Since January 1, 2026, every candidate pursuing the ACAS designation must pass both parts of the CAS's Property and Casualty Predictive Analytics requirement, a $300 Pearson VUE exam and a separate $700 take-home project. 67.5% of the 126 candidates in PCPA's first mandatory administration passed both components.

No prior CAS requirement has combined an on-demand computer-based test with a quarterly, human-graded independent project. The first cohort is navigating that structure with almost nothing to benchmark against.

Key Takeaways

  • 85 of 126 candidates passed both components in Spring 2026, a 67.5% combined rate, with no component-level breakdown published showing how many failures came from the exam against the report.
  • Two weeks against three months. A failed exam can be re-sat after a fortnight. A failed project waits for the next quarterly window, then six to eight weeks for a result.
  • Exactly 1,000 words. "Any project submission that goes beyond 1,000 words will not be scored and will result in a failed attempt," the CAS states. It is a scoring rule, not a style guideline.
  • 80 hours against 20. The CAS estimates the exam at 80 hours of preparation and the project at roughly 20, which makes the project the lighter lift on paper and the tighter one in practice.
  • 10 to 15% is the salary premium DW Simpson attributes to Python, R and SQL fluency, stacking on 15 to 25% for the credential itself. The mandate folds part of that into the ACAS floor.

Two Instruments, One Requirement

The CAS finalized the mandatory transition after extending the original November 1, 2025 deadline to January 1, 2026, giving candidates a final window to complete the legacy ACAS requirements without touching PCPA. Anyone who cleared that slate is grandfathered. Everyone else has PCPA inserted into the pathway as a required, sequenced pair.

The exam half is a two-hour, computer-scored test administered continuously at Pearson VUE centers, with an estimated 80 hours of preparation, no formula sheet, and a $100 no-show fee. It covers GLM diagnostics and distributional selection across Tweedie, Poisson with log-exposure offset and Gamma, tree-based model tuning across Random Forest and XGBoost, variable selection including LASSO, and a situational ethics section built around ASOP 12, 23, 41 and 56 alongside the NAIC Model Bulletin and state instruments like Colorado SB 21-169. A candidate can sit it three times within a rolling twelve months, with a mandatory two-week wait between attempts.

The project is a different instrument entirely. Once the exam is passed, a candidate has up to one year to register before the exam credit expires. The project runs on four fixed windows a year: CAS distributes a dataset and a business problem, and the candidate has two weeks, working independently with no hands-on help from another person, to submit a technical report capped at exactly 1,000 words, working model code in R, Python or SAS, and up to five supporting tables.

The cap is a scoring rule. "Any project submission that goes beyond 1,000 words will not be scored and will result in a failed attempt," the CAS states in its FAQ. Generative AI tools may be consulted for general syntax or conceptual questions, but candidates are barred from sharing the project instructions or dataset with any AI system. Results arrive six to eight weeks after the submission deadline.

The Calendar Is the Binding Constraint

The two components fail in structurally different ways, and the difference is worth more planning attention than the modeling content.

Run the timeline on a badly sequenced failure. A candidate who passes the exam in January and registers for the next available project window sits that quarter's two-week submission period. If the attempt fails, registration does not reopen until the following window roughly three months later, then another six to eight weeks pass before a result. A single failed project attempt at the wrong point in the calendar pushes a target ACAS date back three to six months, not because the candidate needs that long to close a modeling gap but because no faster retry path exists.

That is a different risk profile from a written-answer exam like MAS-II, where a fall failure means sitting again the following spring on a fixed schedule candidates plan around a year ahead. The hedge is treating the exam-to-project handoff as one planned sequence: registration for a window closes roughly three weeks before it opens, so passing the exam with room to spare buys entry to the very next administration rather than the one after it.

The effort estimates point the same way. The exam carries 80 estimated hours; the project roughly 20, spread across preparing, drafting, revising and finalizing inside a two-week window. Compressed timelines punish process failures more than knowledge gaps. A candidate who has not pre-allocated the 1,000-word budget across problem framing, model selection rationale and validation discussion burns time mid-window discovering the draft is 1,400 words with the section explaining why this model over the alternatives still unwritten.

Sequencing carries the same logic. PCPA has no formal prerequisites, but the CAS is explicit that candidates are "highly encouraged to pass MAS-I, MAS-II and Exam 5 before attempting the exam". MAS-I and MAS-II supply the GLM and ensemble foundation the exam tests at a deeper level than either alone.

Exam 5 supplies something the project needs specifically. Two candidates can build technically identical models on the same dataset and produce materially different reports, because the one without ratemaking context describes fit in deviance and AIC terms without connecting selected variables to a rate indication or a classification plan. Graders score against a rubric rewarding actuarial communication alongside technical correctness, so a report reading as a statistics exercise is a real risk.

Nothing to Benchmark Against

The Spring 2026 sitting is the only data point for candidates who faced PCPA as a true requirement, and the CAS has not published what would make it usable.

Eighty-five of 126 candidates passed, tracked on our PCPA pass-rate page, but no component-level breakdown shows how many failures came from the computer-based test against the technical report. Quarterly project statistics, with exam statistics reported annually, will not begin until enough sittings accumulate under the mandatory regime. A candidate entering the project in late 2026 is calibrating against one published number, a very different position from approaching Exam 5 with a decade of pass-rate history.

The adjacent credential does not fill the gap either. The CAS Institute's Certified Specialist in Predictive Analytics predates PCPA and remains separate and deeper, designed for experienced practitioners already in quantitative roles rather than as an entry-level pathway.

Feature PCPA Project CSPA Project
Technical report length Up to 1,000 words Up to 3,000 words
Time to complete 2 weeks 60 days
Business case and dataset Basic More advanced
Mentor option Not available Available, with resubmission window

The two intersect at exactly one point. Candidates who passed CSPA Course 3, Predictive Modeling: Methods and Techniques, along with its case study project, and did so by sitting the course rather than obtaining credit through an IFM exam waiver, may petition for a full waiver of both PCPA components. That matters for candidates who came up through a data science track before entering the ACAS pipeline. For a traditional exam-sequence candidate, pursuing CSPA to waive PCPA means taking a harder, longer credential to avoid an easier one.

What the mandate does settle is the market position. DW Simpson's 2026 survey shows actuaries fluent in Python, R and SQL earning 10 to 15% more than peers without those skills, a premium stacking on the 15 to 25% tied to an ASA or ACAS credential itself. Making predictive analytics a baseline requirement rather than a specialty differentiator folds part of that premium into the credential's floor, so a candidate who previously stood out by pursuing CSPA or building side projects now competes in a pool where every ACAS holder has demonstrated the same competency through a standardized, externally graded instrument.

CAS Exams SOA vs CAS Actuarial Careers

Further Reading

Sources