IAI CP3 common mistakes
IAI CP3 · Understand the traps and how to avoid them.
Writing the answer to a CA1-style long question as a list of unconnected bullet points that restate the scenario, without linking each point back to the specific facts of the case study or to the actuarial principle being tested.
Candidates memorise topic lists during revision and then 'brain-dump' them under time pressure, hoping markers will find enough correct points. They do not practise applying the syllabus to unseen scenarios, so they never learn to select and tailor points.
What to do instead
For every practice question, force yourself to write a one-line 'so what' after each bullet that names the case-specific fact it addresses (e.g. 'because the scheme is mature and cash-flow negative, ...'). Before writing, spend two minutes drafting a short issue tree and strike out any point that does not connect to it. Mark your own answers against the command word (state, explain, discuss) and delete material that does not answer it.
Ignoring the command word and the marks available: writing three sentences for a 'state' question worth two marks, then running out of time on a later 'discuss' question worth twelve marks, leaving it half-finished.
Candidates treat all questions as equally important and write at a constant level of detail. They also over-invest in the first question they can answer confidently, which feels productive but destroys the time budget.
What to do instead
Before starting, allocate minutes proportional to marks (roughly one minute per mark plus reading time) and write the allocation next to each question. Practise with a timer and a rule: 'state' means a short phrase per point, 'explain' means point plus reason, 'discuss' means a balanced argument with both sides. If a question is running over, stop and move on, noting where to return.
Producing a capital or reserving calculation that is arithmetically correct but uses the wrong basis: applying a solvency capital requirement to an accounting reserve, mixing gross and net of reinsurance, or using a discount rate inconsistent with the stated regulatory regime.
Candidates learn formulas in isolation and do not build a mental map of which basis applies to which purpose. Under exam pressure they grab the most familiar formula rather than the one the question asks for.
What to do instead
Build a one-page comparison table for each calculation family (best-estimate vs prudent, gross vs net, statutory vs internal model) and rehearse labelling every input with its source and basis before computing. In the exam, write the basis in words at the top of the answer and check each number against it. Show the formula and intermediate steps so partial credit survives a slip.
Treating a data science question as a coding exercise: describing a model's mechanics (gradient boosting, cross-validation) without addressing data quality, governance, interpretability, or how the output feeds a business or regulatory decision.
Candidates with technical backgrounds over-index on methods they know, while the exam tests judgement about fitness for purpose, limitations, and controls. They under-estimate how much of the mark sits in the 'so what for the insurer' discussion.
What to do instead
Answer in a fixed structure: business objective, data (source, quality, bias, missingness), method and why it suits the objective, validation and monitoring, limitations and interpretability, governance and sign-off. Practise writing the limitations and governance sections first, because they are the most commonly skipped and the easiest marks to secure.
Writing a professionalism or ethics answer that says 'actuaries should act ethically and follow the code' without identifying the specific threat, the affected party, and the concrete safeguard or escalation route.
Candidates treat professionalism as common sense and do not learn the framework of threats and safeguards. They also avoid naming uncomfortable actions, such as escalating to a supervisor or declining to sign, for fear of sounding confrontational.
What to do instead
Learn and apply a threat-and-safeguard structure to every scenario: identify the threat (self-interest, familiarity, intimidation, self-review), state who is harmed (policyholders, employer, profession), then give a sequenced response (disclose, document, seek advice, escalate, decline to certify). Rehearse two or three short scenarios aloud until the structure is automatic.
Answering a communications question by producing technically complete content that ignores the audience: using jargon with a policyholder letter, omitting the key decision from an executive summary, or failing to state assumptions and uncertainty in a board paper.
Candidates equate a good answer with a comprehensive one. They do not practise adjusting tone, structure, and length for a named reader, and they forget that a communication is judged on whether the reader can act on it.
What to do instead
Before writing, state the audience, their technical level, and the one action or decision you want from them. Then structure accordingly: bottom line first for executives, plain language and 'what this means for you' for customers, assumptions and sensitivity for technical peers. Practise rewriting the same content for two different audiences and compare.
Describing a product, market, or regulation generically without the jurisdiction-specific detail the question requires, or asserting a rule or figure that is out of date or invented.
Candidates rely on vague memory and fear leaving a blank, so they pad with plausible-sounding statements. They also revise from outdated summaries and never check which regime (for example a specific solvency or pensions framework) the question is set in.
What to do instead
Anchor each answer to the regime named in the question and, where unsure of a figure, state the principle and flag the assumption rather than inventing a number. Keep a living one-page summary per regime with the date you last verified it. In practice, answer only what you can support and use the remaining time on questions where you have solid knowledge.
Ignoring asset-liability interaction: recommending an investment strategy or product design on the basis of expected return alone, without matching duration, currency, liquidity, or cash-flow timing to the liabilities.
Candidates study investments and liabilities as separate syllabus items and do not internalise that most real recommendations fail on mismatch, not on return. Under time pressure the simple 'higher return is better' answer feels sufficient.
What to do instead
For every investment or product recommendation, write the liability profile first (timing, currency, uncertainty, liquidity needs) and then test the proposal against each dimension. Practise short 'match or mismatch' drills and always include a sentence on the consequence of mismatch and the hedging or diversification response.
Presenting a model output as a single number with no discussion of assumptions, sensitivity, or model risk, and no acknowledgement of what would change the conclusion.
Candidates want to demonstrate they can compute the answer and treat caveats as weakening it. They also run short of time and cut the discussion section, which is where the judgement marks sit.
What to do instead
Adopt a fixed reporting template: purpose, key assumptions and their justification, results, sensitivity to the two or three most material assumptions, limitations and model risk, and recommended action. Even in a short exam answer, include one sentence on sensitivity and one on limitations; these are reliably rewarded.
Misreading the business context: answering a general insurance question with life insurance framing, or a health and care question with pure life-mortality logic, because the underlying maths looks similar.
Candidates revise by technique rather than by practice area, so the distinctive features of each line (claims frequency and severity, morbidity and recovery, incidence versus prevalence, reserving for incurred but not reported claims) get blurred together.
What to do instead
Before answering, write the practice area and its two or three defining features at the top of the page: for general insurance, claim frequency/severity and IBNR; for health and care, incidence, recovery, and benefit design; for life, mortality, longevity, and policyholder behaviour. Build a comparison grid across the practice areas and test yourself on which technique belongs where.