Every insurer watches shoppers change answers during a quote. Most of it is honest: a typo fixed, a date corrected, a driver added and removed while a household works out who needs cover. Fraud rehearses in the same interface, and the session record is how you tell the difference. A ghost broker or a rate evading applicant iterates the rating factors themselves, occupation, address, annual mileage, years licensed, claims history, watching the premium respond, then buys at the combination they cannot document. Honest corrections move toward the applicant's real details and stop. Manipulation moves away from anything verifiable and stops at a price. That directional difference, plus repetition across identities from one device, is the signal set modern quote monitoring is built on, and it works pre sale, which matters, because catching the policy at the quote is cheaper for everyone than voiding it at the claim.
Direction, velocity, and the tell of the rating factors
Not all fields are equal. Nobody commits fraud by iterating their email address. The rating factors, the fields the premium is computed from, are where manipulation lives, and a session that touches six rating factors in sequence, each change lowering the premium, each value unrelated to the last, is not comparison shopping. Comparison shoppers change cars and coverage levels. Manipulators change who they claim to be. Build the monitoring on rating factor edits specifically, scored by direction: edits that raise the premium are almost never fraud, edits that lower it in a chain deserve the look.
Velocity refines it further. A household quote takes minutes and meanders, with pauses, backtracking, a second person consulted. A rehearsed manipulation session is brisk and systematic, testing one factor at a time like a pricing experiment, because that is what it is. Session timing alone proves nothing. Session timing stacked on rating factor direction, stacked on a device that ran the same experiment for a different surname last week, is a case an investigator can open.
Related reading on this site: Home and Friendly Fraud: Chargeback Signals Hiding in Plain Sight cover the neighbouring problems in more depth.
The response ladder, from friction to refusal
The amateur response to a manipulation score is a hard decline, and it teaches the manipulator which combination tripped the wire so the next session routes around it. The professional ladder starts invisible: log and score, let the policy issue where the evidence is thin, and flag the account for document checks at the first claim or mid term adjustment. Where the pattern is mid strength, add friction that honest shoppers pass without noticing and rehearsed identities fail, verification of license history against the issuing record, address confirmation against postal data, payment instrument name matching the policyholder.
Reserve refusal for the stacked cases, and when you refuse, refuse flatly without itemizing the signals. Every declined manipulator who learns the threshold becomes a consultant to the next one. The fraud team's silence about its own scoring is not coyness. It is the control.
False positives are households, and they are owed a path back
The model will flag honest people. A parent quoting for three children on one laptop iterates drivers and details for an evening and looks, to a careless rule, like a one person ring. A family that moved house quotes from the old address and the new one in the same week. If the only outcome of a flag is a voided policy and a fraud marker, the insurer has converted its most ordinary customers into its angriest complainants, and regulators have views about that conversion. Build the appeal path when you build the rule: human review, a document route that clears the flag, and a correction to the customer's record when you were wrong, in writing. The cost of the path is trivial against the cost of the marker wrongly applied.
Frequently asked questions
What is quote manipulation in insurance?
It is iteratively changing rating factors such as occupation, address, mileage, or claims history during a quote to obtain a lower premium based on details that are not true. Insurers detect it from the pattern of edits in the quote session.
How do insurers distinguish manipulation from honest corrections?
Direction and context. Honest edits move toward the applicant's real details and stop; manipulation moves through rating factors in sequence, each change lowering the price. Repetition of the same pattern across identities or devices strengthens the signal considerably.
What happens when a quote is flagged for manipulation?
Responses form a ladder: silent logging and scoring, added verification such as license or address confirmation, document checks at claim or adjustment, and refusal only for strong stacked cases. Flat early refusals teach manipulators the thresholds, so they are used sparingly.
Can an honest customer be flagged by mistake?
Yes, especially households quoting for several drivers from one device. A fair system includes human review and a document based route to clear the flag, with the customer's record corrected when the flag was wrong.