How we score Insurance Companies
Every score we publish is built from public regulatory filings: documents anyone can download and check. No surveys, no sponsored ratings, no insurer-supplied numbers we can't trace.
Every quarter, every health insurer files fourteen regulatory forms. Once a year, three more public sources add the industry picture. We read all of it: every form, every line.
The number everyone quotes can't protect you.
Most comparisons rank insurers on one or two headline figures: a settlement ratio, a claims ratio. Those numbers aren't wrong. They're just not enough, in three specific ways.
One headline ratio splits into the three things it hides: the rupees behind the count, the slow large claim behind the average, and what happens when you push back.
It counts claims, not rupees.
An insurer can settle 95 of 100 claims and still trim every large bill. By count, that's a success story. In rupees, it can be a very different one.
It averages away your worst day.
Being fast on thousands of small claims says nothing about the one big hospital bill. Averages hide exactly the claim you bought insurance for.
It ignores what happens when you push back.
Complaint counts don't tell you who was right. An insurer that loses when customers escalate is telling you something no settlement ratio captures.
The right plan from the wrong insurer is still the wrong choice. So we made the filings answer harder questions.
Every number takes the same journey.
A score you can't trace is just another opinion. So here is the path every figure we publish travels, from a document anyone can download to a plain answer you can act on.
We start with the public filings: every insurer's quarterly public disclosures, plus the IRDAI Annual Report, the IRDAI Handbook, and the GIC Yearbook. Published by insurers and the regulator, not by us, and not by anyone we pay. We read every line, around 800 figures per insurer every quarter. We run more than 120 cross-checks to catch anything that doesn't add up. We translate what survives into plain answers to the questions you would actually ask. We compare each answer against similar insurers, because a number means little on its own. And only then does it become a Trust Score, with simple Green, Yellow, and Red zones for each measure.
That journey is the same for every insurer and every number. No step is skipped, and every step is one you could repeat yourself.
The journey every number takes: public filings, then we read all of an insurer's roughly 800 figures each quarter, run 120-plus cross-checks, translate the result into plain answers, compare it against similar insurers, and turn it into a Trust Score with Green, Yellow, and Red zones.
Four questions we make the filings answer.
Every metric below is computed from the insurer's own regulatory filings, and every formula is open. Not every question weighs the same in our scores: claim denial counts heaviest. That's an editorial position, and we hold it openly.
Will they pay?
The first and heaviest question. Does the insurer actually pay claims, or quietly turn them down?
Claim denial rate
The share of decided claims an insurer rejects outright. We compute it from the insurer's own quarterly claims data, not a marketing page, and track how it moves year on year. A denial rate that's climbing tells you more than one that's merely high.
The same rate, segment by segment
A healthy looking overall denial rate can hide a much higher one on add-on covers, like the personal-accident cover often sold alongside a health policy. So we apply the same calculation to each segment, not just the headline.
Will they pay in full, and on time?
Paying a claim isn't the same as paying all of it, or paying it before the hospital bill becomes your problem.
Deductions (the Deduction Index)
A claim can be settled and still arrive shaved: deductions taken from a covered bill. We measure how much of the billed amount gets disallowed. Today this is honest only industry-wide; a per-insurer version isn't possible yet, and we will tell you that rather than fake it.
Settlement speed, and whether big claims wait longer
First, how often claims are settled quickly: the share settled within a month. Then the question almost nobody asks: are the slow claims the large ones? We compare the average size of claims that took a month or more against those settled within a month.
What happens when you push back?
Anyone can look good until you disagree with them. This is about what happens when you do.
Complaints per 10,000 claims
How often customers are pushed to complain, scaled so insurers of different sizes compare fairly. We recompute it from the raw grievance and claims counts, then check it against the figure the insurer reports itself.
Complaint rejection rate
When customers do complain, how often the insurer rejects the complaint rather than overturning its own decision. The rest is the overturn rate: the times pushing back actually worked.
Do claims stay settled? (the do-over rate)
A settled claim that has to be reopened was a customer who had to fight twice. We measure how much of an insurer's claims activity is previously closed claims coming back.
Will the policy still make sense later?
A policy that's cheap today and unaffordable at renewal wasn't cheap. These signals point to what's coming.
Combined ratio
Whether an insurer pays out and spends more than it earns. Read it in rupees: of every ₹100 of premium, how much goes to claims, plus how much to running the business. Above ₹100, it is paying out more than it takes in, a leading hint that prices may rise. We use claims incurred, not just claims paid, so claims already owed but not yet paid still count.
Financial stability (solvency)
Whether the insurer holds enough capital to honour its promises. The regulator sets a floor every insurer must stay above.
What the regulator already knows
Penalties the regulator has levied are disclosed twice: in the insurer's own accounts and in the regulator's annual report. We read both, check they agree, and track whether penalties are rising year on year.
We check their numbers against themselves.
Insurers don't file one number: they file hundreds, across fourteen forms, and those numbers have to agree with each other. Before anything enters our scores, it has to survive that audit.
Each figure runs the audit: pass and it keeps its receipt of form, filing, and line; fail and it goes to a review queue instead of into a score.
CHECK 01
Parts must sum to wholes.
When an insurer reports claims by how long they took to settle, the buckets must add up to the total it reports elsewhere on the same form. When they don't, something is being miscounted, or moved.
sum of age-wise claim buckets = reported total
Traced to Form NL-37: Claims Data
CHECK 02
The same number must agree with itself.
Health premium appears three times in every quarterly filing: once in the premium schedule, again split by geography, again split by line of business. Three forms, one truth. We verify all three agree.
NL-4 premium = NL-34 geographic total = NL-35 line-of-business total
Traced to Forms NL-4, NL-34, NL-35
CHECK 03
We redo their arithmetic.
Insurers publish their own analytical ratios: claims ratios, expense ratios, commission ratios. We don't take even their arithmetic on trust: we recompute every one from the raw schedules underneath and compare it against what they published.
published ratio = ratio recomputed from raw schedules
Traced to Form NL-20 vs Forms NL-4, NL-5, NL-6, NL-7
120+ checks on every filing
A number that fails doesn't enter our scores: it enters a review queue.
And every number that passes keeps its receipt: the form, the filing, the line it came from.
We never score in a vacuum.
A single settlement or denial rate can sound precise and still tell you nothing on its own. Is it good, or bad? That depends entirely on what is normal for insurers like this one. A number with no comparison is not an answer.
So we never read a number alone. We compare each insurer against a group of similar insurers, for example other standalone health insurers rather than large general insurers that also happen to sell health cover, so like is measured against like. Then we translate where each measure lands into a simple read: Green, Yellow, or Red.
Green, Yellow and Red describe position, not pass or fail. They tell you at a glance whether a measure is strong, worth a closer look, or a real concern for insurers of this kind. We don't publish the exact cutoffs: that is our editorial judgement, held openly as judgement, and kept out of reach of anyone who would rather game the score than earn it.
An insurer's position on a measure shown against similar insurers, with a Green, Yellow, and Red zone legend. Positions are illustrative and no cutoffs are shown.
Claim denial rate
Complaints per 10,000 claims
Compared with all insurers: claim denial rate better than most peers, Green; complaints per 10,000 claims around the middle, Yellow.
What a finished score looks like.
Here is how all of this comes together for a single insurer. The example below is illustrative, not a real insurer's data.
Your Trust Score blends four parts of an insurer's claim track record: how often it denies claims, how often settled claims get reopened, how often it rejects your complaint when you push back, and whether regulatory penalties against it are rising. Claim denials weigh the most. Other things we measure, like settlement speed, financial stability, and the risk of a future premium hike, are shown alongside your score, not folded into it.
Illustrative example, not a real insurer.
An illustrative score card. Trust Score 64 of 100. Verdict: Pays most claims, but pushes back on complaints more than most. In the score: claim denials Yellow, reopened claims Green, complaint rejection Red, rising penalties Green. Shown alongside, not folded in: settlement speed Green, financial stability Green, premium-hike risk Yellow. Not a real insurer.
An interactive explainer of how four parts of an insurer's claim record (claim denials, reopened claims, complaint rejection, and rising regulatory penalties) combine into an overall Green, Yellow, or Red zone.
Tap each part to switch it between good and concerning, and watch the overall zone.
Overall: Green, Strong.
What the score does and doesn't tell you
The score is a track record, not a promise. It tells you how an insurer has actually behaved (with claims, with complaints, with the regulator) not how it will behave with yours. Past behaviour is the best public evidence there is. It is still evidence, not certainty. The data behind it is as fresh as the filings: quarterly for most signals, annual for some.
And when the evidence is incomplete, the score says so. If a filing is missing one of the signals behind a score, you'll see it marked (based on three of four indicators) rather than silently pretending completeness.
Some of what we'd like to measure, we can already publish, with its limits stated plainly. We show an industry-wide deductions figure (how much of billed value gets disallowed across the market) and an insurer-level signal for whether bigger claims wait longer than small ones.
Three things we want to measure and can't do per insurer yet, because the data simply isn't published:
- Per-insurer deductions. How much each insurer shaves from covered bills. Insurer by insurer, that needs per-claim billing data no one publishes; today it is honest only industry-wide.
- True settlement time by claim size. Whether a ₹4-lakh claim waits longer than a ₹40,000 one, measured directly rather than through the strong proxy the filings allow today.
- Per-insurer ombudsman outcomes. Who wins when disputes reach the insurance ombudsman. The regulator publishes this industry-wide, not insurer by insurer.
The day that data exists, we'll compute it. Until then, every gap stays labelled, not papered over. Some of these gaps can only ever be closed by buyers sharing real claim experiences; that's part of where we're headed.
Where this stands today
The scoring system is built and validated end-to-end, and we're rolling it out across every health insurer in the market, filing by filing. Scores go live as each insurer's data clears our checks, not before.
Spot an error? Disagree with a call we made?
Tell us. If it's a number, we'll trace it back to the filing, and either fix it, or show you the source. If it's our judgement, we'll hear you out. Either way, you get an answer.