A premium leakage audit starts with an uncomfortable premise: a meaningful share of the policies on the book are priced on facts that are no longer, or never were, true. Cars garaged somewhere else. Drivers who moved in, grew up, or started delivering food. Mileage bands from a commute that ended two jobs ago. The audit compares policy data against external records and internal claim signals, scores the mismatches, and works them in order of premium impact. What surprises executives the first time is the composition of the findings. Deliberate fraud is a minority of the dollars. The majority is drift: honest changes never reported, agent workarounds at point of sale, data entry habits that round every risk down. That composition matters because it decides the remedy. Fraud wants investigators. Drift wants better questions at renewal, data prefill that shows the customer what the carrier already knows, and agent incentives that stop paying for bound policies and start paying for accurate ones.
Where the dollars actually hide
Unrated drivers are usually the largest single bucket in personal lines audits: a household member of driving age who appears in address linked data, has their own claim history somewhere, and is absent from the policy. Garaging territory mismatches come next, followed by business use signals, vehicles appearing in delivery or rideshare contexts while rated as pleasure use. Each bucket has a signature in third party data, and the audit's craft is in the false positive rate, because every wrongly accused customer who gets a premium hike letter is a retention loss and a regulator complaint with a name on it.
Commercial lines hide leakage differently: payroll and revenue classifications that drifted from the original application, operations described narrowly at binding that broadened in practice. The mechanism is the same, the application froze a business at birth and the business declined to stay frozen.
Related reading on this site: Home and Ghost Broking: The Policy Was Real. The Broker Was Not. cover the neighbouring problems in more depth.
Fix the intake, not just the book
Auditing the existing book recovers premium once. Fixing intake stops the leak refilling. The findings that should scare a carrier most are the ones traced to its own process: a quote flow that defaults garaging to the mailing address, an agent script that never asks about other household drivers because the question slows the close, a renewal that asks the customer to confirm details by showing them nothing. Each is a design decision with a leakage rate attached, and the audit that reports them as customer dishonesty is protecting the process at the customer's expense.
Prefill is the modern answer and it cuts both ways usefully. Show the applicant the drivers and addresses the data already associates with the household and ask them to correct it. Honest customers fix errors in seconds. Dishonest ones now have to actively contradict a record on screen, which is a different act, legally and psychologically, from omitting a fact nobody asked about.
Amnesty beats ambush
The carriers who recover the most run correction windows: a renewal notice that invites customers to update garaging, drivers, and use, with the premium change applied prospectively and no retroactive action for those who come forward. It feels soft and it outperforms enforcement, because drift customers, the majority, self correct when given a door, and the audit list shrinks to the minority who declined a free chance to be honest. That smaller list is a far better fraud investigation queue than the raw mismatch file ever was.
Sequence the whole programme that way: measure, invite correction, reprice what remains, investigate what still refuses to reconcile. The leakage number a board should ask for is not dollars found in the first sweep. It is the leakage rate twelve months later, after intake changed. A programme that finds the same dollars every year is not recovering premium. It is farming it.
Frequently asked questions
What is a premium leakage audit?
It is a review of a carrier's book comparing the facts policies are priced on, garaging, drivers, use, mileage, against external data and claim signals, to find underpriced risk. Findings are worked in premium impact order, with deliberate fraud separated from unreported life changes.
What does a leakage audit usually find most of?
Unreported change and process drift rather than deliberate fraud: household drivers never added, garaging addresses that moved, business use that grew out of a personal policy, and intake defaults that round risk down. Fraud is present but is typically the minority of dollars.
How do carriers reduce leakage without accusing good customers?
By prefilling applications with known data for the customer to correct, asking direct questions at renewal, and running correction windows that apply changes prospectively for customers who come forward. That shrinks the audit list to cases that genuinely need investigation.
Why do insurers care about leakage if loss ratios look fine?
Because leakage is mispricing, and mispriced books attract more of the underpriced risk while overcharging the accurately rated customers who then leave. Over time the book's composition worsens even if a single year's results look acceptable.