Annual mileage15,000
Premium$1,535
Edits after the price0
Answer shopping
The applicant edits the same rating answer again and again while watching the price, then submits the one that worked. The final value looks ordinary. The edit trail doesn't.
Premium leakage & rate evasion
Applicants drop drivers, trim mileage and move the garaging ZIP until the price looks right. The final answers pass every record check. ForMotiv reads how those answers were given, while the quote is still open and before the policy binds.
$40–50Bestimated U.S. auto leakage a year, and probably a floor
~4xmore likely to claim in the first 30 days, for flagged populations
In sessionsignals return in milliseconds, before bind
Try it yourself
Change any answer before you see a price and ForMotiv reads it as an ordinary correction. Change it after, and the same edit means something else. Watch the right-hand panel as you go.
No behavior suggests the answers were changed to move the price.
Illustrative pricing and simplified signals. Real ForMotiv signals look at many more behaviors, like how often a field is revisited and how answers compare with prefill, and they never rely on a single edit.
The size of the problem
The industry still quotes Verisk's 2016 study: about $29 billion a year, roughly 14% of personal auto premium. Nobody has re-measured it since. Scaling that study with today's premium base puts the number far higher.
Holding Verisk's 14% ratio steady against today's premium base gives about $43B. Scaling by average premium gives about $39B. The digital channel, rising MVR costs and GenAI all push toward the top of the range.
of new auto policies are now bought digitally, up from 36% in 2021 (J.D. Power, 2026). The second set of eyes is gone from half of new business.
average MVR fee in late 2024, up nearly 28% in ten years (Verisk). Every pull you skip is leakage you don't price.
a year from underpriced violations and accidents alone, up from $3.4B (Verisk, 2024).
What record checks miss
These three patterns produce clean-looking applications. The behavior is the only place they show up.
The applicant edits the same rating answer again and again while watching the price, then submits the one that worked. The final value looks ordinary. The edit trail doesn't.
After a decline or a referral, the same person comes back and changes exactly the answer that stopped them. In the data it's two unrelated applications. In the behavior it's one person trying again.
Some sessions arrive knowing exactly which prefilled fields to override to suppress the premium. They look nothing like a real applicant correcting a stale address, even when the values match.
Why it matters to loss ratio
When an undisclosed factor turns up later, actual claim costs can run far above the premium you priced. The four biggest are also the easiest for an applicant to change on a web form.
ForMotiv estimate drawn from patterns across its book of business. Not an industry-wide average.
In-session responses
Your rules decide. Most carriers start light and add friction only as signals stack up, so honest applicants never notice ForMotiv is there.
“Does anyone else in your household drive this car?” A light confirmation that costs a real applicant two seconds.
Lowest frictionPull MVR, CLUE or other paid reports only when behavior justifies the cost, instead of on every quote.
Spend where it paysFreeze answers once a price is shown, or interrupt the flow before a second round of edits.
Stop the shoppingSend the application for manual review with the signals that fired and the reasons behind them.
Highest frictionGood applicants move fastFriction where intent warrants it
See where answer changes cluster by channel, state, product and agents and agencies in the ForMotiv Console.
Feed a DataPack of the most predictive leakage features into the pricing and loss models you already run. See the data
Signals are explainable rules. Every one comes back with the behavior that fired it, so underwriters know why an application was routed.
Related
UnderwritingMisrepresentation, leakage and nondisclosure, plus early-loss predictions.Team page → Agents & brokersWhen the driver disappears at the agent's keyboard, not the customer's.Agent channel → Account servicingDrivers removed and garaging moved right after bind.Transaction → Life nondisclosureThe same idea on tobacco, alcohol and health questions.Use case →
FAQ
What is premium leakage in insurance?
Premium leakage is revenue a carrier loses when a policy is priced on inaccurate application answers, such as undisclosed drivers, understated mileage, wrong garaging address or hidden violations. The policy binds at a lower premium than the real risk deserves.
How does ForMotiv detect premium leakage?
ForMotiv captures how rating answers are given, in session and in real time, including edits and answers that change as the applicant goes or on a later visit. Record checks see only the final answer; ForMotiv shows how it got there.
Isn't version history enough to catch rate evasion?
Version history shows what changed between saves. ForMotiv shows the behavior behind each answer as it happens, including changes made before a price is ever shown, so carriers can act during the session rather than after.
What can a carrier do when leakage signals fire?
Carriers can verify the answers that moved, ask for documentation, quote on the original answers, or refer for manual review before the policy is issued.
Does premium leakage happen in agent-written business too?
Yes. Agents and agencies know which answers move the price, so ForMotiv reads agent sessions the same way and benchmarks agents and agencies against their peers.
Bring one journey and one goal. We’ll show you the behavioral signals behind it, the decisions they can drive, and how ForMotiv fits into your stack.
Leakage, fraud, conversion or agent oversight, on the line and channel you care about most.
A walkthrough of the signals, features and scores for journeys like yours.
Where the tag goes, what the API returns and which systems act on it.