Exploratory Data Analysis
5 minutes
5 Questions
In the PMI-CPMAI framework, Exploratory Data Analysis (EDA) is a core activity in the Identify Data Needs domain, which aligns with Phase II (Data Understanding) of the CPMAI methodology. After the business problem and AI pattern are defined in Phase I (Business Understanding), EDA helps the project team examine available data and decide whether it can realistically support the intended AI solution. It works as a reality check before significant time and money go into data preparation and model development. EDA is the systematic investigation of datasets to understand their structure, content, quality, and relationships. Key activities include profiling data types, volumes, and sources; calculating summary statistics such as mean, median, variance, and distributions; and identifying missing values, duplicates, outliers, and inconsistencies. Teams also explore correlations and patterns between variables, check class balance for classification problems, and use visualizations such as histograms, scatter plots, box plots, and heat maps. From a CPMAI perspective, EDA directly informs go/no-go decisions. It shows whether the data has enough quantity, quality, relevance, and representativeness to meet the business objectives. It also exposes gaps that may require new data collection, third-party data acquisition, or labeling efforts. EDA reveals potential bias, privacy concerns, and compliance risks that matter for trustworthy and responsible AI. Its findings help the team estimate the effort needed for data cleansing, transformation, and feature engineering in Phase III (Data Preparation). EDA also supports the iterative, data-centric nature of CPMAI. If the analysis shows that the data cannot support the original goal, the team should loop back to Business Understanding to refine scope, adjust the AI pattern, or reset expectations with stakeholders. This avoids a common cause of AI project failure, which is building on poor or insufficient data. In short, EDA turns raw data into actionable insight about data readiness. It reduces project risk and ensures that data needs are clearly identified, documented, and aligned with business value before modeling begins.
Exploratory Data Analysis (EDA) in PMI-CPMAI: Identify Data Needs
Exploratory Data Analysis (EDA): A Complete Guide for the PMI-CPMAI Exam
1. Introduction
Exploratory Data Analysis (EDA) is a core activity within the Identify Data Needs domain (Phase II: Data Understanding) of the PMI Cognitive Project Management for AI (CPMAI) methodology. CPMAI is an iterative, data-centric approach built on CRISP-DM. It has six phases: Business Understanding, Data Understanding, Data Preparation, Model Development, Model Evaluation, and Model Operationalization. EDA sits in Data Understanding. This is where the team first examines the data in depth to learn whether it can support the AI solution defined during Business Understanding.
2. What Is Exploratory Data Analysis?
EDA is the process of investigating a dataset to:
- summarize its main characteristics
- uncover patterns, trends, and relationships
- detect anomalies, outliers, and errors
- test assumptions
- assess data quality and fitness for purpose
The team does this through statistical summaries and visualizations, before formal modeling begins. The term was popularized by statistician John Tukey. He stressed letting the data "speak" before imposing hypotheses or models.
In CPMAI terms, EDA answers a key question: Do we have the right data, in the right quantity and quality, to solve the business problem we defined?
3. Why EDA Is Important
a) Most AI projects fail because of data. CPMAI stresses that AI projects are data projects first. Many fail because data is insufficient, poor in quality, inaccessible, or not representative. EDA surfaces these issues early, while they are still cheap to fix.
b) It validates feasibility. EDA confirms or challenges the assumptions made in Business Understanding. If the data cannot support the chosen AI pattern (for example, predictive analytics), the team may need to go back to Phase I. This is the CPMAI "go/no-go" mindset.
c) It reduces risk and cost. Finding data problems after model development is expensive. EDA is a risk-mitigation activity.
d) It informs Data Preparation. Findings drive cleansing, transformation, feature engineering, imputation, labeling, and augmentation in Phase III.
e) It reveals bias and ethical concerns. EDA can expose under-represented groups, skewed distributions, and proxy variables for protected attributes. This supports trustworthy and responsible AI.
f) It guides model selection. Data types, distributions, dimensionality, and relationships help decide which algorithms and approaches are appropriate.
g) It improves stakeholder communication. Visual summaries help business stakeholders understand what the data can and cannot do.
4. How EDA Works: Key Steps and Techniques
Step 1: Understand the data sources and context
- Identify where the data comes from (internal, external, third-party, streaming, batch).
- Review metadata, data dictionaries, and collection methods.
- Understand ownership, access rights, privacy, and compliance constraints.
Step 2: Profile the data structure
- Count records (rows) and features (columns).
- Identify data types: numerical, categorical, ordinal, text, image, time-series.
- Classify the data as structured, semi-structured, or unstructured.
Step 3: Univariate analysis (one variable at a time)
- Measures of central tendency: mean, median, mode.
- Measures of spread: range, variance, standard deviation, interquartile range.
- Distribution shape: skewness, kurtosis, normality.
- Visuals: histograms, box plots, bar charts, frequency tables.
Step 4: Bivariate and multivariate analysis
- Correlation analysis (Pearson, Spearman), correlation matrices, and heatmaps.
- Scatter plots, pair plots, cross-tabulations.
- Relationships between features and the target variable.
- Multicollinearity detection.
Step 5: Assess data quality
- Completeness: missing values and null patterns.
- Accuracy: incorrect or implausible values.
- Consistency: conflicting formats and units.
- Timeliness: how current the data is.
- Uniqueness: duplicate records.
- Validity: conformance to business rules.
Step 6: Detect outliers and anomalies
- Use z-scores, the IQR method, and visual inspection.
- Decide whether each outlier is an error, noise, or a meaningful signal (for example, fraud).
Step 7: Check class balance and representativeness
- For classification, check whether the target classes are imbalanced.
- Check whether the sample represents the real-world population the model will serve.
Step 8: Assess data quantity and labeling
- Is there enough data for the chosen approach? Deep learning usually needs far more data.
- For supervised learning, are labels available, accurate, and consistent?
Step 9: Document findings and recommendations
- Produce a data quality report and an EDA summary.
- Feed the results into the go/no-go decision and the Data Preparation plan.
5. Common EDA Tools
- Python (pandas, NumPy, Matplotlib, Seaborn, pandas-profiling/ydata-profiling)
- R (ggplot2, dplyr)
- BI tools (Tableau, Power BI)
- SQL for querying and aggregation
- Automated data profiling tools
The CPMAI exam is methodology-focused, not tool-focused. Know the purpose of the tools rather than their syntax.
6. EDA Within the CPMAI Iterative Approach
- EDA is iterative, not a one-time event. New data sources or refined requirements trigger more exploration.
- EDA findings may send the project back to Business Understanding to rescope, or forward to Data Preparation.
- CPMAI encourages small, quick iterations: "think big, start small, iterate often".
- EDA supports the Phase II go/no-go checkpoint: is the data sufficient to proceed?
7. EDA vs. Related Concepts
- EDA vs. Data Preparation: EDA discovers and diagnoses. Data Preparation fixes and transforms. EDA comes first.
- EDA vs. Confirmatory Data Analysis: EDA is open-ended discovery. Confirmatory analysis tests specific hypotheses formally.
- EDA vs. Model Evaluation: EDA examines the data. Model Evaluation examines model performance.
- EDA vs. Data Collection: Collection gathers the data. EDA examines what was gathered.
8. Exam Tips: Answering Questions on Exploratory Data Analysis
Tip 1: Know where EDA lives. EDA belongs to Phase II: Data Understanding. If a question asks which phase involves profiling data, identifying quality issues, or visualizing distributions, the answer is Data Understanding, not Data Preparation.
Tip 2: Diagnosis before treatment. If the scenario says the team has just acquired data, the best next step is usually to explore and profile it. Cleaning, feature engineering, and model training come later. Watch for distractors that jump ahead.
Tip 3: Data issues mean reassess, not ignore. If EDA reveals insufficient, biased, or poor-quality data, the CPMAI answer is usually to raise it with stakeholders. Options include acquiring more data, revisiting the business problem, or making a go/no-go decision. Avoid answers that proceed to modeling and hope for the best.
Tip 4: Link EDA to business objectives. EDA is not exploration for its own sake. The best answers connect findings back to the defined business problem and success criteria.
Tip 5: Recognize bias and ethics cues. Scenarios about under-represented groups, skewed samples, or sensitive attributes point to EDA as the place to detect bias. The correct response often involves fairness checks, representativeness analysis, or Trustworthy AI considerations.
Tip 6: Class imbalance and outliers. If a question mentions rare events (fraud, defects, disease), think class imbalance found during EDA. For outliers, the best answer investigates whether they are errors or meaningful signals rather than deleting them automatically.
Tip 7: Remember iteration. CPMAI is agile and iterative. Answers suggesting EDA is done once and never revisited are usually wrong.
Tip 8: Know the data quality dimensions. Be ready to match a symptom to its dimension:
- nulls point to completeness
- duplicates point to uniqueness
- conflicting formats point to consistency
- stale data points to timeliness
Tip 9: Choose the most comprehensive, proactive answer. PMI exams favor answers that are proactive, collaborative (involving data owners and SMEs), and documented. A strong choice might be "perform data profiling and document findings in a data quality report shared with stakeholders".
Tip 10: Watch keywords. Words such as profile, summarize, visualize, distribution, correlation, missing values, outliers, anomalies, patterns, data quality assessment, sufficiency signal EDA. Words such as clean, transform, normalize, encode, impute, augment signal Data Preparation.
Tip 11: Use elimination. Remove options that:
- skip data understanding
- treat technical tools as the goal
- ignore stakeholders
- disregard data quality or ethics
9. Sample Exam-Style Question
An AI project team has collected three years of customer transaction data to build a churn prediction model. Before beginning model development, what should the project manager ensure happens next?
A. Train a baseline model to measure accuracy
B. Normalize all numerical features
C. Conduct exploratory data analysis to assess data quality, distributions, and suitability for the business objective
D. Deploy a pilot to production
Answer: C. EDA in the Data Understanding phase must come before preparation (B), modeling (A), and operationalization (D).
10. Key Takeaways
- EDA is the investigative heart of CPMAI Phase II: Data Understanding.
- It uncovers quality issues, patterns, bias, and gaps before money is spent on modeling.
- It directly informs Data Preparation and the go/no-go decision.
- It is iterative and tied to business objectives.
- On the exam, choose answers that explore and diagnose data first, involve stakeholders, address bias, and respect the CPMAI phase sequence.
1. Introduction
Exploratory Data Analysis (EDA) is a core activity within the Identify Data Needs domain (Phase II: Data Understanding) of the PMI Cognitive Project Management for AI (CPMAI) methodology. CPMAI is an iterative, data-centric approach built on CRISP-DM. It has six phases: Business Understanding, Data Understanding, Data Preparation, Model Development, Model Evaluation, and Model Operationalization. EDA sits in Data Understanding. This is where the team first examines the data in depth to learn whether it can support the AI solution defined during Business Understanding.
2. What Is Exploratory Data Analysis?
EDA is the process of investigating a dataset to:
- summarize its main characteristics
- uncover patterns, trends, and relationships
- detect anomalies, outliers, and errors
- test assumptions
- assess data quality and fitness for purpose
The team does this through statistical summaries and visualizations, before formal modeling begins. The term was popularized by statistician John Tukey. He stressed letting the data "speak" before imposing hypotheses or models.
In CPMAI terms, EDA answers a key question: Do we have the right data, in the right quantity and quality, to solve the business problem we defined?
3. Why EDA Is Important
a) Most AI projects fail because of data. CPMAI stresses that AI projects are data projects first. Many fail because data is insufficient, poor in quality, inaccessible, or not representative. EDA surfaces these issues early, while they are still cheap to fix.
b) It validates feasibility. EDA confirms or challenges the assumptions made in Business Understanding. If the data cannot support the chosen AI pattern (for example, predictive analytics), the team may need to go back to Phase I. This is the CPMAI "go/no-go" mindset.
c) It reduces risk and cost. Finding data problems after model development is expensive. EDA is a risk-mitigation activity.
d) It informs Data Preparation. Findings drive cleansing, transformation, feature engineering, imputation, labeling, and augmentation in Phase III.
e) It reveals bias and ethical concerns. EDA can expose under-represented groups, skewed distributions, and proxy variables for protected attributes. This supports trustworthy and responsible AI.
f) It guides model selection. Data types, distributions, dimensionality, and relationships help decide which algorithms and approaches are appropriate.
g) It improves stakeholder communication. Visual summaries help business stakeholders understand what the data can and cannot do.
4. How EDA Works: Key Steps and Techniques
Step 1: Understand the data sources and context
- Identify where the data comes from (internal, external, third-party, streaming, batch).
- Review metadata, data dictionaries, and collection methods.
- Understand ownership, access rights, privacy, and compliance constraints.
Step 2: Profile the data structure
- Count records (rows) and features (columns).
- Identify data types: numerical, categorical, ordinal, text, image, time-series.
- Classify the data as structured, semi-structured, or unstructured.
Step 3: Univariate analysis (one variable at a time)
- Measures of central tendency: mean, median, mode.
- Measures of spread: range, variance, standard deviation, interquartile range.
- Distribution shape: skewness, kurtosis, normality.
- Visuals: histograms, box plots, bar charts, frequency tables.
Step 4: Bivariate and multivariate analysis
- Correlation analysis (Pearson, Spearman), correlation matrices, and heatmaps.
- Scatter plots, pair plots, cross-tabulations.
- Relationships between features and the target variable.
- Multicollinearity detection.
Step 5: Assess data quality
- Completeness: missing values and null patterns.
- Accuracy: incorrect or implausible values.
- Consistency: conflicting formats and units.
- Timeliness: how current the data is.
- Uniqueness: duplicate records.
- Validity: conformance to business rules.
Step 6: Detect outliers and anomalies
- Use z-scores, the IQR method, and visual inspection.
- Decide whether each outlier is an error, noise, or a meaningful signal (for example, fraud).
Step 7: Check class balance and representativeness
- For classification, check whether the target classes are imbalanced.
- Check whether the sample represents the real-world population the model will serve.
Step 8: Assess data quantity and labeling
- Is there enough data for the chosen approach? Deep learning usually needs far more data.
- For supervised learning, are labels available, accurate, and consistent?
Step 9: Document findings and recommendations
- Produce a data quality report and an EDA summary.
- Feed the results into the go/no-go decision and the Data Preparation plan.
5. Common EDA Tools
- Python (pandas, NumPy, Matplotlib, Seaborn, pandas-profiling/ydata-profiling)
- R (ggplot2, dplyr)
- BI tools (Tableau, Power BI)
- SQL for querying and aggregation
- Automated data profiling tools
The CPMAI exam is methodology-focused, not tool-focused. Know the purpose of the tools rather than their syntax.
6. EDA Within the CPMAI Iterative Approach
- EDA is iterative, not a one-time event. New data sources or refined requirements trigger more exploration.
- EDA findings may send the project back to Business Understanding to rescope, or forward to Data Preparation.
- CPMAI encourages small, quick iterations: "think big, start small, iterate often".
- EDA supports the Phase II go/no-go checkpoint: is the data sufficient to proceed?
7. EDA vs. Related Concepts
- EDA vs. Data Preparation: EDA discovers and diagnoses. Data Preparation fixes and transforms. EDA comes first.
- EDA vs. Confirmatory Data Analysis: EDA is open-ended discovery. Confirmatory analysis tests specific hypotheses formally.
- EDA vs. Model Evaluation: EDA examines the data. Model Evaluation examines model performance.
- EDA vs. Data Collection: Collection gathers the data. EDA examines what was gathered.
8. Exam Tips: Answering Questions on Exploratory Data Analysis
Tip 1: Know where EDA lives. EDA belongs to Phase II: Data Understanding. If a question asks which phase involves profiling data, identifying quality issues, or visualizing distributions, the answer is Data Understanding, not Data Preparation.
Tip 2: Diagnosis before treatment. If the scenario says the team has just acquired data, the best next step is usually to explore and profile it. Cleaning, feature engineering, and model training come later. Watch for distractors that jump ahead.
Tip 3: Data issues mean reassess, not ignore. If EDA reveals insufficient, biased, or poor-quality data, the CPMAI answer is usually to raise it with stakeholders. Options include acquiring more data, revisiting the business problem, or making a go/no-go decision. Avoid answers that proceed to modeling and hope for the best.
Tip 4: Link EDA to business objectives. EDA is not exploration for its own sake. The best answers connect findings back to the defined business problem and success criteria.
Tip 5: Recognize bias and ethics cues. Scenarios about under-represented groups, skewed samples, or sensitive attributes point to EDA as the place to detect bias. The correct response often involves fairness checks, representativeness analysis, or Trustworthy AI considerations.
Tip 6: Class imbalance and outliers. If a question mentions rare events (fraud, defects, disease), think class imbalance found during EDA. For outliers, the best answer investigates whether they are errors or meaningful signals rather than deleting them automatically.
Tip 7: Remember iteration. CPMAI is agile and iterative. Answers suggesting EDA is done once and never revisited are usually wrong.
Tip 8: Know the data quality dimensions. Be ready to match a symptom to its dimension:
- nulls point to completeness
- duplicates point to uniqueness
- conflicting formats point to consistency
- stale data points to timeliness
Tip 9: Choose the most comprehensive, proactive answer. PMI exams favor answers that are proactive, collaborative (involving data owners and SMEs), and documented. A strong choice might be "perform data profiling and document findings in a data quality report shared with stakeholders".
Tip 10: Watch keywords. Words such as profile, summarize, visualize, distribution, correlation, missing values, outliers, anomalies, patterns, data quality assessment, sufficiency signal EDA. Words such as clean, transform, normalize, encode, impute, augment signal Data Preparation.
Tip 11: Use elimination. Remove options that:
- skip data understanding
- treat technical tools as the goal
- ignore stakeholders
- disregard data quality or ethics
9. Sample Exam-Style Question
An AI project team has collected three years of customer transaction data to build a churn prediction model. Before beginning model development, what should the project manager ensure happens next?
A. Train a baseline model to measure accuracy
B. Normalize all numerical features
C. Conduct exploratory data analysis to assess data quality, distributions, and suitability for the business objective
D. Deploy a pilot to production
Answer: C. EDA in the Data Understanding phase must come before preparation (B), modeling (A), and operationalization (D).
10. Key Takeaways
- EDA is the investigative heart of CPMAI Phase II: Data Understanding.
- It uncovers quality issues, patterns, bias, and gaps before money is spent on modeling.
- It directly informs Data Preparation and the go/no-go decision.
- It is iterative and tied to business objectives.
- On the exam, choose answers that explore and diagnose data first, involve stakeholders, address bias, and respect the CPMAI phase sequence.
Test mode:
Unlock Premium Access
PMI Certified Professional in Managing AI (PMI-CPMAI)
- Access to ALL Certifications: Study for any certification on our platform with one subscription
- 3129 Superior-grade PMI Certified Professional in Managing AI (PMI-CPMAI) practice questions
- Unlimited practice tests across all certifications
- Detailed explanations for every question
- PMI-CPMAI: 5 full exams plus all other certification exams
- 30-Day Money-Back Guarantee: Full refund within 30 days if unsatisfied
- Risk-Free: 7-day free trial with all premium features!
More Exploratory Data Analysis questions
70 questions (total)
Other Identify Data Needs concepts
Access Controls and User Permissions for DataCloud Storage and Distributed Data RepositoriesCollaboration Tools and Version Control for AI WorkspacesComparing Available Data Against RequirementsCPMAI Phase II: Data UnderstandingData Collection Processes for Ongoing FeedsData Distributions, Bias and GapsData Extraction from Identified SourcesData Freshness and Relevance for TrainingData Gaps and Remediation StrategiesData Governance Fundamentals for AIData Lineage and Usage Documentation for AuditData Ownership and Access PermissionsData Protection Compliance for Project DataData Quality Dimensions: Accuracy, Completeness, ConsistencyData Quality Standards and Acceptance CriteriaData Refresh and Update ProceduresData Representativeness for Target Use CasesData Schema and Structure for Modeling CompatibilityData Stewards and Data Governance TeamsData Storage and Backup for AI ProjectsData Sufficiency for Robust ModelsData Transfers and Migrations to AI EnvironmentsData Types and Formats for AI Model TrainingData Usage Rights and Licensing AgreementsData Volume Requirements and Sampling StrategiesEngaging Business Users on Data Context and MeaningExecutive Summaries of Data Assessment FindingsExternal and Third-Party Data SourcesGo/No-Go Decisions on Data ReadinessLegacy Systems and Historical Data ArchivesLocating Domain Experts for Data SourcesMapping Data Requirements to Business ObjectivesMapping Internal Databases and Data WarehousesPresenting Data Readiness to StakeholdersProvisioning Compute for Data Processing and TrainingSecure Development Environments for AI TeamsTechnical Experts on Data Systems and StructuresTemporal and Granularity Requirements for DataTranslating Technical Data Concepts for Business AudiencesValidating Completeness and Accuracy During CollectionVisualizations and Reports for Data Insights