IAI SA3 common mistakes
IAI SA3 · Understand the traps and how to avoid them.
Answering a module-design or documentation question with general actuarial reasoning instead of naming the specific SA3 deliverable (for example a model specification, an assumption log, or a validation report) and stating what it must contain.
Candidates carry over habits from core actuarial exams where a defensible calculation earns marks, and underestimate how heavily SA3 rewards structured communication of technical work to a specified audience.
What to do instead
Before writing, identify the audience (peer reviewer, regulator, board, client), the document type requested, and the required sections. Practise drafting short specification and validation documents against a timer and self-mark against a checklist of purpose, scope, data, assumptions, method, limitations, and sign-off.
Applying a standard technique without first stating the business problem, the decisions it supports, and the constraints, so the answer reads as a textbook exercise rather than a solution.
Under time pressure candidates jump to formulas and models, and SA3 scenarios are deliberately open so that framing is where marks are earned.
What to do instead
Open every answer with three or four sentences setting out the objective, the decision-maker, the time horizon, the data available, and the key constraints. Then justify each modelling choice against those constraints rather than against convention.
Presenting a single point estimate or a single scenario as the answer, with no sensitivity analysis, no discussion of model risk, and no statement of what would change the conclusion.
Candidates treat sensitivity work as optional garnish and run out of time on the main calculation, so it is dropped first.
What to do instead
Build a short sensitivity section into every answer as standard: vary two or three material assumptions, show the range of outcomes, identify the assumption the result is most sensitive to, and state the practical implication for the decision.
Describing data as clean or appropriate without naming the specific checks performed and what was done about exceptions, and without stating the limitations of the data for the intended purpose.
Data quality is treated as a box-ticking preamble, and candidates lack a rehearsed checklist of concrete checks to name.
What to do instead
Memorise and deploy a named checklist: completeness, accuracy, consistency, timeliness, granularity, credibility of source, reconciliations, outlier treatment, and handling of missing values. For each, state the finding and the action taken, then state the residual limitation.
Ignoring the regulatory, professional-standards, and governance context, so the answer never mentions documentation standards, peer review, model change control, or the applicable supervisory expectations.
Candidates focus on the mathematics and forget that SA3 scenarios are usually set in a governed environment where process and evidence matter as much as the result.
What to do instead
For each scenario, write down who must approve the work, what evidence must be retained, what independent review is needed, and which standards or supervisory expectations apply. Include a short governance paragraph in every long answer.
Writing long unstructured prose with no headings, no signposting, and conclusions buried in the middle, so the marker cannot find the points being made.
Candidates write as they think, and exam pressure encourages a single stream of consciousness rather than a structured document.
What to do instead
Use headings, numbered points, and a short summary at the start or end. Draft a skeleton of section headings before writing content, and keep one idea per bullet. Practise under timed conditions so the structure becomes automatic.
Misallocating time: spending most of the paper on one long modelling question and leaving later questions with only a few lines, or writing far too much on low-mark parts.
Candidates are drawn to the calculation they feel confident about and lose track of the mark-to-minute relationship across the whole paper.
What to do instead
Before starting, convert each question's marks into a target time and write it in the margin. Stick to it, leave space to return, and deliberately write a shorter, structured answer for low-mark parts rather than an essay.
Failing to answer the actual verb of the question, for example giving an explanation where an evaluation or recommendation was asked for, or listing advantages without a conclusion.
Candidates read the scenario topic and start writing about it, rather than parsing the instruction word and the number of required points.
What to do instead
Underline the instruction word and the number of items requested in every question. Check the finished answer against them, and add an explicit concluding recommendation or judgement wherever the verb implies one.
Treating the scenario as generic and not using the specific facts given, so the answer would be equally valid for any company in any market.
Candidates rely on memorised templates and do not extract the distinctive details (product features, distribution channel, capital position, history) that the question deliberately supplies.
What to do instead
Before answering, list the three or four scenario facts that must change the answer, and explicitly refer to them in the response. If a sentence would survive unchanged in a different scenario, cut it or make it specific.
Omitting the practical implementation dimension: no discussion of how the work would be delivered, by whom, with what systems, over what timeline, and how it would be maintained and reviewed after delivery.
Candidates stop at the technical result and do not think of themselves as the person who must get the work used and kept up to date.
What to do instead
Add a closing implementation paragraph to every substantial answer covering ownership, resource and systems needed, milestones, dependencies, communication to stakeholders, and the review or refresh cycle.