A representative sample in an underwriting audit is a selection of files that accurately reflects the full population of risks a coverholder, MGA or other delegated underwriter has written, so that findings from the sample can be reasonably applied to the wider portfolio. It must mirror the book's mix of products, territories, brokers, limits and time periods. Representativeness depends on how files are chosen, not only on how many.
For delegated authority managers, this distinction matters because a well-built sample supports confident conclusions about underwriting control, while a poorly built one can mask problems or exaggerate them. The sections below cover sample size, selection method, risk-based testing, common sampling flaws and how to interpret what the sample shows.
How many files make a sample representative?
There is no single number that makes an underwriting audit sample representative. Sample size depends on the size and complexity of the portfolio, the number of products and classes, the level of risk the arrangement carries and the confidence the insurer needs. Representativeness comes from coverage of the book, not from a fixed percentage.
A coverholder writing a few hundred policies in one class needs a different approach from an MGA writing thousands of risks across several binders and territories. In practice, auditors weigh several factors when setting the underwriting file sample size:
- The total number of bound risks in the audit period
- The number of distinct classes, products and territories
- The size of the delegated authority and the premium written
- Results from previous audits and any known control weaknesses
- Any minimum expectations set by the capacity provider or market
A small sample drawn thoughtfully across the book will usually tell you more than a larger sample drawn from one corner of it. Where a portfolio is very small, auditors may review every file, which removes sampling questions altogether.
How do auditors select files for an underwriting audit?
Auditors select files by first obtaining a complete bordereau or policy listing for the audit period, then drawing a sample from that full population using a defined method. Selecting from the auditor's own list, rather than one supplied pre-filtered by the coverholder, protects the independence of the review.
A typical selection process follows these steps:
- Request the full risk bordereau for the period under review, covering all binders and products.
- Check that the listing is complete by reconciling it to premium and policy counts reported to the insurer.
- Segment the population by class, territory, broker, premium band and inception date.
- Draw files from each segment, using random selection, targeted selection or a combination.
- Agree the file list with the insurer, and release it to the coverholder only as late as practicable.
Limiting advance notice of the exact files matters. If a coverholder can see which files will be reviewed, it can tidy those records before the auditor arrives, which defeats the purpose of sampling. Reconciling the bordereau first also confirms that the population itself is complete, since a sample can only be representative of the data it is drawn from.
Should an underwriting audit sample be risk-based?
Yes, in most delegated authority audits the sample should be risk-based, at least in part. Risk-based sampling directs more attention to files where errors would be most costly or most likely, such as large limits, referrals, unusual risks or outliers, while still including a spread of ordinary business to test routine controls.
Pure random sampling treats every file as equal, which is rarely true in insurance. A single high-limit risk written outside authority matters far more than a routine policy with a minor documentation gap. For this reason, many auditors build a blended sample that combines two elements:
Targeted selections
These cover risks that carry higher exposure or indicate possible control failures. Examples include policies near or above authority limits, risks requiring referral to the insurer, unusual classes of business, high loss ratio accounts, late or backdated binding and risks placed through a new or unfamiliar broker.
Random selections
These cover the remaining routine business. Random selection tests whether everyday controls work consistently and prevents the auditor's judgement from steering the sample away from areas that look unremarkable but may still contain weaknesses.
Combining both approaches is what gives risk-based sampling its value in a binder audit. The targeted files focus on exposure, while the random files give a fairer picture of normal practice.
What makes an underwriting audit sample unrepresentative?
A sample is unrepresentative when it systematically over- or under-weights part of the portfolio, so its results cannot be fairly applied to the whole book. Common causes include selecting from a limited time window, relying on a coverholder-supplied file list, ignoring minority products or territories, and drawing only from the largest accounts.
Typical weaknesses include:
- Time bias: Drawing files only from the most recent months, which may miss earlier control failures or seasonal patterns
- Selection by the audited party: Letting the coverholder choose or pre-filter the files reviewed
- Product or territory gaps: Covering the main class while omitting smaller lines that may be written with less oversight
- Premium bias: Reviewing only the largest risks and overlooking weaknesses in high-volume, low-premium business
- Incomplete population: Sampling from a bordereau that does not reconcile to reported premium
An MGA writing niche products alongside a core account is a useful example. If the audit sample draws almost entirely from the core account, the niche lines, where underwriting guidelines may be least developed, go largely untested.
How should audit findings from a sample be interpreted?
Audit findings from a sample should be read as evidence of how well controls operate across the portfolio, not as a complete list of every error. The pattern and cause of findings matter more than the raw count, and results from a well-designed sample indicate the likely scale of an issue in the wider book.
When reviewing a report, delegated authority managers should look at several things beyond the headline number of exceptions:
- Whether errors are isolated or repeated across files, brokers or underwriters
- Whether the root cause is a process gap, a training issue or a system limitation
- How significant each finding is in terms of financial exposure, contractual breach or regulatory impact
- Whether the issue is confined to one product or territory or likely to affect the whole binder
A clean sample does not guarantee that no issues exist elsewhere, and a failed file does not always point to a systemic problem. Sound interpretation means weighing the sampling method alongside the results. A clearly written report should explain how files were chosen, what the sample covered and what the findings suggest, so that the insurer can decide on proportionate follow-up such as remediation, re-testing or a wider review.
How Otter Audit helps with underwriting audit sampling
Otter Audit designs underwriting audit samples that reflect the real shape of a delegated portfolio, so insurers can rely on the findings. Our ACII-qualified auditors combine technical judgement with practical experience gained across more than 500 firm visits in 50 countries.
- Sample design that blends risk-based and random selection across products, territories and brokers
- Independent file selection from reconciled bordereaux
- Clear reports that explain findings, their significance and the reasoning behind recommendations
- Consistent, on-time delivery against agreed deadlines
To see how we approach delegated underwriting reviews, explore our delegated authority audits, or contact our team to discuss an audit for your coverholders or MGAs.