Why Claim-to-Policy Matching Matters
A loss amount is useful for ratemaking only when it is attached to the policy and coverage that generated it. A claim record may identify an insured and a total loss, while the policy system holds separate records for terms, locations, and coverages. Connecting those records incorrectly can shift losses between products or periods. Repeating a claim’s full amount once for each coverage can also inflate the total experience.
The task is not simply to find a policy with the same insured name. An insured can have several policies, and a policy can have several coverages or locations. A defensible match uses the claim’s identifying information and the applicable policy and coverage records, then checks that each loss amount is counted once. The matching process should also be consistent with the period basis being used: for example, a policy-period analysis must use the relevant policy term, rather than assigning a claim to a term solely because it was reported during that term.
What Evidence Supports a Match?
Start from a stable claim identifier, not from the dollar amount. The claim identifier distinguishes one claim from another; it does not, by itself, establish which policy or coverage applies. A match usually depends on agreement among several fields. Relevant fields include the policy number and term, insured or entity, occurrence date, location or covered item, coverage identifier, and claim description. Claim-file or policy-administration records may clarify conflicting or missing fields.
Use the date appropriate to the analysis and the available contract records. A claim reported after a policy expires may still arise from an occurrence during that policy term. Conversely, a claim reported during a term does not automatically belong to that term. If the matching evidence is incomplete or conflicting, flag the record for resolution rather than selecting a policy by guesswork.
| Matching question | Evidence to check | Why it matters |
|---|---|---|
| Which policy term could apply? | Policy number, term dates, and the date basis used for the analysis | A policy may renew or be replaced; the insured name alone does not distinguish terms. |
| Which insured, location, or item is involved? | Named insured, location identifier, schedule, and claim details | One insured may have multiple policies, sites, or covered items. |
| Which coverage responds? | Coverage code, claim description, and coverage records | A policy can contain multiple coverages that represent different loss categories. |
| Is the amount recorded once? | Claim identifier, component amount, and crosswalk rows | A repeated claim total on multiple coverage rows can create duplicate losses. |
For each claim, a practical control is to ask whether the match is unique, supported, and reconcilable. A claim can match one coverage, or it can have multiple coverage rows when its loss components are separately supported. In the latter case, the rows must carry the component amounts—not a repeated copy of the entire claim total.
Allocating a Claim Across Coverages
When a claim contains loss components assigned to different coverages, use source detail that identifies the components and their amounts. Do not create a split merely because a policy has several coverages. If there is no support for a coverage allocation, the data should remain flagged until the allocation is resolved; an arbitrary percentage can make coverage results appear more precise than the records justify.
Here \(L_c\) is the source total for claim \(c\), and \(L_{cj}\) is the amount of its component assigned to coverage \(j\). This equality is a reconciliation condition, not a rule for deciding the allocation. The components must first be supported by the claim and policy records.
The reason the reconciliation works is straightforward. If the claim is divided into distinct, nonoverlapping loss components, each dollar in the claim total belongs to one component. Adding the component amounts therefore recovers the original total. If the full total is copied onto each of two coverage rows, the rows overlap: their sum counts the same dollars twice. If a component is omitted, their sum falls short of the source total.
A crosswalk should retain enough detail to test this condition: claim identifier, policy identifier and term, coverage identifier, component or allocation basis, assigned amount, and any exception status. Where a claim has only one supported coverage, the crosswalk has one amount row. Where it has several, it has one row per supported component or allocation.
Worked Matching Process
Use the stable claim identifier and the source claim total for the selected loss measure. Keep claim transactions or repeated system rows from being mistaken for separate claims.
Compare the policy and insured information, relevant date, location or item, and policy term. Do not choose a term based only on a matching name or report date.
Use coverage records and claim details to establish whether one coverage applies or whether the claim has distinct components across multiple coverages.
Enter each supported component once, reconcile the assigned amounts to the source claim total, and flag unmatched or conflicting records for review.
Worked Example: Selecting the Correct Policy Term
A hypothetical small-business claim has identifier K-204, a source claim total of $6,250, and an occurrence date of June 28, 2025. The claim was reported on July 8, 2025. The insured has policy P-42, effective July 1, 2024 through June 30, 2025, with building and liability coverages. Its renewal, policy P-43, is effective July 1, 2025 through June 30, 2026, with the same coverages. The claim concerns damage to the insured building. Which policy and coverage should be recorded, and what amount is assigned?
Method. Match the claim to the policy term and coverage supported by the occurrence, insured property, and coverage records. The stated loss concerns building damage, so the relevant coverage is building coverage.
Step 1: Compare the relevant date with the terms. The occurrence date, June 28, 2025, falls within P-42’s term, which ends June 30, 2025. It does not fall within P-43’s term, which starts July 1, 2025.
Step 2: Select the coverage. The claim concerns damage to the building, so the supported coverage is building, not liability.
Step 3: Record and check the amount. Assign $6,250 to policy P-42’s building coverage. There is one supported coverage component, so the assigned total is $6,250, equal to the source claim total of $6,250.
Check. The match is to P-42 even though the claim was reported during P-43’s term. A common mistake is to use the report date and assign the loss to P-43. That would disregard the given occurrence date and place the claim on the wrong term for an occurrence-based match.
Worked Example: Splitting One Claim Without Duplicating It
A hypothetical commercial property claim has identifier M-518 and a source total of $12,500. The applicable policy term and location are confirmed. The claim file documents $8,000 of building damage and $4,500 of equipment damage; the policy schedule identifies building and equipment coverages. Determine the crosswalk amounts and the total assigned to the claim.
Method. Use the documented loss components to assign amounts to their corresponding coverages, then reconcile their sum to the claim total.
Step 1: Assign the building component. The documented building damage is $8,000, so assign $8,000 to building coverage.
Step 2: Assign the equipment component. The documented equipment damage is $4,500, so assign $4,500 to equipment coverage.
Step 3: Reconcile. The assigned total is $8,000 + $4,500 = $12,500. This equals the source claim total of $12,500.
Check. The crosswalk contains two coverage rows, but the claim total remains $12,500. The common mistake is to copy the full $12,500 onto both rows and report $25,000. The correction is to store the component amounts, $8,000 and $4,500, rather than duplicate the aggregate.
Worked Example: Reconciling a Small Claim File
A hypothetical insurer is reviewing two claims for a named insured with several locations and policy records. Claim R-71 occurred at location 10 on February 10, 2025. The file identifies building damage of $6,700 and contents damage of $2,500, for a source total of $9,200. Policy P-A covers location 10 from January 1 through December 31, 2025, with building and contents coverages. Claim R-72 occurred at location 20 on May 4, 2025, and has a source total of $3,100 for liability. Policy P-B covers location 20 from April 1, 2025 through March 31, 2026, with liability coverage. A separate policy record for the same insured covers location 30. Find the supported assignments and the combined assigned total.
Method. Match each claim using its location, term, and coverage evidence; then reconcile each claim separately before combining totals.
Step 1: Match R-71. Its occurrence date falls within P-A’s term, and its location is location 10. The documented components match P-A’s building and contents coverages. Record $6,700 to P-A building and $2,500 to P-A contents. Their sum is $6,700 + $2,500 = $9,200, matching R-71’s source total.
Step 2: Match R-72. Its occurrence date falls within P-B’s term, and its location is location 20. The claim is identified as liability, so record $3,100 to P-B liability. This single assigned amount matches R-72’s source total.
Step 3: Combine the reconciled assignments. The combined assigned total is $9,200 + $3,100 = $12,300. By coverage, the totals are $6,700 for P-A building, $2,500 for P-A contents, and $3,100 for P-B liability.
Check. The separate location 30 policy is not selected merely because it has the same named insured. The assigned total, $12,300, equals the two source claim totals, $9,200 + $3,100. The common mistake is to match on insured name alone or repeat R-71’s $9,200 total on both coverage rows; location and component evidence resolve the first error, while the reconciliation prevents the second.
Exceptions and Common Traps
A matching process should produce exceptions as well as assignments. “Unmatched” is a useful result when evidence is insufficient; it is safer than silently assigning a claim to a plausible but unsupported policy. Similarly, a claim that appears to fit two policies should be investigated rather than counted under both. Resolve the issue using authoritative policy and claim records, and retain enough information to explain the decision.
- Using the insured name as the only key. One insured may have multiple terms, locations, or coverages. Confirm the other relevant fields.
- Confusing a claim’s repeated records with separate losses. Use the claim identifier and the selected claim total; do not treat every row as a new claim.
- Repeating an aggregate amount across coverage rows. Assign distinct component amounts and check their sum against the source claim total.
- Forcing an allocation without support. A policy’s coverage list does not establish how a particular claim’s amount should be divided.
- Ignoring a mismatch. A conflict in policy term, location, or coverage is a reason to investigate and document an exception, not to choose whichever record is easiest to join.
Check Your Understanding
For each hypothetical record, identify the supported assignment or reconciliation. Treat the stated claim amounts and coverage details as source information.
- A claim occurred on December 30, 2024, and was reported on January 3, 2025. Policy T-1 covers the insured from January 1 through December 31, 2024; renewal T-2 begins January 1, 2025. The claim is otherwise confirmed to concern T-1’s covered equipment. Which policy term applies for an occurrence-date match? (Answer: Policy T-1.)
- A claim total of $15,000 is documented as 40% building damage and 60% contents damage, and both coverages are confirmed under the matched policy. Calculate the two crosswalk amounts and check their sum. (Answer: Building $6,000; contents $9,000; total $15,000.)
- A claim has a source total of $4,800. Its file supports $3,300 for one coverage and $1,500 for a second coverage. A data extract instead places $4,800 on each coverage row. What should the corrected assigned total be, and by how much is the extract overstated? (Answer: Correct total $4,800; extract total $9,600, overstated by $4,800.)
- An insured has two policies active on the occurrence date: one covers location A and one covers location B. A claim is linked to location B and has a supported total of $2,700 for the liability coverage on that policy. What policy record and amount belong in the crosswalk, assuming the location and coverage records are confirmed? (Answer: The location B policy’s liability coverage, $2,700.)
- A claim total is $7,200, but the only available record says that the insured has building and contents coverages. No claim detail identifies the damaged property or supports a split. What should the analyst record before allocating the amount? (Answer: Flag the coverage assignment as unresolved and seek supporting claim or policy evidence; do not split or duplicate the $7,200.)