Study guide

1.2 The PCPA Project: Windows, Deliverables & Grading

CAS Property & Casualty Predictive Analytics PCPA

Track your progress as you read

Finished sections get a check mark.

1 Chapter 1: PCPA Exam & Project Orientation

1.1 PCPA Exam Format, Scoring & Registration 1.2 The PCPA Project: Windows, Deliverables & Grading

2 Chapter 2: Dealing with Data — Sourcing, Importing & Assessing

2.1 Requirements Gathering & Assessing Data Relevance 2.2 Importing, Joining & Reshaping Data Sets 2.3 Exploratory Data Analysis & Evaluating a Data Set 2.4 Data Quality Under ASOP 23 & Modeling Under ASOP 56

3 Chapter 3: Dealing with Data — Conditioning P&C Data for a GLM

3.1 Exposure Bases, Weights & Data Grain 3.2 Missing Data Mechanisms & Imputation Strategies 3.3 Outlier Diagnostics, Trimming & Winsorization 3.4 Variable Transformation & Scaling

4 Chapter 4: Model Diagnostics & Selection — Building a GLM

4.1 Exponential Dispersion Family & Link Functions 4.2 Poisson Regression & Exposure Offsets 4.3 Overdispersion, Quasi-Poisson & Negative Binomial 4.4 Gamma & Inverse Gaussian Models for Claim Severity

5 Chapter 5: Model Diagnostics & Selection — Model Form & Predictors

5.1 Tweedie Compound Poisson-Gamma Modeling 5.2 Binomial & Logistic GLMs for Binary Outcomes 5.3 Categorical Predictors, Base Levels & High-Cardinality Grouping 5.4 Interaction Terms & Non-Linear Effects 5.5 Interpreting Coefficients & Multiplicative Relativities

6 Chapter 6: Model Diagnostics & Selection — Evaluating and Improving the Fit

6.1 Deviance, Likelihood Ratio Tests & Residual Diagnostics 6.2 Stepwise Selection & Information Criteria Penalties 6.3 Multicollinearity, Correlated Predictors & Aliasing 6.4 Under-fitting, Over-fitting & Cross-Validation

7 Chapter 7: Model Diagnostics & Selection — Validation & Model Choice

7.1 Actual vs. Predicted, Calibration & Loss Ratio Charts 7.2 Simple Quantile Plots & Measuring Lift 7.3 Double Lift Charts & Competing Model Evaluation 7.4 Gini Index, Lorenz Curves & ROC Curves

8 Chapter 8: Model Interpretation & Presentation — Tables, Charts & Exhibits

8.1 Designing Tables & Statistical Exhibits 8.2 Chart Selection & Graphical Perception 8.3 How Charts Mislead 8.4 One-Way, Two-Way & Partial Dependence Exhibits

9 Chapter 9: Model Interpretation & Presentation — Communicating the Analysis

9.1 Communicating Methodology & Modeling Decisions 9.2 Explaining Why a Variable Was or Was Not Included 9.3 Communicating to Non-Technical Audiences 9.4 The PCPA Technical Report: Structure & Rubric

1.2 The PCPA Project: Windows, Deliverables & Grading

Key Takeaways Explain This Topic AI

The PCPA Project runs in four windows a year, each opening on the 16th of March or December and the 15th of June or September and closing on the last day of that month.

The technical report is capped at 1,250 words inclusive of the report and its appendices, and exceeding the limit is an automatic failed attempt.

Candidates submit code written in R, Python or SAS, up to five tables or supporting graphics as appendices, and an attestation of independent work.

The project fee is $700 per attempt and candidates may re-register within one year of their exam pass date, after which they must pass the exam again.

Generative AI may be used for general concept learning but not to interpret the business problem, analyse the data, or draft, edit or review any part of the submission.

Last updated: September 2026

The PCPA Project

Quick Answer: The project is a remote, open-book, independent modelling assignment. CAS gives you a business problem and one or two data sets; you build a GLM , then submit a technical report of no more than 1,250 words including appendices , your code, up to five tables or graphics, and an attestation. It costs $700 per attempt and is graded pass/fail by humans against a rubric.

You may only register for a project window after your official exam pass is posted to the candidate portal . Because official results take 15 days and registration closes about a week before the window opens, CAS states the practical rule plainly: plan to pass the exam at least three weeks before the window you want.

The Four Annual Windows

  • Exam deadline Registration deadline Window opens Submission deadline Results
  • February 23 March 9 March 16 March 31 May 31
  • May 25 June 8 June 15 June 30 August 31
  • August 25 September 8 September 15 September 30 November 30
  • November 25 December 9 December 16 December 31 February 28/29

The Exam Deadline column is the last day you can sit the exam and still be eligible for that window. Results are released roughly 6-8 weeks after the submission deadline. CAS states there are no extensions and no exceptions , and the religious/national holiday policy that applies to the windowed MAS and 5-9 exams does not apply here.

Each window is roughly two weeks of calendar time. That is not two weeks of full-time work for most candidates, so treat the window as a project-management exercise: read the business problem on day one, finish exploratory data analysis and a first GLM inside the first weekend, and reserve the final third of the window for refinement, exhibits, and the word-count fight.

What You Receive and What You Submit

At the start of the window, through the project portal, you receive:

  • a statement of the business problem ,
  • one or two data sets to explore and model,
  • scope parameters for the project, and

guidelines on what to submit for grading.

Your submission must contain:

A technical report, maximum 1,250 words inclusive of the report and appendices , describing how you conducted exploratory data analysis and handled data issues, how you built and improved a GLM, and how you interpreted the findings both technically and for the business decision — including your rationale for building the model.

Your predictive analytics code in R, Python or SAS . The code is not scored, but it may be run by graders or reviewed for originality, and it should reproduce your data preparation and modelling results programmatically.

Up to five tables or supporting graphics as appendices.

A Candidate Agreement and Project Attestation confirming the work is your own.

Answers to a set of questions about your data, process and outputs, posed at the point of submission.

Warning

Three hard limits cause an automatic failed attempt, independent of model quality: exceeding the 1,250-word cap, submitting more supporting materials than allowed, and not following the file-format instructions. A brilliant model in a 1,400-word report fails. Count words including appendices, and consult the final checklist in the project instructions before uploading.

The Grading Rubric

The project is human-scored against a rubric built from the same three domains as the exam, but weighted 30% / 30% / 40% . The published performance criteria tell you exactly what a grader looks for:

  • Criterion Domain (task)
  • Describes how variables were transformed and why Dealing with Data (A-3)

Describes anomalous data characteristics such as outliers and missing data, and how they were addressed Dealing with Data (A-4)

The model performs reasonably well on an assessment data set Model Diagnostics & Selection (B-1)

Correctly interprets diagnostics such as AIC, BIC and Type I errors, spurious relationships, multicollinearity and correlated variables, and uses them to improve fit Model Diagnostics & Selection (B-2)

  • Describes which model was selected and why Model Diagnostics & Selection (B-2)
  • Generates and provides technical output, such as model coefficients Model Diagnostics & Selection (B-2)

Explains the iterative build, including which diagnostics were used and why Model Diagnostics & Selection (B-2)

Justifies the choice of data and visual presentation: labelling, clarity of purpose, fit to data, model, audience and business question Model Interpretation & Presentation (C-1)

  • Describes what methods were used and why Model Interpretation & Presentation (C-2)
  • Describes why a variable was or was not included Model Interpretation & Presentation (C-2)
  • The model works and is appropriate for the business question Model Interpretation & Presentation (C-3)

Makes a persuasive argument to a non-technical audience, including the rationale for building the model Model Interpretation & Presentation (C-3)

Read that list as a report outline . Every criterion is a sentence or short paragraph you owe the grader, and with only 1,250 words to spend, roughly 100 words per criterion is the budget. That is why vague prose is so expensive here.

Independence, Open Book, and Generative AI

The project is open book. You may consult colleagues for general questions, reference materials, and online resources. What you may not do is let any of them do the work.

CAS permits generative AI only for general learning or clarification of concepts unrelated to the specific business problem. It explicitly prohibits using AI to:

  • interpret or solve the specific business problem,
  • analyse the project data,
  • develop calculations, approaches, recommendations or conclusions,
  • draft, rewrite, summarise, edit or improve project responses, or

review the work for correctness or completeness.

All PCPA instructions, prompts and data sets are proprietary to CAS and may not be copied, uploaded, shared, or entered into a generative AI tool. Security measures include plagiarism software, grader execution of submitted code, and the submission questionnaire.

Retakes

A failed project can be re-registered for $700 per attempt, and a candidate may attempt it multiple times within one year of the date they passed the PCPA Exam . The last opportunity to register falls on or before that one-year mark; after that, the candidate must pass the exam again before re-registering for the project.