SOA Exam PA Guide
Exam PA tests predictive analytics in a business problem setting: data, modeling choices, interpretation, and communication.
- Role
- Exam Guide
- Level
- Core
- Time
- Reference
- Freshness
- Stable
What Exam PA Entails
- Predictive analytics, data analysis, model interpretation, and business communication.
- Less about memorizing formulas and more about explaining a defensible modeling workflow.
- Core practice: report writing, validation, limitations, stakeholder recommendations, and model comparison.
SOA Exam PA
Official syllabus facts are mapped; released project statements are topic-tagging inputs only.
What the official PDFs establish
- Format
- 3.5-hour exam focused on data analysis in a business context and written responses.
- Exam software
- Word and Excel are available; R and RStudio are not available at the exam.
- Assumed knowledge
- Probability, mathematical statistics VEE, and SRM are assumed.
Topic and domain coverage
| Topic | Weight | Source |
|---|---|---|
| Predictive Analytics Problem Definition | 10-20% | Source: Exam PA syllabus, p. 2 |
| Data Exploration and Visualization | 20-30% | Source: Exam PA syllabus, p. 2 |
| Data Transformations and Unsupervised Learning | 10-20% | Source: Exam PA syllabus, p. 3 |
| Generalized Linear Models | 25-35% | Source: Exam PA syllabus, p. 3 |
| Tree-Based Models | 10-20% | Source: Exam PA syllabus, p. 4 |
Chapter and reading intelligence
- SOA e-Learning modules
The PA syllabus includes required e-Learning module readings.
- ISLR, Frees, R for Everyone, and Data Visualization
Official readings include selected chapters and sections; page content should map topics rather than copy the reading list verbatim.
Official files used by the map
- Official syllabussyllabus
Primary source for format, software, topic weights, and readings.
Source: Exam PA - April 2026 Syllabus - Released PA project statementreleased-exam
Use to map task styles and reporting expectations; do not republish project text.
Quick Answer
PA is an ASA-path predictive analytics exam, not an FSA fellowship exam. It sits after the statistics foundation and rewards clear modeling judgment, not just coding.
Official Format, Software, And Scope
The April 2026 syllabus maps PA as a 3.5-hour exam focused on data analysis in a business context and written responses. The official overview says the exam combines e-learning modules with an exam that asks candidates to solve a business problem using a data set.
The syllabus also sets a practical software boundary: Word and Excel are available during the exam, while R and RStudio are not. That means preparation should focus on statistical interpretation, report structure, and workflow judgment rather than assuming you will have a full coding environment at exam time.
What The Syllabus Emphasizes
The weight structure puts most of the exam in business framing, data exploration, transformations, generalized linear models, and tree-based models. In other words, the exam is less about squeezing out every possible model and more about making a defensible analytic case.
- Predictive analytics problem definition: 10-20%.
- Data exploration and visualization: 20-30%.
- Data transformations and unsupervised learning: 10-20%.
- Generalized linear models: 25-35%.
- Tree-based models: 10-20%.
How PA Differs From SRM
SRM is where you learn the modeling ideas. PA is where you have to explain and defend their use under business constraints. That difference is why candidates who did well in SRM can still struggle in PA if they write like they are answering a statistics worksheet instead of advising a stakeholder.
The strongest PA answers are structured. They identify the business objective, explain the chosen analytic path, interpret results in plain language, state limitations honestly, and make a recommendation that follows from the evidence rather than from generic model enthusiasm.
What It Tests
The distinctive skill is translating a data or modeling workflow into a coherent actuarial report: what you did, why it is reasonable, what the limitations are, and what a stakeholder should do with the result.
That makes PA one of the most communication-sensitive pages on the site. Good preparation includes report outlines, model-comparison language, and clear ways to discuss tradeoffs, not just lists of modeling techniques.
Where ATPA Fits Next
ATPA is the next ASA-side predictive-analytics component after PA. The official ATPA page assumes VEE Mathematical Statistics, P, SRM, and PA, which means PA should be treated as the point where your foundational modeling and communication skills become advanced-project prerequisites.
Common Candidate Mistakes
The biggest PA mistake is giving technically plausible answers with weak business framing. A response can mention the right model family and still lose force if it does not connect the output to decision-making, limitations, and stakeholder action.
Another mistake is over-writing. PA answers should be precise and well-organized, not padded with generic machine-learning narration.