Problem Formulation in Data Science
Understanding Domain Problems, Problem Statements, and
Hypotheses
Department of Statistics
University of Lagos
First Semester 2025
Last updated: December 10, 2025
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 1 / 17
Contents
1 Understanding the Domain Problem
2 Problem Statement Formulation
3 Hypothesis Formulation
4 The Integrated Process
5 Tools and Templates
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 2 / 17
The Foundation of Any Data Science Project
Key Principle
No amount of sophisticated analytics can compensate for a poorly
understood problem.
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 3 / 17
What is a Domain Problem?
A real-world challenge within a specific field that:
Has measurable impact
Can be addressed through data-driven insights
Exists within constraints (technical, ethical, regulatory, budgetary)
Examples:
Healthcare: Reduce patient readmission rates by 15% within 12
months
Retail: Increase customer lifetime value by identifying high-value
segments
Manufacturing: Predict equipment failure to reduce downtime
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 4 / 17
The Critical First Step: Problem Understanding
Question Purpose
Who Identify stakeholders
What Define problem clearly
Why Quantify impact
Where Determine context
When Identify temporal patterns
How Understand current solutions
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 5 / 17
Characteristics of a Well-Formulated Problem Statement
Follows the SMART framework:
Specific: Clear and unambiguous
Measurable: Quantifiable success criteria
Achievable: Realistic given constraints
Relevant: Aligns with business objectives
Time-bound: Includes timeframe considerations
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 6 / 17
Problem Statement Template
Template
In the context of [DOMAIN], we are facing [SPECIFIC PROBLEM] which
affects [STAKEHOLDERS] resulting in [QUANTIFIED IMPACT]. A
successful solution would [DESIRED OUTCOME] measured by [SUCCESS
METRICS] within [TIMEFRAME].
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 7 / 17
Example Transformations
Poor Formulation Improved Formulation
We need to improve sales Q3 sales in Northeast decreased
by 18% due to churn in 25–35 age
group
Make our website better Mobile users have 65% higher
bounce rate and 40% shorter ses-
sions
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 8 / 17
Types of Hypotheses
1 Research Hypotheses:
Hr : “Customer satisfaction affects retention rates”
2 Statistical Hypotheses:
H0 : “No correlation between satisfaction and retention”
H1 : “Higher satisfaction correlates with higher retention”
3 Machine Learning Hypotheses:
Hml : “Feature X improves prediction accuracy by at least 5%”
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 9 / 17
Hypothesis Development Framework
Observation → Question → Literature Review → Hypothesis
Formulation
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 10 / 17
Hypothesis Prioritization: ICE Scoring
Impact: Expected improvement over baseline
Confidence: Evidence strength
Ease: Feasibility and cost of testing
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 11 / 17
Data Science Process Flow
1 Domain Understanding
2 Problem Statement Formulation
3 Hypothesis Generation
4 Data Requirements
5 Experimental Design
6 Execution & Validation
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 12 / 17
Case Study: E-commerce Platform
Problem Statement
“Our e-commerce platform has a conversion rate 40% below industry
average, especially for mobile users. We need to identify and address the
top 3 barriers to increase mobile conversion to at least 2.0% within 6
months.”
Hypotheses:
1 Mobile checkout has too many steps
2 Slow page load times reduce conversion
3 Difficulty applying discount codes
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 13 / 17
Hypothesis Testing Matrix
Hypothesis Test Design Data Required Success Criteria
Mobile checkout steps A/B Test User logs, funnel data p < 0.05, lift > 5%
Page load times Correlation analysis Performance logs r > 0.3
Discount codes Usability study Survey data, recordings 90% success rate
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 14 / 17
Summary: Key Takeaways
1 Domain understanding comes first
2 Problem statements should be SMART
3 Hypotheses must be testable and falsifiable
4 Iterate and refine continuously
5 Document assumptions and decisions
“The most important step in data science isn’t building mod-
els—it’s asking the right questions.”
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 15 / 17
References
CRISP-DM: Cross-Industry Standard Process for Data Mining
OSEMN Framework: Obtain, Scrub, Explore, Model, Interpret
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 16 / 17
Thank You
First Semester 2025 Last updated: Decemb
Department of Statistics University of Lagos Problem Formulation in Data Science 17 / 17