In Part 1 I covered what SRM tests and how the syllabus fits together. This post is about materials — what each type of resource is for, which official resources are free, and how to assemble a stack you’ll actually finish.
I’ll say upfront: there’s no single perfect set of SRM materials. The right combination depends on your statistical background, your time, your budget, and how you like to learn. What follows is the framework I used to decide, rather than a sales pitch for any one product.
Every resource needs a job
The mistake I see most often — and nearly made myself — is collecting resources: two manuals, three question banks, a YouTube playlist, someone’s notes from a forum. A long list feels like progress. It isn’t. Each item in your stack should have one clear job:
| Job | What fills it |
|---|---|
| Teach the syllabus | One study manual or course — one, not two |
| Provide enough practice | A question bank, plus the official sample questions |
| Settle disputes | The syllabus and official textbooks |
| Connect it all | Your own notes and error log |
If a resource doesn’t have a job, it’s a distraction. The best resource is the one you’ll use consistently to the end.
Start with what’s free and official
Before spending anything, collect these:
The syllabus for your sitting. Not last year’s. The SRM study page links the current one. The syllabus is the source of truth for topic weights, learning outcomes, and required readings — whatever a commercial provider emphasises, the syllabus decides what’s examinable.
An Introduction to Statistical Learning (ISLR). The September 2026 syllabus lists two textbooks: Frees, Regression Modeling with Actuarial and Financial Applications, and James, Witten, Hastie & Tibshirani, An Introduction to Statistical Learning (second edition). ISLR is legally free as a PDF from the authors’ website, and it’s genuinely one of the best-written statistics books at any price. If you’re coming from a data-science background, it will feel familiar; if you’re not, it’s still the clearest explanation of trees, ensembles, PCA, and clustering you’ll find. The syllabus specifies selected chapters, not the whole book.
The official sample questions. The SOA doesn’t release past SRM exams, but it publishes sample questions and solutions through the study page. My advice: don’t burn through all of them early. Save a portion for late in your preparation, as a check that your provider’s question style and emphasis actually match the SOA’s.
The study manual question
Most candidates use a commercial manual or course, and for SRM I think that’s the right call for one reason: the exam pulls from two textbooks plus additional readings, and a good manual condenses that into one sequenced path with practice attached. That’s worth paying for once — and only once. Buying a second manual because the first one felt hard is almost always a mistake; the difficulty is the syllabus, not the book.
When comparing providers, the things that mattered in practice were: whether explanations lead to worked questions immediately (not fifty pages later), whether the question bank tracks your performance by topic, how long the access period is, and whether the questions are written in the SOA’s interpretive style or are just calculation drills. Ask other candidates what the questions felt like relative to the real exam — that’s more informative than star ratings.
Question practice is the actual preparation
Whatever bank you use, the volume of questions matters less than what you do after getting one wrong. Every miss falls into one of four buckets:
- a concept I didn’t understand;
- a formula or definition I couldn’t recall;
- a misreading or interpretation error; or
- a time-management or calculation slip.
Each bucket has a different fix — rereading a chapter only repairs the first one. Logging which bucket each miss falls into is what turns a question bank from a scorekeeper into a study plan. I’ll go deeper on the error log in Part 3.
One warning while practising: providers’ difficulty scales are their own inventions. Don’t treat any practice score as a guaranteed pass threshold, and don’t compare your numbers with someone using a different bank.
Notes: small, comparative, and yours
I kept my own notes deliberately small. Not chapter summaries — the manual already is one — but the things that kept tripping me up:
- definitions that are easy to confuse;
- model assumptions and their diagnostics;
- side-by-side comparisons (lasso vs. ridge, bagging vs. random forests, k-means vs. hierarchical);
- common R-output patterns; and
- formulas I repeatedly forgot.
Near exam day, a one-page comparison sheet beats rereading a chapter, every time. The act of building it is most of the value, which is also why borrowing someone else’s notes never works as well as it should.
The stack I’d recommend
For someone starting SRM today, working full time like I was:
- One study manual or course to teach the syllabus, chosen for teaching style and question quality, not price alone;
- Its question bank plus the official sample questions for practice, with the official ones held back for the final stretch;
- ISLR (free) for any topic that needs a second, better explanation; and
- Your own comparison sheets and error log to tie it together.
That’s it. Four jobs, four resources. Anything beyond that should have to argue its way in.
Next in the series: How to actually pass — the study strategy
Sources: SOA SRM study page · September 2026 syllabus · ISLR free download. I purchased my own materials and this post contains no affiliate links.