A model that predicts churn accurately does not necessarily establish why customers churn or which intervention will reduce churn.

2. The Research Problem Statement and Purpose

A strong problem statement normally contains five components:

1. Context: the organization, industry, market, or process.
2. Undesirable condition or decision need: the issue requiring attention.
3. Evidence of significance: financial, operational, customer, social, or strategic consequences.
4. Knowledge gap: what is unknown or inadequately understood.
5. Researchable focus: the population, variables, and setting to be examined.

A useful structure is:

In [context], [problem or decision need] is occurring, creating [consequence]. Although [existing knowledge or practice] is available, [specific gap] remains unresolved. This study will investigate [population, variables, and setting] to support [decision or outcome].

The purpose statement then expresses what the study will do. For example:

The purpose of this study is to examine whether perceived service reliability, price fairness, and platform usability are associated with renewal intention among small-business subscribers, while considering customer tenure and plan type.

The statement is appropriately cautious: “associated with” does not claim causation. Causal language should be used only when the design and evidence justify it.

2.1 Feasibility and significance

A problem may be important but not feasible. Before finalizing it, assess:

• Access: Can the researcher obtain the required data and participants?
• Time: Can the study be completed within the available period?
• Resources: Are software, expertise, and funding adequate?
• Measurement: Can the concepts be observed or measured credibly?
• Ethics: Can privacy, consent, fairness, and potential harm be managed?
• Decision relevance: Will the findings inform a real decision?

A practical research problem balances importance with answerability.

3. Research Objectives, Questions, and Hypotheses

3.1 Research objectives

Research objectives convert the purpose into observable tasks. A broad objective states the overall intention; specific objectives divide it into achievable components.

Weak objective: “To understand customer experience.”

Stronger objectives:

1. To describe customer perceptions of service reliability and platform usability.
2. To examine the association between perceived service reliability and renewal intention.
3. To determine whether the relationship differs by customer tenure.
4. To identify the factors most strongly associated with actual renewal behavior.

Objectives should use precise verbs such as describe, compare, estimate, examine, assess, test, evaluate, and identify. Avoid vague verbs such as know or learn unless they are clarified.

3.2 Research questions

A research question is a focused question that can be answered using evidence. It should identify the population, context, key concepts, and sometimes the relevant time period.

Examples:

• What proportion of subscribers renewed during the 2024 contract cycle?
• How do subscribers rate service reliability and platform usability?
• Is perceived service reliability associated with renewal intention?
• Does customer tenure change the relationship between service reliability and renewal intention?
• Which factors best predict actual renewal among eligible subscribers?

A study may include one central question and several subquestions. Each question should connect to at least one objective and to a feasible source of evidence.

3.3 Hypotheses

A hypothesis is a testable proposition about an expected relationship or difference. A directional hypothesis specifies the expected direction; a nondirectional hypothesis states that a relationship or difference exists without specifying direction.

Examples:

• H1: Perceived service reliability is positively associated with renewal intention.
• H2: Perceived price fairness is positively associated with renewal intention.
• H3: The positive association between service reliability and renewal intention is stronger among newer customers.

The corresponding null hypothesis for H1 would state that there is no association in the population. Researchers do not “prove” a hypothesis through a sample; they assess whether the evidence is sufficiently inconsistent with a specified null model under stated assumptions.

Hypotheses should be:

• conceptually justified;
• empirically testable;
• stated in terms of measurable constructs;
• neither so broad that they are untestable nor so narrow that they lack significance;
• consistent with the study design.

A cross-sectional observational study can test associations, but it usually cannot establish temporal precedence or causation by itself.

3.4 Alignment matrix

Alignment prevents a study from asking questions that its data cannot answer.

4. Scope, Population, and Unit of Analysis

4.1 Scope

Scope defines what the study includes and excludes. It may specify:

• organization or industry;
• geographic location;
• period of study;
• customer or employee segment;
• products or services;
• variables and outcomes;
• research design;
• level of analysis.

A scope statement might specify that the study concerns small-business subscribers in one country who held an active annual contract between January and December 2024. It might exclude enterprise customers, monthly contracts, and post-renewal satisfaction.

Explicit exclusions are valuable because they prevent unjustified generalization.

4.2 Population, sample, and sampling frame

The target population is the complete group to which the researcher wishes to generalize. The accessible population is the portion practically available. A sampling frame is the operational list from which cases are selected. The sample is the set of cases actually studied.

For example, the target population may be all small-business subscribers, the accessible population may be subscribers in three regions, the sampling frame may be the customer database, and the sample may be 1,200 randomly selected accounts.

A sample size does not automatically make findings representative. Coverage errors, nonresponse, duplicate records, and systematic exclusion can threaten generalizability.

4.3 Unit of analysis and unit of observation

The unit of analysis is the entity about which conclusions are drawn: a customer, employee, store, transaction, team, organization, or country. The unit of observation is the entity from which data are collected. They may be the same, but need not be.

If transaction records are used to make conclusions about customers, the transaction is the observation but the customer is the unit of analysis. If employees are nested within branches, employees may be observations while branches are a higher-level unit of analysis.

An analytical error occurs when the level of evidence does not match the level of the claim. Store-level averages cannot automatically establish that individual customers with a particular characteristic behave similarly. This is related to the ecological fallacy: relationships observed for groups may not hold for individuals.

Repeated observations also require care. Ten transactions from one customer are not equivalent to ten independent customers. Dependence among observations affects estimation and inference.

5. Concepts, Constructs, Variables, and Operationalization

A concept is a general idea, such as trust or loyalty. A construct is a theoretically meaningful concept defined for research, such as perceived service reliability. A variable is a characteristic that can take different values across cases. An indicator is an observable measure used to represent a construct.

For example, customer loyalty may be represented by renewal intention, recommendation intention, purchase frequency, or actual renewal. These are related but not identical indicators.

5.1 Types of variables

• Independent variable: a presumed predictor or explanatory factor.
• Dependent variable: the outcome to be explained or predicted.
• Control variable: a factor included to reduce alternative explanations or improve precision.
• Mediator: a mechanism through which a predictor may influence an outcome.
• Moderator: a variable that changes the strength or direction of a relationship.
• Confounder: a variable related to both predictor and outcome that may create a misleading association if not addressed.

Consider a proposed relationship between training and employee productivity. Employee skill may be a confounder if skilled employees are more likely to receive advanced training and are also more productive. Motivation may be a mediator if training increases motivation, which then increases productivity. Department type may be a moderator if training is more effective in technical departments than in administrative departments.

5.2 Operational definitions

An operational definition explains exactly how a construct will be measured. It should specify the indicator, scale, reference period, coding, and source.

Operationalization involves judgment. A single item may be inexpensive but less reliable than multiple items. Behavioral measures may avoid self-report bias but may not capture attitudes. Researchers should consider reliability, validity, sensitivity, missingness, and fairness.

6. Theoretical and Conceptual Frameworks

6.1 Theory and conceptual framework

A theory is an organized set of concepts and propositions that explains relationships and mechanisms. Examples relevant to business research include the Technology Acceptance Model, Theory of Planned Behavior, resource-based views of the firm, expectancy theory, and diffusion of innovation.

A conceptual framework is a study-specific representation of the concepts and relationships the researcher expects to investigate. It may be derived from one or more theories, prior empirical findings, professional knowledge, or qualitative evidence. A conceptual framework is not merely a list of variables; it explains why the variables are connected.

A useful framework should identify:

1. the outcome of interest;
2. principal explanatory variables;
3. expected direction or nature of relationships;
4. mediating or moderating mechanisms;
5. relevant controls and contextual conditions;
6. boundaries of the proposed explanation.

6.2 Causal logic and alternative explanations

Arrows in a framework often imply a proposed direction, but an arrow alone does not prove causation. A causal interpretation generally requires plausible mechanism, temporal ordering, appropriate comparison, and attention to alternative explanations. In observational research, researchers should use careful language and disclose limitations.

A framework should ask:

• Why should X relate to Y?
• What process connects them?
• Could Y influence X instead?
• Is a third variable responsible for both?
• Under what conditions might the relationship weaken or reverse?

6.3 Illustrative conceptual framework

The following framework concerns subscription renewal.

Figure: An illustrative conceptual framework for studying subscription renewal.

The framework proposes that reliability, price fairness, and usability influence satisfaction, which is associated with renewal intention. Satisfaction is a mediator in the proposed explanation. Tenure is a moderator: the strength of a relationship may differ between newer and longer-standing customers. Plan type, industry, and account size may be controls or confounders, depending on the research logic.

6.4 Translating a framework into hypotheses

The framework may generate the following hypotheses:

• H1: Service reliability is positively associated with customer satisfaction.
• H2: Perceived price fairness is positively associated with customer satisfaction.
• H3: Platform usability is positively associated with customer satisfaction.
• H4: Customer satisfaction is positively associated with renewal intention.
• H5: Customer satisfaction mediates the association between service reliability and renewal intention.
• H6: Customer tenure moderates the association between service reliability and renewal intention.

Each hypothesis requires an operational definition and an appropriate analytical strategy. A framework should not include relationships merely because they appear interesting; every relationship should be theoretically or empirically defensible.

7. Worked Example: Aligning a Business Analytics Study

Background

An illustrative online learning company has noticed that course-completion rates have declined. Managers propose sending more reminder messages, but they do not know whether reminders address the main barrier.

Problem diagnosis

The observable condition is lower completion. Plausible explanations include workload, course difficulty, poor interface usability, weak learner motivation, and inadequate reminders. The information gap is the relative importance of these factors and whether reminders improve completion.

Purpose and objectives

Purpose: to examine factors associated with course completion and evaluate whether reminder exposure is related to completion among enrolled adult learners.

Objectives:

1. Describe completion rates by course and learner segment.
2. Examine associations between perceived course difficulty, platform usability, motivation, reminder exposure, and completion.
3. Compare completion rates for learners receiving different reminder schedules, if assignment was systematic or randomized.
4. Develop recommendations for improving completion without increasing message fatigue.

Questions and hypotheses

Question 1: What is the completion rate across courses and learner segments?

Question 2: Which learner and course characteristics are associated with completion?

Question 3: Does reminder exposure relate to completion after accounting for prior engagement?

If reminders were randomly assigned, a causal hypothesis may be appropriate: learners assigned to the reminder condition will have a higher completion probability than learners assigned to the control condition. If reminders were sent based on prior inactivity, the study should state an association rather than a causal claim because inactivity may influence both reminder exposure and completion.

Scope and unit of analysis

The target population is adult learners enrolled in the company’s professional courses during one six-month period. The unit of analysis is the learner-course enrollment, not the individual click or message. This distinction matters because one learner may enroll in several courses. The analysis must account for repeated enrollments or restrict the study to one defined enrollment per learner.

Operationalization

Completion is coded as 1 if the learner meets the company’s stated completion criterion within 90 days and 0 otherwise. Prior engagement may be measured as the number of learning sessions during the first seven days. Usability and motivation may be measured using multi-item survey scales. Reminder exposure is defined by the number and timing of messages recorded in the communication system.

Analysis and decision use

Descriptive analysis establishes where the problem is concentrated. A predictive model may identify learners at high risk of noncompletion, but prediction should not be confused with intervention effectiveness. If the organization wants to know whether reminders cause improvement, a randomized experiment or a credible quasi-experimental design is preferable. The final recommendation should consider completion, learner experience, privacy, and the risk of excessive messaging.

Learning points

The example demonstrates that a single business symptom can support several different studies. It also shows why the unit of analysis, timing of measurement, assignment mechanism, and decision purpose must be specified before analysis begins.

8. Common Errors and Quality Checks

8.1 Common framing errors

Starting with a method. “We will use machine learning” is not a research problem. The method follows the question.

Overly broad questions. “What causes customer loyalty?” is usually too broad for a bachelor-level project.

Leading questions. “How does our excellent service increase loyalty?” assumes the conclusion.

Construct confusion. Satisfaction, loyalty, retention, and advocacy may be related but should not be treated as interchangeable.

Causal overclaiming. A correlation or predictive association does not automatically show that changing the predictor will change the outcome.

Mismatched levels. Individual conclusions drawn from branch-level data can be misleading.

Post hoc frameworks. Building a framework only after seeing the results can encourage selective explanations. Exploratory findings may generate new hypotheses, but they should be identified as such.

Unbounded scope. Including too many variables, populations, and outcomes reduces depth and feasibility.

8.2 A practical quality checklist

Before data collection, ask:

• Is the problem stated as an evidence gap rather than merely a symptom?
• Is the intended decision clear?
• Do objectives, questions, hypotheses, and measures align?
• Are the population, time period, setting, and unit of analysis explicit?
• Are key constructs operationally defined?
• Are alternative explanations considered?
• Does the design support the intended level of causal language?
• Are data access, quality, missingness, and privacy addressed?
• Could the framework be explained in a few coherent paragraphs?
• Would a different stakeholder interpret the terms in the same way?

9. Ethics, Governance, and Contemporary Analytics

Problem framing is also an ethical activity. The way an organization defines a problem determines who is measured, who is classified as risky, and which interventions are considered legitimate.

Researchers should apply data minimization: collect only information necessary for the stated purpose. Personal data should be handled according to applicable privacy law and organizational policy. Sensitive attributes require particular care, even when they are excluded from the final model, because proxy variables may reproduce their effects.

A framework involving customer scoring or employee analytics should consider:

• informed consent or another lawful basis for data use;
• confidentiality and access controls;
• retention and deletion policies;
• fairness across relevant groups;
• explainability appropriate to the decision;
• human review for consequential decisions;
• documentation of assumptions and limitations.

Artificial intelligence can assist with literature discovery, coding, text classification, and pattern detection, but it does not replace conceptual reasoning. Generative systems may produce plausible but unsupported relationships, fabricated references, or biased categories. Researchers remain responsible for verifying sources, documenting data provenance, protecting confidential information, and ensuring that an AI-assisted framework reflects defensible theory and evidence.

Contemporary business research increasingly combines transactional, survey, sensor, text, and digital trace data. This increases analytical opportunity but also makes construct validity more difficult. A click is not necessarily engagement, a sentiment score is not necessarily satisfaction, and a prediction is not necessarily an explanation. The framework must state what each data element represents and what it does not represent.

Chapter Summary

A managerial concern becomes a research problem when the decision need, evidence gap, population, variables, and context are specified clearly. A strong problem statement distinguishes observed symptoms from possible causes and identifies the information required for action.

Research objectives translate purpose into achievable tasks. Research questions express what the study will answer, while hypotheses state testable expectations about relationships or differences. These elements must be aligned with the available data and design.

Scope defines the boundaries of the study. Population, sampling frame, sample, unit of analysis, and unit of observation determine what can reasonably be concluded. Constructs must be operationalized through valid and reliable indicators. Independent, dependent, control, mediator, moderator, and confounding variables serve different analytical roles.

A theoretical framework draws on established explanations; a conceptual framework adapts relevant concepts and relationships to a particular study. Frameworks should show mechanisms, alternative explanations, contextual conditions, and limits. Arrows suggest proposed relationships but do not by themselves establish causation.

Finally, problem framing has ethical consequences. Privacy, fairness, transparency, data minimization, and responsible use of AI should be considered from the beginning rather than added after analysis. The best business analytics research is not simply technically sophisticated; it is relevant, answerable, transparent, and responsible.

Key Terms

Research problem: A clearly defined uncertainty or evidence gap requiring systematic investigation.

Managerial problem: A business situation requiring a decision or intervention.

Problem statement: A concise account of context, condition, significance, knowledge gap, and research focus.

Research purpose: The overall intention of a study.

Research objective: A specific, achievable task undertaken to fulfill the research purpose.

Research question: A focused question answerable through systematic evidence.

Hypothesis: A testable proposition about an expected relationship or difference.

Scope: The stated boundaries of a study.

Target population: The full group to which the researcher intends to generalize.

Sampling frame: The operational list or source from which cases are selected.

Unit of analysis: The entity about which conclusions are drawn.

Unit of observation: The entity from which data are collected.

Construct: An abstract, theoretically meaningful concept defined for research.

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

Independent variable: A presumed explanatory or predictor variable.

Dependent variable: The outcome being explained or predicted.

Mediator: A variable representing a mechanism through which a relationship operates.

Moderator: A variable that changes the strength or direction of a relationship.

Conceptual framework: A study-specific model of relevant concepts and proposed relationships.

Ecological fallacy: The error of inferring individual relationships from group-level evidence.

Review Questions

Short Answer Questions

1. What is the difference between a managerial problem and a research problem?
2. Why should a problem statement distinguish symptoms from possible causes?
3. Define unit of analysis and provide one business example.
4. What is the difference between a mediator and a moderator?
5. Why does a predictive relationship not necessarily establish causation?

Long Answer Questions

1. Explain the components of a strong business research problem statement and illustrate each component with an example.
2. Discuss how research objectives, research questions, and hypotheses should be aligned.
3. Explain the roles of theory and conceptual frameworks in business analytics research.
4. Compare exploratory, descriptive, explanatory, evaluative, and predictive research purposes.
5. Discuss the ethical issues that may arise when conceptual frameworks are used to design customer or employee scoring systems.

Analytical / Application Questions

1. A retailer observes declining sales but has data on price, promotions, stockouts, competitor activity, store location, and customer visits. Formulate a research problem, one purpose statement, three objectives, and three research questions.
2. A bank uses past repayment behavior to send additional reminders to customers predicted to be at risk. Explain why comparing repayment rates between reminded and non-reminded customers may not identify the causal effect of reminders. Propose a stronger design.
3. Construct a conceptual framework for employee adoption of a new analytics platform. Identify the dependent variable, at least three explanatory variables, one mediator, one moderator, and two controls. State two testable hypotheses and define how each major construct could be measured.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes a research problem from a managerial problem?

A. A research problem identifies the evidence gap underlying a decision need.

B. A research problem is always expressed as a financial loss.

C. A managerial problem must be tested with a hypothesis.

D. A managerial problem is narrower than every research problem.

Correct Answer: A. A research problem identifies the evidence gap underlying a decision need.

Question 2

Which objective is written most appropriately for a business analytics study?

A. To know everything about customer loyalty.

B. To examine the association between service reliability and renewal intention.

C. To prove that the company has excellent service.

D. To understand customers in a general sense.

Correct Answer: B. To examine the association between service reliability and renewal intention.

Question 3

A researcher concludes that customers with more support contacts are less loyal. Which factor could be a confounder if dissatisfied customers both contact support more and are more likely to leave?

A. Customer dissatisfaction

B. The font used in the report

C. The sample identifier

D. The research title

Correct Answer: A. Customer dissatisfaction

Question 4

What is the unit of analysis when transaction records are used to draw conclusions about individual customers?

A. The transaction only

B. The database table

C. The customer

D. The spreadsheet column

Correct Answer: C. The customer

Question 5

Which statement correctly describes a moderator?

A. It is the outcome variable in every study.

B. It changes the strength or direction of a relationship.

C. It is always caused by the independent variable.

D. It is an alternative name for a sampling frame.

Correct Answer: B. It changes the strength or direction of a relationship.

Question 6

Which item is an operational definition of customer tenure?

A. The general concept of long-term loyalty

B. A manager's belief that older customers are valuable

C. Months since the customer's initial activation

D. A theory explaining customer retention

Correct Answer: C. Months since the customer's initial activation

Question 7

What is the primary purpose of a conceptual framework?

A. To list every variable available in a database

B. To show and justify relevant concepts and their proposed relationships

C. To replace the need for data collection

D. To guarantee that a causal relationship exists

Correct Answer: B. To show and justify relevant concepts and their proposed relationships

Question 8

A cross-sectional observational study finds that usability is associated with renewal intention. Which conclusion is most defensible?

A. Improving usability definitely causes renewal.

B. Renewal intention causes usability perceptions in every case.

C. Usability is associated with renewal intention, although causation is not established.

D. No relationship can be reported under any circumstances.

Correct Answer: C. Usability is associated with renewal intention, although causation is not established.

Question 9

Which practice best supports responsible AI-assisted research?

A. Accepting generated references without verification

B. Uploading confidential customer data to an unrestricted system

C. Verifying sources, documenting assistance, and protecting sensitive data

D. Treating model predictions as explanations of behavior

Correct Answer: C. Verifying sources, documenting assistance, and protecting sensitive data

Question 10

Which research purpose primarily asks whether an intervention achieved its intended effect?

A. Descriptive

B. Evaluative

C. Exploratory

D. Conceptual

Correct Answer: B. Evaluative