Learning Objectives
By the end of this lesson, learners should be able to:
- Define diagnostic analytics.
- Explain the purpose of root-cause analysis.
- Distinguish symptoms from underlying causes.
- Apply structured approaches to business problem investigation.
- Use segmentation and drill-down techniques in diagnosis.
- Explain the role of correlation and comparative evidence.
- Evaluate competing explanations.
- Develop evidence-based diagnostic conclusions.
1. Meaning of Diagnostic Analytics
Diagnostic analytics is the systematic examination of data to identify relationships, contributing factors and possible explanations for observed business outcomes.
It moves beyond:
What happened?
toward:
Why did it happen?
For example:
Descriptive finding:
Customer cancellations increased by 14%.
Diagnostic investigation:
The increase was concentrated among customers using a particular service package and coincided with a deterioration in service response times.
The second statement provides a possible explanation that can be investigated further.
2. Root-Cause Analysis
Root-cause analysis (RCA) is a structured approach to identifying underlying factors responsible for a problem.
A root cause differs from a symptom.
Example
Symptom:
Customer complaints increased.
Immediate factor:
Service response times increased.
Potential underlying cause:
The organization experienced a shortage of trained support staff following rapid customer growth.
A strong diagnostic process attempts to move beyond visible symptoms.
3. Symptoms Versus Causes
Consider:
Sales declined.
This is an outcome, not necessarily a cause.
Further investigation might reveal:
Sales decline → Lower order volume → Customer losses → Increased cancellations → Service quality deterioration
Each stage may represent a contributing factor rather than the final root cause.
Analysts should therefore avoid stopping at the first plausible explanation.
4. The Five Whys Technique
The Five Whys approach repeatedly asks “why?” to move from a visible problem toward underlying causes.
Example:
Problem: Orders are being delivered late.
Why?
Because dispatch is delayed.
Why?
Because orders remain in the processing queue.
Why?
Because processing capacity is insufficient.
Why?
Because staffing levels were based on outdated demand assumptions.
Why?
Because workforce planning did not incorporate recent demand growth.
The objective is not necessarily to ask exactly five questions. The principle is to continue investigating until the underlying contributing factors become clearer.
5. Fishbone Analysis
A fishbone diagram, also called an Ishikawa diagram, organizes potential causes into categories.
Common categories may include:
- People.
- Processes.
- Technology.
- Materials.
- Measurement.
- Environment.
The technique encourages teams to consider multiple possible explanations instead of immediately selecting one.
6. Drill-Down Analysis
Diagnostic analytics frequently begins with a high-level metric and progressively examines more detailed dimensions.
Example:
Revenue decline
↓
Region
↓
Country
↓
Branch
↓
Product
↓
Customer segment
↓
Transaction
Drill-down analysis helps identify where the change is concentrated.
7. Segmentation in Diagnosis
Segmentation can reveal relationships hidden by aggregate figures.
For example, overall customer satisfaction may decline only among:
- New customers.
- Mobile users.
- Customers in one region.
- Customers using a specific product.
Such patterns can substantially narrow the diagnostic investigation.
8. Comparative Analysis
Comparisons can help identify unusual patterns.
An analyst might compare:
- Current versus previous period.
- Actual versus budget.
- Branch versus branch.
- Product versus product.
- Customer segment versus customer segment.
- Organization versus industry benchmark.
Comparative analysis provides context for determining whether a result is unusual.
9. Correlation Analysis
Correlation measures the degree to which variables move together.
For example, an analyst might investigate the relationship between:
- Advertising expenditure and sales.
- Delivery time and customer satisfaction.
- Employee absence and productivity.
However, correlation alone does not establish causality.
A strong diagnostic investigation should consider:
- Alternative explanations.
- Confounding variables.
- Timing.
- Data quality.
- Business context.
10. Variance Analysis
Variance analysis compares actual results against a benchmark.
Examples include:
Actual sales − Budgeted sales
or
Actual cost − Standard cost
Significant variances can identify areas requiring investigation.
However, a variance is a signal for investigation rather than automatically a cause.
11. Hypothesis-Based Diagnosis
A disciplined analyst should develop competing hypotheses.
For example:
Problem: Customer churn increased.
Possible explanations:
- Service quality declined.
- Prices increased.
- Competitors introduced stronger offers.
- Customer demographics changed.
- A system problem affected service delivery.
The analyst should then evaluate evidence supporting or contradicting each hypothesis.
12. Avoiding Confirmation Bias
Confirmation bias occurs when analysts favor evidence that supports an existing belief while overlooking contradictory information.
For example, if management believes that high prices caused customer losses, analysts may focus exclusively on pricing data.
A stronger process deliberately investigates alternative explanations.
13. Correlation Does Not Equal Causation
Suppose a company observes:
Employees who receive more training have higher productivity.
Possible explanations include:
- Training improves productivity.
- More capable employees are more likely to receive training.
- Managers assign training to high-performing departments.
- Another variable affects both training and productivity.
Therefore, the observed relationship requires careful interpretation.
14. Diagnostic Analytics Workflow
A practical workflow is:
Step 1: Define the problem
Clearly specify the outcome requiring investigation.
Step 2: Quantify the problem
Determine its magnitude.
Step 3: Identify when and where it occurred
Analyze time and relevant segments.
Step 4: Generate hypotheses
Develop plausible explanations.
Step 5: Gather evidence
Use relevant internal and external data.
Step 6: Test competing explanations
Determine which explanations are most consistent with the evidence.
Step 7: Identify contributing factors
Distinguish major drivers from incidental relationships.
Step 8: Communicate the diagnosis
Present evidence, limitations and recommended areas for action.
15. Diagnostic Analytics and Decision-Making
The purpose of diagnosis is not simply to produce an explanation.
A useful diagnosis should help management determine:
- What requires attention?
- Where should resources be allocated?
- What additional evidence is required?
- Which interventions should be tested?
- What risks should be monitored?
The strongest diagnostic analysis connects evidence to managerial action.
16. Limitations of Root-Cause Analysis
Root-cause analysis can be affected by:
- Incomplete data.
- Incorrect assumptions.
- Complex causal relationships.
- Multiple simultaneous causes.
- Measurement errors.
- Organizational bias.
Some business problems do not have a single root cause.
A more realistic conclusion may identify several interacting contributing factors.
Lesson Summary
Diagnostic analytics investigates why observed outcomes occurred.
Important techniques include:
- Root-cause analysis.
- Five Whys.
- Fishbone analysis.
- Drill-down analysis.
- Segmentation.
- Comparative analysis.
- Variance analysis.
- Correlation analysis.
- Hypothesis testing.
A strong analyst distinguishes symptoms from causes, investigates competing explanations and avoids treating correlation as proof of causation.
The objective is not merely to find an explanation, but to develop an evidence-based understanding that improves decision-making.
References
- ASQ — Root Cause Analysis
American Society for Quality - NIST/SEMATECH — Statistical Methods
NIST Statistical Methods Handbook - IBM — Business Analytics
IBM Analytics
Review Questions
- What is diagnostic analytics?
- How does diagnostic analytics differ from descriptive analytics?
- What is root-cause analysis?
- What is the difference between a symptom and a cause?
- How does the Five Whys technique support diagnosis?
- What is a fishbone diagram?
- How does drill-down analysis help identify potential causes?
- Why is segmentation important in diagnostic analytics?
- How can variance analysis identify areas requiring investigation?
- Why does correlation not establish causation?
- What is confirmation bias and why is it dangerous in analytics?
- Why might a business problem have several contributing causes rather than one root cause?