CPDA Certification Specification | Data Science Institute

Certification specification

Certified Professional Data Analyst

The Certified Professional Data Analyst (CPDA®) is the Data Science Institute's professional credential in applied data analytics.

CPDA certification page

CPDA specification at a glance
Level	Professional
Structure	4 Parts, 13 Units
Assessment	Exam A, then Exam B
Pass mark	50% in each exam
Exam delivery	Online, remotely proctored
Progression	CPDS®

Section 1Certification overview

The Certified Professional Data Analyst (CPDA®) is the Data Science Institute's professional credential in applied data analytics. It validates practical competence in preparing, analysing, visualising, reducing, interpreting and modelling data to support evidence-based decisions in organisational settings.

CPDA® develops applied capability across the analytics lifecycle using Python, Excel and associated analytical methods. Each unit is grounded in realistic datasets, worked examples and practice tasks that reflect the analytical work carried out by professional data analysts.

CPDA® is designed for learners who already have some exposure to data work or who are progressing from the Certified Associate Data Analyst (CADA™). Its audiences include working professionals moving into analytical roles, analysts seeking formal validation of applied capability, domain specialists who need deeper analytics competence, and learners preparing to progress to the Certified Professional Data Scientist (CPDS®) and onward to advanced DSI qualifications.

CPDA® sits above CADA™ on the DSI pathway and below CPDS® in technical depth. It is intentionally more substantial than the associate-level credential, with greater emphasis on Python analytics, statistical inference, data reduction, unsupervised analytics and predictive modelling.

The certification is delivered online through the Certifications Platform, with structured learning activities and two summative certification examinations delivered online with remote proctoring.

Section 2Programme structure

2.1 Structural model

The programme is organised as Parts → Units → Lessons, with unit scope defined in the unit-level syllabus and lesson-level descriptors defined in the lesson-by-lesson breakdown.

2.2 Parts and weightings

Parts and weightings for CPDA
Part	Share of the certification	Weighting
Data Analytics FoundationPart 1		35%
Data VisualisationPart 2		15%
Data Reduction and Unsupervised AnalyticsPart 3		20%
Predictive ModellingPart 4		30%

2.3 Parts and units

Part 1 · 35%

Data Analytics Foundation

Unit 1 Python for Data Analysis
Unit 2 Data Management
Unit 3 Descriptive Statistics
Unit 4 Inferential Statistics

Part 2 · 15%

Data Visualisation

Unit 1 Introduction to Data Visualisation
Unit 2 Data Visualisation Using Python

Part 3 · 20%

Data Reduction and Unsupervised Analytics

Unit 1 Dimensionality Reduction
Unit 2 Principal Component Analysis
Unit 3 Clustering and Segmentation

Part 4 · 30%

Predictive Modelling

Unit 1 Predictive Modelling Basics
Unit 2 Linear Regression
Unit 3 Classification Models
Unit 4 Model Validation

2.4 Programme scale

This consolidated specification includes 4 Parts and 13 Units. The certification is supported by lesson learning materials, knowledge checks, unit practices, and part-level case studies delivered through the Certifications Platform.

Section 3Learning and assessment requirements

3.1 Standard learning pathway

The programme combines instruction, applied practice, and objective checks at lesson level; consolidation at unit level; part-level applied case work where present; and certification readiness activities supported by two summative certification examinations at certification level.

3.2 Progression requirements

To be eligible for award:

All Unit Practices must be completed (ungraded but mandatory).
At least 90% of platform learning activities must be marked complete, including applied activities and case studies where present in the pathway.
Both CPDA® summative certification examinations must be passed.

3.3 Summative certification exams

Exam A

Type: Knowledge
Format: Computer-based, objective examination
Duration: 120 minutes
Pass mark: 50%
Sequence: Must be passed before Exam B can be attempted

Exam B

Type: Practical application
Format: Applied practical examination
Duration: 180 minutes
Pass mark: 50%
Sequence: Taken after Exam A is passed

Assessment delivery environment is online with remote proctoring.

CPDA® certification exam structure
Attribute	Exam A	Exam B
Type	Knowledge	Practical application
Format	Computer-based, objective examination	Applied practical examination
Duration	120 minutes	180 minutes
Pass mark	50%	50%
Sequence	Must be passed before Exam B can be attempted	Taken after Exam A is passed

3.4 Use of the Certifications Platform

CPDA® is delivered through the Institute's Certifications Platform, which structures each certification consistently. Each Part contains units, units contain lessons, and supporting practice and case-study material is provided at the appropriate level of the structure.

At Part level — a Case Study integrating the units within the Part, where appropriate to the structure and learning outcomes.
At Unit level — a Unit Practice that demonstrates capability against the unit's ILOs.
At Lesson level — Learn content, Lesson Practice, and a Knowledge Check aligned to the lesson's ILOs.
Across the programme — curated datasets, worked examples, code-based activities and quizzes that support lesson-level practice, applied consolidation and revision for the certification examinations.

Section 4Part descriptors

Each part below sets out its scope, its intended learning outcomes, and the units it contains. Unit entries give the unit purpose, the applied outputs expected, the tools and environment used, the scope of the unit, and the unit intended learning outcomes.

4.1 Part 1 · 35% of the certificationData Analytics Foundation

Part 1 establishes foundational competence for CPDA®: practical Python fluency for analytics, end-to-end data management workflows, descriptive understanding of datasets, and inferential reasoning for drawing conclusions from data.

Part intended learning outcomes

Use a Python analytics environment to run reproducible analysis workflows and apply core programming constructs to analytics tasks.
Prepare and manage datasets for analysis using reproducible data-management steps, including import, inspection, cleaning, transformation, merging, aggregation and export.
Summarise and interpret datasets using appropriate descriptive statistics and exploratory techniques, including distribution and relationship interpretation.
Select and apply appropriate inferential methods, including confidence intervals and hypothesis tests, and interpret results with consideration of assumptions.
Communicate statistical and analytical outputs clearly, with correct interpretation of what results do and do not imply.

Unit 1

Python for Data Analysis

Unit purpose: Build foundational Python capability for analytics, including core programming constructs and introductory data structures used throughout later CPDA® units.
Applied outputs: Reproducible Python artefacts demonstrating correct use of types, functions, control flow, date/time handling, directories and common analytics-support tasks.
Tools and environment: Python in a notebook or IDE environment, including core Python libraries and introductory use of pandas data structures for data handling tasks.
Unit scope: Python environment and workflow; introductory pandas data structures; importing and using libraries; core data types; numeric and string functions; control flow; dates and times; and file and directory operations used in analytics projects.

Unit intended learning outcomes

Set up and use a Python analytics environment and run basic programs for data analysis tasks.
Explain and use Python libraries and modules, including importing packages and describing the purpose of key analytics libraries introduced.
Use core Python data types and structures, including numbers, strings, lists, tuples and dictionaries, with indexing and basic operations.
Apply numeric and string functions and operators to compute results and perform basic text handling tasks.
Implement control flow, including conditionals and loops, to automate repeated steps and decision rules.
Work with directories and date/time data using the methods introduced.

Unit 2

Data Management

Unit purpose: Develop practical competence in preparing and managing datasets for analysis, including importing, inspecting, cleaning, combining, summarising and exporting data in a reproducible way.
Applied outputs: Analysis-ready datasets and reproducible preparation steps, including joins or merges, aggregations, filtered extracts and documented handling of missing values.
Tools and environment: Python with pandas for importing, cleaning, transforming, combining and exporting datasets.
Unit scope: End-to-end data management workflow: import and export; inspection and modification; subsets and sorting; joins, concatenation and aggregation; and missing data detection and strategies introduced.

Unit intended learning outcomes

Import data from multiple formats and sources using reproducible workflows.
Identify and resolve data-quality issues, including missingness, duplicates and data type problems.
Create subsets, filter records and sort data to prepare analysis-ready views.
Merge, join, append, aggregate and engineer features for analytical tasks.
Prepare clean, structured datasets and document key data preparation decisions.

Unit 3

Descriptive Statistics

Unit purpose: Build capability to summarise and interpret datasets using descriptive statistics and exploratory methods, including understanding distributions and relationships and using automated EDA approaches.
Applied outputs: Descriptive summaries, distribution and relationship outputs, automated profiling artefacts, and clear written interpretations of key patterns.
Tools and environment: Python with pandas for statistical summaries, plus standard Python statistics and visualisation tooling for exploratory analysis and automated EDA profiling where applicable.
Unit scope: Descriptive statistics foundations; broader distribution measures; relationship exploration; automated profiling; and sampling method introduction and rationale.

Unit intended learning outcomes

Identify variable types and select appropriate summary measures.
Compute and interpret central tendency, spread, skewness and correlation.
Summarise datasets using numerical profiles and charts.
Generate descriptive statistics in Python and interpret outputs in context.
Explain basic sampling methods and why sampling choices affect interpretation.

Unit 4

Inferential Statistics

Unit purpose: Develop competence in inferential reasoning and hypothesis testing in Python, including selecting appropriate tests, checking assumptions, interpreting results, and applying ANOVA and non-parametric alternatives.
Applied outputs: Executed inferential tests and ANOVA workflows, justified test choice based on assumptions and data type, and correctly interpreted and reported statistical conclusions.
Tools and environment: Python statistical analysis tooling for hypothesis testing, variance testing, ANOVA workflows and non-parametric testing.
Unit scope: Hypothesis testing foundations; normality diagnostics; parametric tests and variance equality testing; one-way and two-way ANOVA; non-parametric alternatives; and categorical association testing approaches covered.

Unit intended learning outcomes

Explain sampling designs, sampling distributions and the central limit theorem.
Construct and interpret confidence intervals.
Conduct hypothesis tests, including t-tests, chi-square tests, ANOVA and non-parametric tests introduced.
Check relevant assumptions and select appropriate inferential methods.
Perform inferential analysis in Python and interpret statistical significance, assumptions and limitations.

4.2 Part 2 · 15% of the certificationData Visualisation

Part 2 develops professional visual communication capability: principles of effective data visualisation, responsible chart selection and storytelling, and applied implementation of visuals using Python and Excel for analytical and stakeholder contexts.

Part intended learning outcomes

Select appropriate visual encodings and chart types based on analytical question, data type and audience needs.
Apply core visual-design principles to improve clarity, interpretability and credibility of charts and dashboards.
Identify misleading or low-quality visualisations and propose corrections aligned to responsible presentation expectations.
Create and interpret common analytical charts in Python using the charting methods introduced.
Create and interpret reporting charts in Excel using the summarisation and charting methods introduced.

Unit 1

Introduction to Data Visualisation

Unit purpose: Establish visual communication judgement: selecting appropriate charts, designing for clarity and credibility, avoiding misleading presentation, and structuring insight narratives.
Applied outputs: Critique and improve visuals; justify chart selection for audience and question; and propose design changes that improve clarity, accuracy and impact.
Tools and environment: Tool-agnostic principles that apply to any charting environment used in analytics and reporting.
Unit scope: Principles of data visualisation; common mistakes; storytelling structure; and responsible or ethical presentation expectations.

Unit intended learning outcomes

Explain why data visualisation is essential for communicating analytical insights and supporting decisions.
Select appropriate chart types for common analytical questions based on data type and purpose.
Apply core visual-design principles to improve clarity, interpretability and credibility.
Identify common visualisation mistakes that reduce clarity or mislead and propose corrections.
Apply a simple data-storytelling structure to communicate insight and implication.
Apply responsible presentation expectations, including transparency and avoidance of misleading communication.

Unit 2

Data Visualisation Using Python

Unit purpose: Build practical capability to create, refine and interpret common chart types in Python for analysis and communication, including categorical, distribution and relationship visuals.
Applied outputs: Reproducible visual analysis outputs with appropriate chart choice and clear interpretation of distributions, comparisons and relationships.
Tools and environment: Python visualisation environment using common charting libraries such as matplotlib, seaborn and related Python charting tools.
Unit scope: Categorical and composition visuals; distributions; relationship visuals; chart-ready summaries; and formatting decisions that improve analytical communication.

Unit intended learning outcomes

Build and customise common Python charts for categorical comparison and composition using the methods introduced.
Create and interpret distribution plots, including box plots, histograms and density plots.
Create and interpret relationship visuals, including scatter plots and extensions introduced.
Prepare chart-ready summaries using grouping and reshaping approaches introduced for plotting.
Apply labels, layout, readability and formatting adjustments to improve interpretability.

4.3 Part 3 · 20% of the certificationData Reduction and Unsupervised Analytics

Part 3 develops learners' ability to simplify, summarise and interpret complex datasets using data reduction and unsupervised analytics methods. Learners apply dimensionality reduction methods, including principal component analysis, and clustering methods, including K-Means, to support segmentation, profiling, modelling preparation and decision-making.

Part intended learning outcomes

Explain the purpose of data reduction and identify when dimensionality reduction is useful in analytics workflows.
Apply principal component analysis to reduce high-dimensional datasets into a smaller number of interpretable components.
Interpret variance explained, component scores and component loadings in applied analytical contexts.
Explain how principal components can support regression workflows through principal component regression.
Apply K-Means clustering to identify natural groupings in data.
Interpret cluster outputs for segmentation, profiling and decision-support use cases.
Communicate the benefits and limitations of data reduction and clustering methods clearly to technical and non-technical audiences.

Unit 1

Dimensionality Reduction

Unit purpose: Introduce the need for data reduction in analytics and modelling, especially where datasets contain many variables, correlated features or complex patterns that are difficult to interpret directly.
Applied outputs: Exploratory summaries of high-dimensional datasets, identification of redundancy or correlation between variables, and a clear explanation of where dimensionality reduction may be useful.
Tools and environment: Python analytics environment using pandas and relevant visual or statistical libraries.
Unit scope: High-dimensional data; feature redundancy; correlation and multicollinearity; motivation for data reduction; overview of unsupervised methods; and the relationship between data reduction, exploratory analysis and modelling preparation.

Unit intended learning outcomes

Explain what dimensionality means in a dataset.
Identify problems caused by many variables, correlated features and redundant information.
Explain why reducing dimensionality can improve interpretability, visualisation and modelling workflows.
Distinguish data reduction from feature selection at an introductory level.
Describe common use cases for data reduction in analytics and professional decision-making.

Unit 2

Principal Component Analysis

Unit purpose: Develop applied competence in principal component analysis as a method for transforming correlated variables into a smaller number of components that summarise the main patterns of variation in the data.
Applied outputs: PCA outputs, reduced-dimension datasets, variance-explained summaries, component loadings, component score plots and introductory principal component regression workflows.
Tools and environment: Python-based PCA workflow using appropriate data preprocessing, scaling, decomposition and interpretation methods.
Unit scope: PCA motivation and workflow; scaling and preprocessing; component intuition; explained variance; component loadings; component scores; component visualisation; and use of principal components in regression workflows.

Unit intended learning outcomes

Explain the purpose of PCA and identify situations where it is appropriate.
Prepare data for PCA, including appropriate scaling where required.
Compute principal components using Python.
Interpret explained variance and decide how many components to retain.
Interpret component loadings and component scores in context.
Create reduced-dimensional representations of datasets.
Explain how principal components can be used in regression workflows.
Interpret principal component regression outputs at an applied level.

Unit 3

Clustering and Segmentation

Unit purpose: Introduce clustering as an unsupervised analytics method for identifying natural groupings in data and using those groupings for segmentation, profiling and insight generation.
Applied outputs: K-Means cluster assignments, cluster summaries, visualisations of cluster structure, interpreted segmentation profiles and an applied customer segmentation case study.
Tools and environment: Python-based clustering workflow using pandas, preprocessing tools and clustering or visualisation libraries.
Unit scope: Unsupervised learning; clustering purpose and workflow; distance-based grouping; K-Means algorithm intuition; selecting and interpreting clusters; profiling cluster characteristics; and use cases such as customer segmentation, operational grouping and behavioural analysis.

Unit intended learning outcomes

Explain the purpose of clustering and distinguish it from supervised modelling.
Describe the basic intuition of the K-Means algorithm.
Prepare data for clustering using appropriate preprocessing steps.
Apply K-Means clustering in Python.
Interpret cluster assignments and cluster centres.
Use visual and numerical summaries to profile clusters.
Explain how clustering supports segmentation and decision-making.
Identify limitations of K-Means, including sensitivity to scaling, choice of k and cluster shape assumptions.

4.4 Part 4 · 30% of the certificationPredictive Modelling

Part 4 develops predictive modelling capability from problem framing through model development and validation: regression versus classification, multiple linear regression diagnostics and selection, logistic regression and extensions, and out-of-sample validation with appropriate metrics.

Part intended learning outcomes

Frame predictive modelling problems and select appropriate modelling approaches based on target variable type and problem context.
Develop and interpret regression models, including specification, coefficient interpretation and performance evaluation using appropriate metrics.
Diagnose key regression assumptions using the diagnostic methods introduced and apply model-selection principles to select a final model.
Develop and interpret classification models using logistic regression within a GLM framing, including probability interpretation and performance evaluation.
Validate regression and classification models using out-of-sample approaches introduced and draw metric-driven conclusions about generalisability and overfitting risk.

Unit 1

Predictive Modelling Basics

Unit purpose: Introduce predictive modelling and establish method selection between regression and classification based on problem framing and target variable type.
Applied outputs: Framed predictive problems, identification of appropriate modelling families, and description of the workflow required to develop a predictive solution.
Tools and environment: Tool-agnostic workflow and method selection concepts used in subsequent Python implementation units.
Unit scope: Predictive modelling concepts; supervised versus unsupervised learning; regression versus classification; overfitting and underfitting; bias-variance trade-off; modelling workflow; and preparation for applied model development.

Unit intended learning outcomes

Distinguish supervised and unsupervised learning.
Select models for regression and classification tasks based on target variable type.
Explain overfitting, underfitting and the bias-variance trade-off.
Describe the core workflow for preparing data and developing predictive models.
Assess basic modelling assumptions and data preparation requirements.

Unit 2

Linear Regression

Unit purpose: Develop end-to-end capability in multiple linear regression including specification, interpretation, diagnostics and model selection.
Applied outputs: Built and compared regression models, diagnostic evidence, selection of an appropriate final model, and communication of model meaning and limitations in context.
Tools and environment: Python modelling environment for multiple linear regression, including regression fitting and diagnostic tooling suitable for assumptions checks and model selection.
Unit scope: Multiple linear regression; categorical and dummy variables; multicollinearity; normality and homoscedasticity assumptions; influential observations; and model selection.

Unit intended learning outcomes

Fit simple and multiple regression models.
Interpret coefficients, intervals and significance.
Diagnose assumptions using residual plots and tests introduced.
Manage multicollinearity, transformations and interactions where introduced.
Evaluate performance using regression metrics such as MAE, RMSE and R².
Select and communicate a reasonable final regression model based on evidence and context.

Unit 3

Classification Models

Unit purpose: Develop classification modelling capability using logistic regression within a GLM framing, including binary logistic regression and extensions to multinomial and ordinal contexts, with appropriate performance evaluation.
Applied outputs: Fitted and interpreted classification models, generated predicted probabilities, evaluated performance using multiple metrics and curves introduced, and communicated results in decision-support terms.
Tools and environment: Python modelling environment for logistic regression and GLM-based classification, including tools for generating predictions and evaluating classification performance.
Unit scope: Binary logistic regression; predicted probabilities and threshold logic; classification tables and metrics; ROC/AUC and related concepts; GLM framing; multinomial logistic regression; and ordinal logistic regression.

Unit intended learning outcomes

Explain when binary logistic regression is appropriate and interpret odds, odds ratios and test concepts introduced.
Convert predicted probabilities to class predictions using threshold logic introduced.
Interpret classification tables and metrics, including accuracy, precision, recall or sensitivity and specificity.
Explain and interpret ROC/AUC and other performance concepts introduced, including lift and KS statistics where presented.
Describe the GLM framework and position logistic regression within GLMs at a high level.
Explain when multinomial and ordinal logistic regression are appropriate and interpret predicted probabilities for multi-class or ordered outcomes as introduced.

Unit 4

Model Validation

Unit purpose: Develop robust validation practice using out-of-sample evaluation for both regression and classification.
Applied outputs: Validation summaries and metric-driven conclusions about generalisability and overfitting risk.
Tools and environment: Python validation environment supporting train/test evaluation and cross-validation workflows, including metric computation for regression and classification.
Unit scope: Out-of-sample validation; hold-out validation; k-fold and repeated k-fold approaches introduced; regression validation using R² and RMSE; classification validation using accuracy, precision and recall or sensitivity; and conclusions about generalisability.

Unit intended learning outcomes

Explain why validation should be out-of-sample and distinguish hold-out validation from k-fold or repeated k-fold approaches introduced.
Interpret regression validation outputs using R² and RMSE in a validation context.
Compare training and testing results to identify possible overfitting.
Interpret classification validation outputs using accuracy, precision and recall or sensitivity in a validation context.
Produce metric-driven conclusions about generalisability and overfitting risk based on validation results.

Section 5Appendices

Appendix A Position in the DSI Certification Pathway

CPDA® is the professional-level data analyst credential in the DSI certification pathway. It builds on the associate-level foundations of CADA™ and prepares learners for progression to CPDS® where the emphasis moves into advanced data science, machine learning and deep learning.

Learners who enter CPDA® directly should have sufficient digital, numerical and analytical readiness to work with Python-based analytics tasks. Learners progressing from CADA™ will have already encountered the business reporting, SQL, Excel, Power BI and applied analytics foundations that support progression into CPDA®.

CPDA certification page CADA specification CPDS specification

Appendix B Platform Learning Terms

Lesson Practice refers to lesson-level applied activity that helps learners practise the concept or workflow immediately after learning it. Knowledge Checks are objective checks aligned to lesson learning outcomes. Unit Practices are mandatory consolidation activities linked to the unit's intended learning outcomes. Part-level Case Studies integrate learning across units and support applied professional judgement.