IAI SP6 common mistakes
IAI SP6 · Understand the traps and how to avoid them.
Treating 'fraud' as a single concept and defining it loosely (e.g. 'dishonest behaviour by a claimant'), instead of distinguishing fraud from abuse, error, exaggeration, and opportunism, and without anchoring to the elements of deception and intent to gain.
Candidates learn the topic through claims anecdotes and conflate the lay meaning of fraud with the technical definition. They also skip the distinction between criminal fraud and civil/contractual fraud, which changes the burden of proof and the response options.
What to do instead
Before answering, write a two-line working definition: a dishonest act or omission intended to obtain a financial advantage (or cause a loss) by deception. Then classify the scenario on a spectrum: error, abuse/opportunism, exaggeration, planned fraud, organised fraud. Practise sorting 10 short scenarios into these categories and justify each with the specific element that decides it.
Reciting typologies as a list of labels (e.g. 'staged accident', 'fronting', 'ghost employee') without being able to explain the mechanism, the point in the lifecycle where the fraud occurs, and which control detects it.
Typology names are memorable and easy to revise from summaries, so candidates memorise labels rather than the underlying scheme. Exam questions often present a scenario and ask which typology fits, or ask where a typology sits in underwriting versus claims.
What to do instead
Build a one-page table with columns: typology, who commits it, at which stage (application, underwriting, claim, renewal, broker/agent), the false statement made, the loss mechanism, and the detection signal. Write the mechanism in one sentence in your own words for each typology. Test yourself by describing the scheme without using its label.
Answering legal-framework questions by instinct ('that's obviously a criminal offence', 'we must report it to the regulator') without identifying the specific legal basis, the applicable jurisdiction, and the threshold or duty that triggers action.
Fraud law is jurisdiction-specific and layered: criminal fraud offences, insurance-specific statutory provisions, regulatory conduct rules, data protection constraints, and anti-money-laundering reporting duties all interact. Candidates revise principles but not the source of each duty, and they mix civil recovery with criminal referral.
What to do instead
Create a source map for your jurisdiction: offence provisions, insurance statute/regulation, regulator conduct rules, AML reporting duty, data protection rules on sharing investigation data, and employment/contract law for agents. For each, note who it binds, what triggers it, and the consequence of breach. When answering, name the instrument and the trigger rather than asserting a general duty.
Assuming every red flag is evidence of fraud and writing answers that treat indicators as proof, or conversely dismissing indicators because innocent explanations exist.
Red-flag lists are taught as checklists, so candidates equate presence of an indicator with guilt. The exam tests whether you understand indicators as probabilistic signals that shift suspicion and justify further enquiry, not conclusions.
What to do instead
For each indicator, practise the paired response: 'this raises suspicion because X; it can also be explained by Y; therefore the next step is Z'. Always state the benign alternative and the verification step. Use the language of correlation and escalation, not proof, and reserve the word 'evidence' for material that has been verified and documented.
Confusing underwriting-stage and claims-stage indicators, or applying claims red flags (e.g. claim timing, repair estimate inflation) to an underwriting scenario and vice versa.
Many indicators look similar across stages, and candidates revise a single undifferentiated list. The exam deliberately places indicators in the wrong lifecycle context to test whether you know when the control can act.
What to do instead
Split your indicator notes into three columns: application/underwriting (non-disclosure, misrepresentation, identity, sum insured mismatch, intermediary behaviour), claims (loss circumstances, timing, documentation, provider involvement), and post-claim/renewal (repeat patterns, cancellation after claim). For each, note what information is available at that stage and which control can act on it.
Describing an investigation as a sequence of activities ('interview the claimant, get documents, refer to police') without addressing evidence handling: continuity/chain of custody, contemporaneous notes, secure storage, lawful acquisition, and preservation of digital material.
Candidates focus on the investigative narrative and treat evidence handling as administrative detail. In practice, mishandled evidence destroys the case and can create legal exposure, which is what the exam is testing.
What to do instead
Learn a standard evidence lifecycle: identify, lawfully obtain, record (what, where, when, from whom), preserve integrity (sealed, logged, access-controlled), maintain continuity, and disclose appropriately. For every investigative scenario, add two sentences on how the evidence would be recorded and preserved, and flag any material that must not be altered (e.g. original documents, device data).
Writing interview answers that jump to confrontation or accusation, or that ignore the distinction between an information-gathering interview and an interview under caution/formal recorded interview, and the safeguards attached to each.
Candidates model interviews on dramatic confrontation and forget that the purpose is to obtain accurate, admissible information. They also overlook voluntariness, the right to representation, cautions or equivalents, and the risk of coercion invalidating a statement.
What to do instead
Structure every interview answer as: objective, status of interviewee (witness, claimant, employee, suspect), legal basis and warnings required, preparation (documents, timeline, open questions), conduct (open then funnel, no leading, no promises or threats), and close (statement taken verbatim, read back, signed, dated). State explicitly when a formal caution or legal advice would be required.
Treating fraud risk management as a list of controls (hotline, referral team, analytics) without linking them to a risk assessment, ownership, proportionality, and measurable outcomes, or without recognising the tension between fraud control and customer fairness.
Control lists are easy to memorise and appear comprehensive. Candidates miss that exam questions test governance: who owns the risk, how appetite is set, how controls are tested, and how false positives and customer detriment are managed.
What to do instead
Frame answers as a cycle: risk assessment and appetite, preventive controls (vetting, verification, wording, incentives), detective controls (analytics, audit, tip-offs), response (investigation, recovery, referral), and governance (ownership, reporting, testing, lessons learned). For each control, state the risk it addresses, its cost/false-positive trade-off, and how effectiveness would be measured.
Describing data analytics as a magic detection layer ('use AI to spot anomalies') without specifying the data, the rule or model, the threshold, the output, and the human review that follows.
Candidates lack hands-on analytics experience and default to buzzwords. The exam tests whether you understand that detection depends on data quality, labelled outcomes, base rates, false positives, and a defined triage process.
What to do instead
For any analytics question, specify: data sources and fields, the detection logic (rule, anomaly score, network/link analysis, text mining), the threshold and why, the expected false-positive rate, the triage and referral path, and the feedback loop that refines the model. Mention governance: data protection, explainability, bias testing, and audit of model changes.
Ignoring conduct and professional-standards dimensions when answering operational questions, e.g. recommending covert surveillance or data sharing without addressing proportionality, privacy, conflicts of interest, or the duty to act in good faith towards the customer.
Candidates treat ethics as a separate topic and do not integrate it into investigation, analytics, or underwriting answers. Exam scenarios often embed a conduct dilemma inside a technical question, and marks are lost when it is not addressed.
What to do instead
Adopt a habit of adding a short conduct check to every answer: is the action lawful, proportionate, non-discriminatory, documented, and consistent with treating the customer fairly? Identify conflicts (e.g. investigator incentivised by savings), privacy limits on data use, and the need for escalation and independent review. State the check explicitly rather than assuming it is implied.