Investigators kept recognizing the same people behind supposedly unrelated claims but could not prove it without days of manual digging. A single view of every claimant across policies and claims surfaced 290 suspicious clusters and cut the time to build a case by more than half.
INDUSTRY
Insurance (property and casualty)
BUSINESS PROFILE
Personal & commercial P&C · Sheboygan, WI · Policyholder-owned mutual
Acuity Insurance is a property and casualty insurer headquartered in Sheboygan, Wisconsin. A mutual company owned by its policyholders, Acuity writes personal and commercial lines across more than 30 states through a network of independent agents, and employs roughly 1,700 people. It is rated A+ for financial strength and ranks among the larger regional carriers in the country.
An Acuity investigator opened a new bodily-injury claim and felt he had seen the claimant before. The name was spelled differently and the address had a new apartment number, but the date of birth and the treating clinic were familiar. Confirming the hunch meant most of a day cross-referencing two systems by hand, and by the time he had, more claims like it had come in.
The special investigations unit had always suspected that organized claims activity was larger than it could prove. The problem was never instinct; it was evidence. Linking a claimant to prior claims, shared addresses, repeated providers, and the same few vehicles took manual work that did not scale past one case at a time.
Acuity's claims history and policy records lived in separate systems, and a claimant could appear in each under a slightly different name, address, or date of birth. To connect them, an investigator pulled records from both, lined them up by hand, and made a judgment call. A single suspicious file could absorb most of a day before anyone knew whether it was worth pursuing.
Individual claims rarely look fraudulent on their own. The signal is in the pattern: the same garaging address across unrelated policies, one clinic feeding dozens of injury claims, a vehicle that keeps reappearing. Without a way to connect parties across the whole book, those patterns stayed invisible, and only the clumsiest schemes were caught.
Referrals arrived faster than investigators could work them, so the queue grew and the lower-value files were quietly dropped. The unit was not short on suspicion or skill. It was short on the hours that manual cross-referencing consumed before real investigation could begin.
Two pressures turned a long-standing frustration into a project. Claim volume was climbing, which stretched the manual approach further, and the state insurance department had sharpened its expectations for how carriers detect and report suspected fraud. Acuity needed to connect claimants across its book reliably, and to show the reasoning behind every link.
Acuity's SIU did not want a black-box fraud score. A model that flagged a policyholder as suspicious without explaining why would be worse than useless, because acting on it could mean wrongly accusing an honest customer. The unit wanted to see the connections and the reasons behind them.
Acuity chose MatchLogic to resolve claimants across its claims and policy systems and licensed the Server tier, with the Workflow Scheduler rebuilding the linked view every night. Investigators would start each morning with the connections already drawn, not with a blank cross-referencing task. Five capabilities decided it.
Anchored by Jaro-Winkler distance and phonetic encoding, the engine recognized the same claimant or insured across both systems despite different spellings, addresses, and dates of birth. The variations that had stalled investigators became something the matching handled in seconds.
Using its own rule definitions, the SIU connected parties that shared an address, a phone number, a bank account, or a vehicle, so a cluster of related claims surfaced as one network rather than as isolated files. Patterns that only appear at the level of a ring became visible for the first time.
A reason behind every link
Every connection carried the rule and the field-level scores that produced it. An investigator could see exactly why two records were tied together, and the same explanation supported a referral or a report to the state without anyone reconstructing it later.
Caution built into the matching
Because wrongly tying an honest policyholder to a fraud network is a serious harm, Acuity tuned the rules conservatively and routed anything short of a high-confidence link to an investigator. In a hand-checked sample of a thousand candidate pairs, false matches came in under one percent.
Regulated data that stayed in-house
The matching ran on a server inside Acuity's own data center, so claims and policyholder data never left the company's control. For a regulated carrier handling sensitive personal and medical information, keeping that processing in-house avoided a new round of vendor security review.
The work was sequenced around the unit's biggest fear, a false accusation, so the early weeks went to tuning and review rather than speed.
Claims data is written under pressure, and it showed. The same claimant appeared as Robert in one file and Bob in another; addresses carried apartment numbers in one system and not the other; dates of birth were transposed often enough to matter. The team built the match rules against this reality and validated them against a set of pairs investigators had already judged by hand.
The team resolved twelve years of claims history first, which gave the SIU a back catalog of networks to investigate, then set the nightly job to fold in new claims as they were filed. From kickoff to a nightly linked view the investigators trusted, the work took about ten weeks.
The match drew on 1.8 million records, the claims and policy systems plus twelve years of party history, and resolved them to about 420,000 unique people and businesses. Every legacy record stayed linked to the unified party, so a claim, a policy, and a prior loss now resolve to one identity. The nightly job flags when a new claim belongs to a party the SIU is already watching.
Run against the back catalog, the matching connected roughly 1,300 claims into about 290 clusters the unit had never seen as related. Some were coincidence and closed quickly; others became active investigations into organized activity. The value was not a single figure but a queue of real leads where there had been guesswork.
Because the connections and their evidence were drawn each morning, the hours an investigator once spent proving that two records were the same person mostly disappeared. The unit measured the time to assemble a case file and found it fell about sixty percent. Investigation, the part that needs human judgment, now starts almost at once.
With the manual cross-referencing removed from the front of every case, the SIU worked about forty percent more referrals without adding staff. The backlog of lower-value files that used to be dropped is now triaged properly. Capacity, not budget, had been the constraint, and the matching returned it.
Acuity now investigates against a connected view of its claimants rather than a pile of separate files. The unit is extending the same matching to its agent and vendor records, where duplicate and related parties carry their own risk, and the nightly job keeps the picture current as new claims arrive. For a mutual that has answered to its policyholders since 1925, catching organized fraud faster protects the people who fund every honest claim.
Founded in 2003, MatchLogic has matched over 2 billion records across 4,500+ enterprise installations, powering production-scale entity resolution pipelines.
MatchLogic combines deterministic and fuzzy matching with visual profiling and drag-and-drop cleansing workflows to identify duplicates and maintain trusted golden records.
Built for fintech, healthcare, insurance, and government teams, the platform provides REST API integrations, cross-platform deployment, and full auditability across every match operation.
.svg)
