When a Company's Claims History Becomes Portable: The Insurance Risk Score in IRDAI's Registry Paper

IRDAI released the Public Insurance Registry consultation paper on 1 September 2026, with comments open until 30 September 2026. Early reading treated it as infrastructure: a consent-based, authoritative record of policies, claims and grievances replacing the reconciliation work everyone currently does by email. That reading is correct and incomplete.

The paper also proposes that the registry may enable an Insurance Risk Score, described as a unified consent-based score that synthesises insurance history from the Insurance Information Bureau of India (IIB), credit information companies and other permitted external data sources, as a decision-support input for underwriting. One sentence, and it moves the registry from a records project into a pricing project.

The wording matters. The score is framed as decision support, not as an automated accept or decline and not as a filed rate. An underwriter still holds the pen. Anyone who watched credit bureau scores enter Indian lending, though, knows what decision support becomes once it is available on every file: the default anchor, with the discussion reduced to why this risk should be treated differently from what the score says.

For a commercial buyer the practical question is what the score would be built from, and what a broker can do about a number that follows the company from insurer to insurer.

The consultation describes the registry as a connector rather than a copy. Business Standard reported on 2 September 2026 that the PIR could connect to the Central KYC Registry, Aadhaar authentication, vehicle-registration systems, health and mortality registries, weather and disaster databases and court case-management systems, without copying their data. A federated query model changes the security profile of the project, because no single breach exposes a consolidated national dataset, but it does not change what an underwriter can see at the moment of a query.

Translate that into a mid-size manufacturing account. The pieces a score could reach include:

• Claim count, paid quantum, outstanding reserves and policy year for every claim reported across insurers, not only the ones the buyer chose to disclose.
• Policy churn: how often the account moved insurer, and whether each move followed a loss year.
• Vehicle registration records for a commercial fleet, which underpin motor and fleet rating independently of what the proposal form says.
• Weather and disaster data mapped to the insured location, which is exposure evidence the buyer cannot dispute with a narrative.
• Litigation history from court case-management systems, which is directly relevant to liability and directors and officers underwriting.
• Credit information from companies regulated under the Credit Information Companies (Regulation) Act, 2005, which introduces payment behaviour into insurance pricing for the first time at scale.

That last item deserves attention. Credit data carries predictive power for claims frequency in several markets, and it imports every known fairness problem attached to credit scoring. Ask, in the comment window and later at renewal, how heavily the credit component weighs and whether a thin credit file for a young company reads as neutral or negative.

The commercial remarketing playbook for a bad three years is well understood. The broker collects claims experience certificates from the incumbent, which arrive late and in inconsistent formats, frames the loss record around a single dominant event, and takes the submission to markets that have no independent way to test the framing before quoting. The incumbent holds an information advantage; everyone else works from what is in the submission.

A portable score removes the gap between what the broker presents and what the market already knows. The loss record arrives at the underwriter before the submission does. That has three effects worth planning for.

First, the buyer with genuinely poor experience loses the option of quietly moving. Rate correction that would previously have taken two or three renewal cycles to catch up with the loss record arrives at the first renewal after the score does.

Second, the incumbent loses its information advantage, which cuts in the buyer's favour more often than brokers expect. Incumbent insurers currently price partly on the assumption that no competing market can verify a good record either. A buyer with a clean five years and no way to prove it against market averages gains from a system that proves it.

Third, the duty of utmost good faith changes character. Non-disclosure of a prior loss is currently discovered at claim stage, when the consequence is a repudiated claim on a live loss. Under a registry, the same gap surfaces at underwriting stage, when the consequence is a question in an email. Detection moving earlier is better for the buyer, even though it feels like exposure.

The score is the visible half. The paper's second data proposition operates on the portfolio and will reach buyers who never see a score at all. The draft says the PIR

can enable collection, collation, aggregation and analysis of anonymized loss and claim data from all insurers, and provision of actionable anonymized information

and adds that it could set up structured alerts on unusual spikes or concentration of losses across parts of the country, segments and sectors.

Read that as a market-wide accumulation view no single insurer can currently build. An insurer sees its own book. It knows its exposure in a flood-prone industrial belt but not the industry's, and cannot see how many other carriers write the same cluster of chemical plants on the same river. Concentration alerts change that, and the first thing an insurer does with a concentration signal is price for it.

The timing lines up with a live pricing argument. Economic Times reported on 3 September 2026 that non-life insurers are discussing a floor rate for the natural-catastrophe component of fire cover after steep discounting, with the IRDAI chairman voicing concern about insurer financial health. Floor rates are hard to defend on assertion and easy to defend on pooled loss evidence. A registry producing industry-wide loss aggregates gives the discipline argument a factual base, and gives a broker arguing against a catastrophe load a harder job.

For a buyer, the consequence is that part of next year's rate movement will have nothing to do with the buyer's own record. It will come from a concentration alert on the district, the occupancy class or the sector. Separating those two components in a renewal discussion becomes a core broking skill rather than a refinement, in the same way that the [return of technical floors through reinsurance treaties](/underwriting-risk/de-tariffing-correction-stfi-iib-minimum-rates-fire-treaty-india-2026) already forces buyers to read a fire quote component by component.

Business Standard reported that consent-based access under the PIR must be specific, informed, revocable and auditable, with role-based access and purpose limitation. Those four adjectives are the same test the Digital Personal Data Protection Act, 2023 applies to consent, and they are not satisfied by anything most broking firms currently hold.

A common objection is that commercial insurance deals with company data, so the DPDP Act does not bite. For most Indian commercial accounts that is wrong in practice. A proprietorship or partnership is its proprietors. A corporate file carries KYC of authorised signatories and directors. A fleet policy carries driver records. A group health or workers compensation placement carries employee and dependant data, including health data. Once any of that flows through a consent-based registry query, the broker is processing personal data and needs the artefacts to prove it did so lawfully.

What the broker has to be able to produce, per client and per query:

1. A consent record naming the specific purpose (placement, renewal remarketing, claims verification), not a standing clause in the terms of engagement saying data may be shared with insurers.
2. Evidence the notice was given in a form the client could understand, with the categories of data and recipients identified.
3. A working revocation path, and a documented answer to what happens to a placement that is mid-market when consent is withdrawn.
4. An access log showing which staff member queried what, when, and under which mandate, because auditable means someone will eventually ask.
5. Role-based access inside the broking firm, so that a servicing executive on a motor account cannot pull a score on an unrelated client.

If the loss record travels regardless, the only remaining lever is the explanation attached to it. That is a real lever, because a score is a summary and an underwriter prices a risk. The broker's job shifts from controlling what the market learns to controlling how fast the market learns why the record improved.

A mitigation file that actually moves a rate has six components, and each one has to be evidenced rather than asserted.

The six components

1. A loss register reconciled to insurer records. Every claim, with date, cause, paid amount, outstanding reserve and status, tied to the insurer's own statement. If the client's register and the insurer's differ, fix that before an underwriter finds it, because a mismatch reads as either poor control or selective disclosure.
2. Root cause per material loss. For every claim above a threshold the client sets, the surveyor or loss adjuster reference and the actual cause, not the peril label. "Fire" is a peril. "Failed thermal cut-out on a 2011 dust collector" is a cause an underwriter can price against.
3. Corrective action with dates and cost. What was changed, when it was commissioned, and what it cost. A hydrant upgrade with a commissioning certificate and an invoice is evidence. A line in a submission saying fire safety has been improved is not.
4. A post-remediation experience window. The loss record since the fix, stated separately from the full period. A three-year record containing an INR 4 crore loss in year one and nothing since is a different risk from one losing INR 1.3 crore every year, and a headline frequency figure hides the difference.
5. Changed exposure. Sites closed, processes discontinued, a hazardous storage relocated, fleet composition changed. Exposure changes should be quantified against the sum insured and location schedule, because a score built on history will not know the plant that burned no longer exists.
6. A retention offer that prices the argument. If the client believes the attritional losses are behind them, a higher voluntary deductible or an aggregate retention converts that belief into a number the underwriter can accept. It is the most credible signal in the file, because it costs the client money if the belief is wrong.

One discipline sits above all six: separate attritional frequency from single-event severity in every presentation. A score compresses both into one figure, and the mitigation file exists to decompress it.

The comment window closes on 30 September 2026. Responses from brokers and corporate buyers are thin in most IRDAI consultations, so the design settles around insurer and technology-provider preferences by default. A few questions are worth putting on record.

• Can the insured see their own score, and contest it? A score used in underwriting that the subject cannot see is a decision the subject cannot argue with. Credit bureau practice in India already settled this in favour of disclosure and dispute rights.
• What is the correction path for an erroneous IIB entry? Claims data quality across insurers is uneven. A closed-no-payment claim recorded as paid, or a claim booked against the wrong policy, becomes a permanent rating penalty once it is a scored input.
• What happens to a placement when consent is revoked mid-market? Revocable consent means little unless the operational consequence is defined.
• Is non-personal corporate data inside or outside the consent architecture? If a company's claims record is not personal data, it may sit outside DPDP consent entirely, and the specific and revocable promise then covers less than a buyer assumes.
• What retention applies to a score once it is pulled? Purpose limitation without a retention rule leaves copies of scores in insurer and broker systems long after the purpose expires.

While that plays out, the preparation is the same whether the registry ships in 2027 or later. Reconcile your claims register against every insurer statement you hold, this quarter, before anyone else does it for you. Fix the identifiers, because a group whose subsidiaries sit under three naming conventions will be scored inconsistently until it resolves to one entity. Build the mitigation file for your worst account now, while no renewal deadline is attached to it. Firms already working through broker data readiness for the registry or standing up a claims data warehouse are doing the same work from the operations side.

The underlying shift is straightforward. Commercial insurance in India has been priced with an information asymmetry that favoured whoever held the file. A registry with a score removes that asymmetry. Brokers whose value came from managing what the market knew will feel the loss. Brokers whose value comes from evidence, risk engineering and structuring will find the argument easier to win, because for the first time the counterparty can verify the good news too.