Exam SRM Sample Questions, Free and Interactive
SRM is the data-science actuarial exam. Where P and FM test math and finance with actuarial framing, SRM is where you start working with the toolkit a modern actuary uses for predictive modeling: generalized linear models, decision trees, time series, principal component analysis, clustering. The exam style is different too. More conceptual, less calculation-heavy, more "which method is appropriate here" than "compute this integral."
The SOA's published SRM sample set is smaller than the older preliminaries: 66 items as of 2026, the entire publicly-released body of sample work. The exam is younger (introduced in its current form in 2018) and the SOA has accumulated fewer publicly-released items. With 66 items, the sample set works best as a final-stretch calibration tool, not as primary practice.
FreeFellow hosts all 66 inside the practice surface with worked solutions alongside, plus an original SRM bank (1,000-plus questions) for the bulk of practice. Free, no signup to browse, no credit card.
Start practicing Exam SRM samples
What the 66 Sample Items Cover
The full topic distribution:
| Topic area | Approx. sample count | What you'll see |
|---|---|---|
| Generalized linear models | ~20 | Logistic regression, Poisson regression, link functions, deviance, log-likelihood interpretation |
| Decision trees | ~12 | CART, splitting criteria (Gini, entropy), pruning, ensemble methods, bagging/boosting |
| Time series | ~12 | Stationarity tests, AR, MA, ARMA, model identification, forecasting |
| Principal component analysis | ~8 | Variance explained, loadings, biplot interpretation, application to dimension reduction |
| Cluster analysis | ~6 | K-means, hierarchical clustering, dendrogram interpretation |
| Model selection and validation | ~8 | Cross-validation, AIC/BIC, bias-variance tradeoff, train/test splitting |
The recognizable patterns: GLM questions favor scenario problems where the candidate identifies the appropriate link function given response data structure (binary, count, continuous-positive). Time-series questions favor stationarity diagnosis and model identification from autocorrelation patterns. Tree questions favor splitting-criterion comparisons and ensemble logic. The exam style is unmistakably data-analytic rather than calculation-heavy.
Why the SRM Sample Set Is a Calibration Tool, Not a Primary Bank
Three concrete reasons specific to SRM:
The 66-item set is too small to drill repeatedly without memorizing the answers. Once you have seen each item once, the calibration value drops sharply.
The exam style rewards conceptual understanding far more than pattern-matching. Reading ISLR (the canonical SRM textbook, available free from its authors at statlearning.com) is a much better use of time than re-drilling 66 sample items.
The SRM live exam pacing is more forgiving than P/FM. You have time to think, which means raw practice volume matters less than for the calculator-heavy exams.
Save all 66 SRM samples for the final two weeks. Take them in a single 100-minute timed block (live exam pacing). If you score 70 percent or higher, you are ready. If below 60, schedule another week of topic review before sitting.
Three Tactics for the 66 Items
Read ISLR first, work samples second. ISLR (An Introduction to Statistical Learning, James/Witten/Hastie/Tibshirani) is the conceptual foundation that the SOA samples assume you have. Read chapters 3 through 10 before touching the sample set. Samples without ISLR foundation produce frustration rather than learning.
Write the model-selection reasoning before checking the solution. SRM rewards "which model is appropriate" reasoning. When you work a sample item, write down (on scratch paper or in a note) why you chose the model you chose. Then check the worked solution. Self-grade on the reasoning quality, not just the multiple-choice answer.
Focus on GLM link functions specifically. Of the 20 GLM items, roughly half hinge on identifying the appropriate link function (logit for binary, log for Poisson, identity for continuous-positive with constant variance). Mastery here disproportionately translates to live-exam points.
What the Samples Don't Cover Well
Honest gaps for SRM:
- Recent syllabus additions (2024+). Model interpretability methods (SHAP values, partial dependence plots) and fairness/bias metrics are syllabus topics with sparse sample coverage. Supplement from the FreeFellow original SRM bank, which is calibrated against the current syllabus.
- Real-world data-cleaning scenarios. The samples assume tidy data. Live exam can include questions where the answer depends on recognizing data-quality issues.
- Newer ensemble methods (gradient boosting, random forests). Sparse in older sample batches; better covered in the FreeFellow originals.
How SRM Practice Compares Across Free and Paid
| Source | SRM Sample Questions | Format | Cost |
|---|---|---|---|
| FreeFellow | All 66, interactive | Same surface as topic practice | $0 |
| ISLR (textbook) | None directly | Free PDF and printed book | $0 |
| ASM SRM Manual | All 66 in study manual | $130 to $200 | |
| Coaching Actuaries Learn + Adapt SRM | All 66 + originals + video lessons | Interactive within Adapt | $300 to $400 per exam |
| SOA only | All 66 | Static PDF on soa.org | $0 |
ISLR deserves the explicit callout: it is the most important free SRM resource, written by the canonical authors of the underlying statistical-learning material. The SOA samples calibrate your readiness; ISLR builds the conceptual base.
Start Practicing
Practice all 66 Exam SRM samples alongside the FreeFellow original SRM bank. Both are free. Pair with ISLR for the conceptual foundation.
For the broader actuarial exam prep landscape: Best Free Actuarial Exam Prep Resources (2026).