- What the CDS Domain Outline Actually Is
- Domains 1 and 5: Framing the Question and the Data
- Domains 2, 3 and 4: Storage, Math, and Programming
- Domains 6 and 7: Cleaning and Exploring Data
- Domains 8, 9 and 10: Inference, Prediction, and Causality
- Domains 11 and 12: Written Analysis and Reproducibility
- Domain Map at a Glance
- Sequencing Your Study Around the Domains
- Registration, Fees, and Unverified Details
- Frequently Asked Questions
- The CDS outline from ICCP lists 12 official topics, published without percentage weights or a revision year.
- Domain 1 starts with the business question, not the algorithm, so framing skills are testable content.
- Domains 8 through 10 (inference, prediction, causality) form the conceptual core; learn to tell them apart.
- Written analysis and reproducibility are full domains, not afterthoughts, so practice explaining results in prose.
What the CDS Domain Outline Actually Is
The Certified Data Scientist - Associate/Practitioner (CDS) credential is issued by the Institute for the Certification of Computing Professionals (ICCP). Its currently accessible Data Science Exam outline lists 12 official topics. That number is the firmest structural fact available, and it shapes how you should plan.
Two caveats matter before you build a study plan around the list. First, the outline publishes no percentage weights, so nobody can honestly tell you that one domain is worth a specific share of your score. Second, the outline carries no revision year, and it has not been verified as a complete multi-exam pathway blueprint. Treat it as the authoritative topic list, and treat any site claiming exact domain percentages with suspicion. If you want the broader picture of the credential first, start with What Is CDS Certification? and then come back here.
The domain order itself tells a story. It follows the life of a data science project: understand the business problem, find and store the data, apply math and code, define the analytic question, clean, explore, model, predict, ask causal questions, write it up, and make the whole thing reproducible. Reading the list as a project lifecycle makes the 12 areas far easier to retain than memorizing them as isolated headings.
Domains 1 and 5: Framing the Question and the Data
These two domains bracket the technical work. One asks whether you are solving the right problem; the other asks whether you have classified the question and the data correctly before you touch a model.
Domain 1: Business & Technology Issues: Starting with the Question First - What Problem are you Trying to Solve?
The title is itself a study instruction. Expect scenario-style content where the correct response is to clarify the objective before choosing a technique.
- Translating a vague stakeholder request into a defined, answerable question
- Recognizing when a problem does not need a complex model at all
- Connecting technology constraints and business context to the analytic approach
- Judging what success looks like before analysis begins
Domain 5: The Data Analytic Question and Types of Data and Reporting
This domain is about matching the kind of question to the kind of analysis, and understanding the data you hold and how results get reported.
- Distinguishing question types such as descriptive, exploratory, inferential, predictive, and causal
- Identifying data types and how they constrain which methods are valid
- Reporting findings in a form that fits the audience and the question asked
A common failure pattern is treating these as "soft" domains and spending all study time on code and algorithms. The outline places them at the front of the list deliberately. Misclassifying a causal question as a predictive one is a conceptual error that no amount of programming skill will rescue, and it connects directly to Domains 9 and 10 below.
Domains 2, 3 and 4: Storage, Math, and Programming
This cluster is the technical bedrock. Candidates from software backgrounds usually find Domain 4 comfortable and Domain 3 harder; candidates from statistics backgrounds often have the reverse experience.
Domain 2: Data Storage, Big Data and Sources
Expect questions about where data lives and how it is accessed. Study the differences between structured and unstructured storage, how large-scale data changes your options, and how the origin of a dataset affects its reliability. Source awareness feeds directly into Domain 6, because data quality problems usually trace back to how the data was collected and stored.
Domain 3: Mathematics and Statistical Data Science
This is the quantitative foundation: the probability and statistical ideas that every later modeling domain assumes. Do not read it as a standalone math exam. The practical test is whether you can reason about distributions, variability, and uncertainty well enough to understand why a model behaves the way it does.
Domain 4: Programming Skills
The outline names this domain generically, so the safest preparation is language-agnostic fluency: manipulating data structures, writing clear control flow, reading and reasoning about code, and understanding how analysis steps are expressed programmatically. Because the published outline does not specify a required language, confirm any language expectations with ICCP rather than assuming one.
Domains 6 and 7: Cleaning and Exploring Data
Domain 6: Tidying the Data - Data Cleaning and Quality
The word "tidying" signals an emphasis on putting data into a consistent, analysis-ready structure, not just fixing typos.
- Handling missing values and understanding why they are missing
- Detecting and treating outliers and inconsistent records
- Restructuring data so each variable and observation is cleanly represented
- Assessing data quality before trusting any downstream result
Domain 7: Exploratory Analysis
Exploration is where you learn what the data can and cannot support, before formal modeling.
- Summarizing distributions and relationships with statistics and visualizations
- Spotting structure, anomalies, and candidate hypotheses
- Understanding that exploration generates hypotheses rather than confirming them
The conceptual trap here is confusing exploration with confirmation. A pattern you discover while exploring is a lead, not a conclusion, and the outline's progression into Domain 8 (inference) exists precisely to formalize that distinction.
Domains 8, 9 and 10: Inference, Prediction, and Causality
If any part of the outline deserves your deepest study, it is this trio. They look similar on the surface, since all three involve models and data, but they answer fundamentally different questions. Scenario questions are likely to test whether you can tell them apart.
Domain 8: Statistical Modeling and Inference
Using a sample to draw conclusions about a larger population, with honest accounting for uncertainty.
- Fitting models and interpreting their parameters
- Quantifying uncertainty rather than reporting point estimates alone
- Understanding what assumptions a model requires and what happens when they fail
Domain 9: Prediction and Machine Learning
Here the goal shifts from explaining to forecasting accuracy on new data.
- Training versus evaluating models, and why performance must be judged on unseen data
- Overfitting, generalization, and the tradeoffs involved in model complexity
- Choosing evaluation approaches that fit the prediction task
Domain 10: Causality
The hardest conceptual jump: moving from "these things move together" to "this change produces that effect."
- Why correlation, even strong correlation, does not establish cause
- The role of confounding and how study design addresses it
- Why randomized experiments carry a different evidentiary weight than observational data
Key Takeaway
Build a one-line test for each domain: inference asks "what is true about the population?", prediction asks "what will happen next?", and causality asks "what would change if we intervened?" When a scenario question appears, name which of the three it is before reading the answer options.
For a sense of how demanding this conceptual material is relative to other certifications, see How Hard Is the CDS Exam?
Domains 11 and 12: Written Analysis and Reproducibility
Many data certifications stop at modeling. The CDS outline does not. Giving communication and reproducibility their own domains reflects what employers actually need from practitioners.
Domain 11: Written Analysis
Turning technical results into a clear, honest narrative.
- Stating the question, method, finding, and limitations in plain language
- Matching the level of detail to the audience
- Avoiding overstated claims, especially where evidence supports only association
Domain 12: Reproducibility
Ensuring another person can rerun your work and arrive at the same result.
- Documenting data sources, transformations, and decisions
- Structuring analysis so steps are repeatable rather than manual and undocumented
- Recognizing why irreproducible work undermines trust in findings
These two domains interlock with the rest of the outline. A reproducible analysis is only valuable if it was framed correctly (Domain 1), and a well-written analysis is only trustworthy if its causal and inferential claims (Domains 8 and 10) were handled carefully. Practice writing a short methods-and-limitations paragraph after every practice problem you work.
Domain Map at a Glance
| Domain | Core Question It Answers | Typical Weak Spot |
|---|---|---|
| 1. Business & Technology Issues | What problem are we solving? | Jumping to methods too early |
| 2. Data Storage, Big Data and Sources | Where does the data come from and live? | Ignoring source reliability |
| 3. Mathematics and Statistical Data Science | What quantitative ideas underpin everything? | Weak probability intuition |
| 4. Programming Skills | Can I express analysis in code? | Unreadable, unrepeatable scripts |
| 5. Analytic Question, Data Types, Reporting | What kind of question and data is this? | Misclassifying question type |
| 6. Tidying the Data | Is the data clean and trustworthy? | Skipping quality checks |
| 7. Exploratory Analysis | What does the data suggest? | Treating leads as conclusions |
| 8. Statistical Modeling and Inference | What is true of the population? | Ignoring uncertainty |
| 9. Prediction and Machine Learning | What will happen on new data? | Overfitting, weak evaluation |
| 10. Causality | What would an intervention change? | Confusing correlation with cause |
| 11. Written Analysis | Can I communicate this honestly? | Overstated claims |
| 12. Reproducibility | Can someone else rerun this? | Undocumented manual steps |
Sequencing Your Study Around the Domains
Because the outline follows a project lifecycle, studying in outline order mostly works, with one adjustment: do the math and programming foundations early, since everything later leans on them. A reasonable arrangement, scaled to however many weeks you have:
Framing and foundations
- Domains 1 and 5: practice classifying questions by type
- Domain 3: refresh probability and statistics fundamentals
Data handling
- Domains 2, 4 and 6: storage concepts, code fluency, cleaning workflows
Exploration and modeling
- Domains 7 and 8: explore a dataset, then formalize findings with inference
The conceptual core
- Domains 9 and 10: drill the prediction versus causality distinction hardest
Communication and integration
- Domains 11 and 12, then mixed review across all 12 areas
Weeks 4 and 5 deserve extra buffer if causal reasoning is new to you. For a fuller planning framework, the CDS study guide covers pacing in more depth, and the CDS cheat sheet works well as a final-week compression of the definitions you need at your fingertips. When you want to test retention across all 12 areas, use the CDS practice tests and review every miss by domain.
Registration, Fees, and Unverified Details
Domain knowledge is only half the preparation. The administrative side has several points that cannot be asserted with confidence from available information, and it is better to know that now than to discover it late.
- Exam fees: ICCP's current store lists $350 per professional exam including proctoring and $249 per foundational exam. An older FAQ shows $299, so confirm the current figure. Which exam SKUs apply to CDS, and how many exams the credential requires, are not verified, so no total certification-exam fee can be stated. See the CDS cost breakdown for how to work through this.
- Application: ICCP lists a generic candidate application fee of $45, and the general application requires a resume and an ethics agreement.
- Prerequisites: Exact CDS education, experience, training-hour, and reference requirements require confirmation from the issuer. Check the CDS requirements guide and then verify directly with ICCP.
- Proctoring: General ICCP examinations use approved proctoring, including arranged remote or in-person options. The CDS-specific testing vendor and workflow are not verified.
- Exam rules: Open-book status, calculator rules, and whether delivery is adaptive are not verified. Do not assume reference materials will be allowed.
- Maintenance: An active certification runs on a 3-year professional-development cycle. General ICCP tier rules specify Associate at 60 hours with a $35 annual fee and Practitioner at 120 hours with a $75 annual fee, with candidate-specific applicability to be confirmed.
Frequently Asked Questions
The currently accessible ICCP Data Science Exam outline lists 12 official topics, from Business & Technology Issues through Reproducibility. The outline does not publish percentage weights or a revision year.
Not in the outline currently available. Any source quoting exact weights for individual CDS domains is going beyond what the published outline supports, so plan for broad coverage of all 12 areas instead.
Allocate time based on your background. Most candidates find Domains 8 through 10 (inference, prediction, and causality) the most conceptually demanding, because distinguishing the three question types takes deliberate practice.
The published outline names the domain simply as Programming Skills and does not specify a language. Confirm any language expectations with ICCP, and in the meantime build general fluency in data manipulation and readable, repeatable code.
Open-book status, calculator rules, and adaptive delivery have not been verified for this credential. Ask ICCP directly before exam day rather than assuming reference materials are permitted.
Mapping your strengths against all 12 domains is the most useful first step you can take. Once you know where you are thin, the ROI analysis and the pass rate discussion will make more sense, and the practice test site lets you check each domain against realistic questions.