Decision-Making, Problem-Solving and Managerial Analytics

Introduction

Managers make decisions continuously: allocating staff, responding to customer complaints, selecting suppliers, approving budgets, managing risk and deciding whether performance requires corrective action. Leadership is therefore not only the ability to inspire people; it is also the disciplined ability to define situations accurately, evaluate alternatives and act responsibly when information is incomplete.

Decision-making concerns choosing among alternatives. Problem-solving is the broader process of understanding an undesirable condition, identifying its causes and designing and implementing a response. Managerial analytics supports both activities by converting data into useful evidence. However, data does not eliminate judgement. Managers must decide what to measure, assess the reliability of information, recognise bias and consider ethical consequences.

This chapter presents rational, bounded-rationality, evidence-based and participative approaches. It also introduces root-cause analysis, SWOT analysis, risk assessment, basic data interpretation, dashboards and key performance indicators (KPIs). The emphasis is practical application at diploma level: managers should be able to use these tools appropriately, explain their reasoning and recognise their limitations.

Learning objectives

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

1. distinguish decision-making from problem-solving;
2. apply rational, bounded-rationality, evidence-based and participative decision models;
3. diagnose problems using root-cause tools and SWOT analysis;
4. assess risks using probability, impact and control responses;
5. interpret basic managerial data without overstating conclusions;
6. design useful KPIs and dashboards;
7. identify cognitive, organisational and data-related bias; and
8. make transparent, ethical and accountable recommendations under uncertainty.

1. Decision-making and problem-solving

A decision is a commitment to a course of action. It may be strategic, tactical or operational. Strategic decisions shape long-term direction; tactical decisions translate strategy into programmes and resource choices; operational decisions govern day-to-day work. Decisions may also be programmed, where established rules apply, or non-programmed, where novel judgement is required.

Problem-solving normally contains several linked decisions. A manager first decides how to define the problem, then how to investigate it, which alternatives to consider, what action to take and how to evaluate results. A poorly defined problem can make technically excellent analysis irrelevant. For example, “staff morale is low” is a symptom or broad diagnosis, not a sufficiently precise problem statement. A stronger statement might be: “Voluntary turnover among customer-service employees has risen from 8% to 15% over six months, concentrated among employees with less than one year of service.”

A useful problem statement identifies:

• the gap between expected and actual performance;
• who or what is affected;
• the time period and location;
• available evidence; and
• constraints that limit possible responses.

Managers should distinguish symptoms, causes and consequences. Late deliveries may be a symptom; inaccurate inventory records may be a cause; lost customers may be a consequence. Treating a symptom may provide temporary relief without correcting the underlying system.

2. Major decision models

The rational model

The rational model assumes that a decision-maker can define objectives, obtain relevant information, identify alternatives, predict consequences and select the alternative that best meets agreed criteria. Its typical sequence is:

1. define the problem or opportunity;
2. establish objectives and decision criteria;
3. gather and verify information;
4. generate feasible alternatives;
5. evaluate alternatives against the criteria;
6. choose an option;
7. implement the decision; and
8. monitor outcomes and learn.

The model is valuable because it encourages explicit reasoning and reduces impulsive action. A weighted decision matrix can help when several criteria matter. If alternative i receives a score sᵢⱼ on criterion j, and criterion j has weight wⱼ, the weighted score is:

Here, Sᵢ is the total score for alternative i; wⱼ is the weight assigned to criterion j; sᵢⱼ is the score of alternative i on criterion j; and n is the number of criteria. Weights should normally sum to 1, and all scores must use the same direction—for example, higher scores must consistently mean better performance. The method assumes that criteria are sufficiently clear and that scores and weights are reasonable. It does not create objectivity where judgements are arbitrary.

For example, a manager evaluates two suppliers using quality, cost and reliability. Suppose the weights are 0.40, 0.30 and 0.30. Supplier A scores 8, 6 and 7 out of 10; Supplier B scores 7, 9 and 8. The totals are:

• A: (0.40 × 8) + (0.30 × 6) + (0.30 × 7) = 7.1
• B: (0.40 × 7) + (0.30 × 9) + (0.30 × 8) = 7.9

Supplier B has the higher score, but the manager should still test whether its cost score reflects total cost, including delivery delays, defects and contract risks.

Bounded rationality

Herbert Simon’s concept of bounded rationality recognises that managers have limited time, information, attention and computational capacity. In complex settings, it is often impossible to identify and compare every alternative. Managers therefore “satisfice”: they seek an option that is good enough to meet important requirements rather than proving that it is the absolute optimum.

Bounded rationality is not carelessness. It is a realistic response to constraints. Good practice is to define minimum acceptable standards, identify the most consequential uncertainties, compare a manageable number of alternatives and establish a review point. A manager might select a supplier that meets quality, legal and delivery thresholds, even if a theoretically better supplier could exist but would require months of investigation.

Evidence-based management

Evidence-based management combines the best available evidence with professional expertise, stakeholder knowledge and organisational values. Evidence may include internal performance data, controlled trials, customer research, employee experience, professional standards and relevant research literature.

A practical evidence-based sequence is:

• ask a focused question;
• acquire relevant and credible evidence;
• appraise its quality and limitations;
• combine evidence with context and experience;
• act, document the rationale and evaluate results.

Evidence quality depends on validity, reliability, relevance, timeliness and freedom from systematic bias. A large dataset is not automatically good evidence. For instance, a customer survey with thousands of responses may still be misleading if dissatisfied customers were less likely to respond or if the questions were leading.

Participative decision-making

Participative decision-making involves affected employees, specialists, customers or other stakeholders in defining the problem, generating alternatives or evaluating consequences. Participation can improve information quality, acceptance and implementation. It can also be slower and may be dominated by confident or senior voices.

The appropriate level of participation depends on the importance of the decision, the manager’s information advantage, the need for commitment, time available and the consequences of error. A manager may decide alone in an emergency, consult specialists on a technical choice, or use a team process where implementation depends heavily on employee commitment.

Participation must be genuine. Asking for opinions after a decision has already been made damages trust. Useful methods include structured workshops, nominal-group techniques, anonymous surveys and “devil’s advocate” reviews. The manager should state who has decision authority, how input will be used and why some recommendations were not adopted.

Comparing decision models

These models are complementary rather than mutually exclusive. A manager can use participation to define criteria, evidence to compare alternatives, bounded rationality to set a search limit and rational analysis to document the final choice.

3. Problem-solving and root-cause analysis

Root-cause analysis (RCA) seeks to identify conditions that allow a problem to occur or recur. It should not be used to assign blame. A useful RCA distinguishes immediate causes, contributing factors and systemic causes.

The five-whys technique

The five-whys method repeatedly asks why an event occurred until the investigation reaches a controllable process or system condition. Five is not a compulsory number. The questioning should stop when the explanation is supported by evidence and is sufficiently actionable.

Illustrative chain:

• Why was a customer order late? The order waited two days for approval.
• Why did approval take two days? The designated approver was absent.
• Why was there no substitute? The process had no backup role.
• Why was no backup role defined? The procedure assumed one-person availability.
• Why was that assumption not reviewed? The process had no resilience review.

The final finding is not simply “the approver was absent”; it concerns process design and continuity controls. Investigators should verify each link rather than accepting a plausible story.

Fishbone analysis

A fishbone, or cause-and-effect, diagram groups possible causes under categories such as people, methods, machines or technology, materials, measurement and environment. The categories are prompts, not a universal law. Teams should adapt them to the context. The tool is especially useful during brainstorming, but each proposed cause requires evidence before it is treated as a root cause.

Pareto analysis

Pareto analysis ranks causes or categories by frequency, cost or impact. The familiar “80/20” idea is a heuristic, not a law. A small number of categories may account for a large share of harm, but the actual distribution must be measured. A Pareto chart can help managers focus limited resources on the most consequential contributors.

Process mapping

Process maps show activities, decisions, handoffs and delays. They reveal duplication, unclear ownership and control gaps. A process should be mapped as it actually operates, not merely as policy says it operates. Staff who perform the work are often essential sources of accurate process knowledge.

4. SWOT analysis and strategic diagnosis

SWOT classifies internal strengths and weaknesses and external opportunities and threats. Strengths and weaknesses concern the organisation’s resources, capabilities, processes and culture. Opportunities and threats arise from the environment, including competitors, regulation, technology, economic conditions and social change.

SWOT is a framing tool, not a complete strategy. Statements should be specific and supported by evidence. “Good reputation” is weak unless linked to evidence such as repeat-customer rates or independent ratings. Managers can develop options by matching strengths to opportunities, using opportunities to address weaknesses, using strengths to reduce threats and preparing safeguards for serious weaknesses and threats.

A common error is to list factors without prioritising them. A stronger analysis identifies the significance, likelihood and controllability of each factor and connects it to a decision.

5. Risk assessment under uncertainty

Risk is commonly understood as the effect of uncertainty on objectives. In practical management, risk assessment involves identifying hazards or events, estimating likelihood and consequence, evaluating existing controls and deciding on further treatment.

A simple expected-loss estimate is:

EL is expected loss, p is the probability of the event, expressed between 0 and 1, and I is the financial or other quantified impact if it occurs. This calculation assumes that probability and impact estimates are meaningful and that the consequences can reasonably be represented by one measure. It is not a prediction of the exact outcome; it is a prioritisation aid.

If a system failure has an estimated probability of 0.10 in a year and an impact of $40,000, then:

EL = 0.10 × $40,000 = $4,000 per year.

A control costing $1,500 annually may appear attractive, but the manager must also consider control effectiveness, implementation costs, secondary effects, legal duties and risks that are not easily monetised.

Risk matrices classify likelihood and impact, often as low, medium or high. They are easy to communicate but can create false precision. A “medium” rating may hide major disagreement or uncertainty. Managers should record assumptions, use ranges where appropriate, identify risk owners and define trigger points for escalation.

Risk responses include avoiding the activity, reducing likelihood, reducing impact, transferring or sharing risk through contracts or insurance, and accepting risk with informed approval. Risk acceptance should never be confused with ignoring risk.

6. Managerial analytics and data interpretation

Managerial analytics is the systematic use of data, models and interpretation to support decisions. It is often described in four forms:

• descriptive analytics: what happened;
• diagnostic analytics: why it may have happened;
• predictive analytics: what may happen; and
• prescriptive analytics: what action may be preferable.

The forms build on one another, but prediction and prescription require assumptions beyond historical description.

Basic measures

The mean is calculated as:

where xᵢ is each observation and n is the number of observations. The mean is sensitive to extreme values. The median is the middle value after ordering observations and is often more representative of a skewed distribution.

A percentage change is:

The old value must not be zero, and the result should be interpreted with attention to units, time period and measurement definition. If complaints rise from 40 to 50, the increase is ((50 − 40) ÷ 40) × 100% = 25%.

Managers should distinguish counts from rates. Ten incidents may be serious in a small operation but minor in a very large one. A rate uses a relevant denominator, such as incidents per 1,000 transactions. Comparisons also require consistent definitions and comparable periods.

Correlation indicates that two variables move together; it does not by itself establish causation. A third factor may influence both, the direction may be reversed, or the relationship may be coincidental. Before claiming that training reduced errors, a manager should consider seasonality, changes in staff composition, workload and reporting practices.

Data quality and bias

Useful questions include:

• Who collected the data and for what purpose?
• Are definitions consistent across teams and periods?
• Are important cases missing?
• Does the sample represent the population?
• Could incentives have changed reporting behaviour?
• Are privacy and consent requirements satisfied?

Common biases include confirmation bias, availability bias, anchoring, recency bias, survivorship bias and sunk-cost bias. Organisational bias can arise from targets that encourage under-reporting or from cultures that punish bad news. Countermeasures include asking for disconfirming evidence, using multiple sources, separating data collection from evaluation where feasible, recording assumptions and inviting independent review.

7. KPIs and dashboards

A KPI is a carefully selected measure linked to an important objective. A metric is any measure; not every metric deserves KPI status. Effective KPIs have a clear definition, owner, data source, frequency, target, tolerance and action rule.

A balanced performance set may include:

• outcome measures, such as customer retention;
• process measures, such as average response time;
• quality measures, such as defect rate;
• people measures, such as regretted turnover; and
• leading indicators, such as completion of preventive maintenance.

A dashboard should support a decision, not merely display attractive numbers. It should show trends, comparisons with targets, meaningful segmentation and exceptions requiring attention. Excessive indicators create noise. Each KPI should answer: “What decision will change if this value moves?”

Targets can create dysfunctional behaviour. If staff are measured only on speed, quality may decline. If sales are rewarded without regard to customer suitability, complaints may rise. Managers should use a balanced set of measures and review whether indicators remain valid as circumstances change.

8. A practical decision cycle

A disciplined managerial cycle can integrate the approaches in this chapter:

1. Frame the decision and define success.
2. Identify stakeholders, constraints and ethical obligations.
3. Collect and validate relevant evidence.
4. Diagnose causes and distinguish symptoms from system conditions.
5. Generate alternatives, including a “do nothing” baseline.
6. Assess benefits, costs, risks and implementation requirements.
7. Consult appropriate stakeholders and test assumptions.
8. Decide, document reasons and assign accountability.
9. Implement with milestones and controls.
10. Monitor KPIs, review outcomes and revise the decision when evidence changes.

This is a cycle rather than a rigid staircase. New evidence may require returning to problem definition or diagnosis.

Illustrative case: reducing missed appointments

Background

A fictional community health centre has 1,200 scheduled appointments each month. Missed appointments increased from 6% to 10% over four months. The centre loses capacity, patients wait longer for service and staff spend time contacting absent patients.

Problem and analysis

The manager initially considers blaming patients or imposing a penalty. A process map shows that appointments are often booked six weeks in advance, reminders are sent only by email and cancellation procedures are unclear. A review finds that 35% of patients have not opened the email reminder. Staff interviews reveal that some patients work irregular hours and cannot easily telephone during office hours.

The manager uses a simple cause analysis. Contributing factors include long lead time, unsuitable reminder channels, unclear cancellation options and incomplete contact preferences. The data is segmented by appointment type and communication preference rather than assuming all patients behave similarly.

A risk assessment identifies potential unintended consequences. A strict penalty could disproportionately affect low-income patients and discourage necessary care. Privacy rules also require careful handling of contact information.

Approach and solution

The centre pilots a multi-channel reminder process for two months: text messages where consent exists, email reminders and a simple cancellation link. Patients are asked at booking which channel they prefer. The pilot uses a comparable service group where the previous process remains temporarily in place, while recognising that the groups may not be perfectly equivalent.

The manager defines KPIs: missed-appointment rate, cancellation notice rate, appointment utilisation and complaints about reminders. A process owner reviews results weekly. Staff are invited to identify operational problems, and patients are consulted about accessibility.

Suppose the pilot group records 90 missed appointments out of 1,000 appointments, while the previous process recorded 100 missed appointments out of 1,000. The missed-appointment rate falls from 10% to 9%, a relative reduction of 10% but an absolute reduction of 1 percentage point. The manager should avoid claiming that the intervention alone caused the change without considering other differences and random variation. Nevertheless, the result supports a wider, monitored implementation if costs and equity effects are acceptable.

Outcome and learning points

The centre expands the reminder process, retains non-digital alternatives and reviews results quarterly. The case demonstrates that a visible problem may have several process causes; participation improves practical design; rates are more informative than counts; and ethical analysis must consider access and unintended effects. It also shows why monitoring after implementation is part of decision-making, not an optional final step.

9. Ethical, governance and technology considerations

Managers are accountable for decisions even when software or consultants provide recommendations. Data should be collected and used lawfully, fairly and proportionately. Personal information requires appropriate access controls, retention practices and transparency. Sensitive decisions—such as recruitment, discipline, credit or service eligibility—require particular care because historical data may reproduce past discrimination.

Artificial intelligence can support forecasting, anomaly detection, summarisation and scenario analysis. It can also generate inaccurate outputs, obscure assumptions, amplify biased data and create confidentiality risks. Human review should be proportionate to the consequences of error. Managers should verify sources, test outputs across relevant groups, document the tool’s role and provide a route for challenge or correction. Automation should support accountable judgement rather than become an excuse to avoid responsibility.

10. Limitations and best practices

No decision framework removes uncertainty. Data may be incomplete, objectives may conflict, stakeholders may have unequal influence and outcomes may take time to appear. Quantitative scores can give an illusion of precision, while group discussion can create conformity. Managers should therefore:

• define terms and measures before reviewing results;
• triangulate data from different sources;
• distinguish fact, assumption and judgement;
• use ranges and scenarios when exact estimates are unreliable;
• invite disagreement and search for disconfirming evidence;
• consider distributional and ethical effects, not only averages;
• assign decision ownership and review dates; and
• learn from outcomes without using hindsight unfairly.

A sound decision is not necessarily one that produces a favourable result, because chance and external events matter. It is one whose reasoning was proportionate, evidence was responsibly assessed, risks were considered and learning was captured.

Chapter Summary

Decision-making is the selection of a course of action, while problem-solving includes diagnosis, intervention and evaluation. The rational model provides structure; bounded rationality recognises limits; evidence-based management connects action to credible information; and participative decision-making improves knowledge and commitment when used appropriately. Root-cause analysis helps managers move beyond symptoms, while SWOT frames internal and external conditions. Risk assessment supports prioritisation but should not create false precision. Managerial analytics requires careful interpretation of averages, rates, percentages, trends and correlations. KPIs and dashboards are valuable only when linked to objectives and decisions. Bias, data quality, privacy, equity and accountability must be managed throughout the process. Effective managers combine analysis with judgement, consultation, ethical awareness and continuous review.

Key Terms

Decision-making: Choosing among alternatives and committing to a course of action.

Problem-solving: A systematic process of defining, analysing and addressing an undesirable condition.

Rational model: A structured approach based on objectives, alternatives, criteria and evaluation.

Bounded rationality: The recognition that decisions are limited by information, time and cognitive capacity.

Satisficing: Selecting an option that meets important requirements rather than proving it optimal.

Evidence-based management: Combining credible evidence, professional judgement, stakeholder knowledge and values.

Participative decision-making: Involving relevant stakeholders in defining, analysing or choosing a response.

Root-cause analysis: Investigation of underlying conditions that allow a problem to occur or recur.

Five whys: A questioning method used to explore successive causes of an event.

SWOT analysis: Classification of strengths, weaknesses, opportunities and threats.

Risk: The effect of uncertainty on objectives.

Expected loss: An estimated probability multiplied by the associated impact.

Analytics: The use of data and interpretation to support understanding and action.

Correlation: A statistical association between variables that does not by itself establish causation.

Key performance indicator: A selected measure linked to an important objective.

Dashboard: A visual display of relevant measures, trends and exceptions.

Cognitive bias: A systematic pattern of judgement that can distort reasoning.

Triangulation: Comparing multiple sources or methods to strengthen understanding.

Review Questions

Short Answer Questions

1. How does problem-solving differ from decision-making?
2. What is bounded rationality, and why is it relevant to managers?
3. State two conditions of a useful KPI.
4. Why should managers distinguish counts from rates?
5. What is the main purpose of root-cause analysis?

Long Answer Questions

1. Compare the rational, evidence-based and participative decision models, including one limitation of each.
2. Explain how a manager could use five-whys analysis and process mapping to investigate recurring service failures.
3. Discuss how SWOT analysis can support, but cannot replace, strategic judgement.
4. Explain the role of data quality, correlation and bias in managerial analytics.
5. Describe how a dashboard can encourage either effective performance management or dysfunctional behaviour.

Analytical / Application Questions

1. A department’s complaints increased from 80 to 100 while transactions increased from 4,000 to 6,000. Calculate the percentage change in complaints and compare complaint rates per 1,000 transactions. What conclusion should the manager draw?
2. A manager must choose between two suppliers using cost, quality and delivery reliability. Design a weighted decision matrix and explain how you would test whether the result is sensitive to changes in the weights.
3. A proposed control costs $2,000 per year and is expected to reduce a 0.15-probability annual loss of $30,000 by half. Estimate the avoided expected loss and discuss at least two non-financial factors before recommending the control.

Multiple Choice Questions with Answers

Question 1

Which decision model explicitly recognises limits in time, information and cognitive capacity?

A. Rational model

B. Bounded-rationality model

C. Participative model

D. SWOT model

Correct Answer: B. Bounded-rationality model

Question 2

In a weighted decision matrix, what must be true for scores to be combined meaningfully?

A. Every criterion must have the same weight

B. All scores must use a consistent direction and scale

C. Only financial criteria may be included

D. The highest raw score must always determine the decision

Correct Answer: B. All scores must use a consistent direction and scale

Question 3

A risk has probability 0.20 and impact of $10,000. What is its estimated expected loss?

A. $200

B. $500

C. $2,000

D. $10,200

Correct Answer: C. $2,000

Question 4

Which statement about correlation is correct?

A. It always proves causation

B. It shows that one variable causes another

C. It indicates association but does not by itself prove causation

D. It can only be used with financial data

Correct Answer: C. It indicates association but does not by itself prove causation

Question 5

Which SWOT element is normally external to the organisation?

A. Strength

B. Weakness

C. Opportunity

D. Internal capability

Correct Answer: C. Opportunity

Question 6

A complaint count rises from 40 to 50. What is the percentage increase?

A. 10%

B. 20%

C. 25%

D. 50%

Correct Answer: C. 25%

Question 7

What is the primary purpose of a KPI?

A. To display as many measurements as possible

B. To link an important objective to a meaningful measure

C. To eliminate managerial judgement

D. To guarantee improved performance

Correct Answer: B. To link an important objective to a meaningful measure

Question 8

Which action best reduces confirmation bias?

A. Considering only information supporting the initial view

B. Ignoring conflicting evidence until implementation

C. Actively seeking credible disconfirming evidence

D. Allowing the most senior person to decide without review

Correct Answer: C. Actively seeking credible disconfirming evidence

Question 9

What is a key ethical concern when using AI for high-consequence decisions?

A. It always produces transparent reasoning

B. It may reproduce bias in historical data

C. It removes the need for accountability

D. It cannot process any organisational data

Correct Answer: B. It may reproduce bias in historical data

Question 10

Why is a rate often more useful than a raw count when comparing performance?

A. A rate removes the need for accurate data

B. A rate accounts for the relevant size or volume of activity

C. A rate always proves causation

D. A rate cannot be affected by measurement definitions

Correct Answer: B. A rate accounts for the relevant size or volume of activity