How we compare insurance

Independence

Best Worst Insurance holds no carrier panel, takes no commissions, sells no leads, and runs no affiliate links. Nothing on this site is paid placement, and no insurer has been given sight of a verdict before publication. That is what lets us name the worst product in a category as readily as the best — a publication that earns per policy sold structurally cannot.

We are a publisher, not a broker or an agent. In India we are not registered with the IRDAI; in the United States we are not a licensed agent, broker or web-broker and cannot enrol anyone in a marketplace plan. In neither market do we quote, bind or place cover, and we do not tell any individual reader which policy is right for them. Everything here is general information about publicly filed products. Before you buy or rely on any cover, confirm the current terms with the insurer, the marketplace, or a licensed advisor.

Sourcing standard

Every source is graded, and the grade governs how the claim may be written:

We work from primary documents wherever one exists. Where a carrier’s own site blocks automated readers or has dropped its filings, we recover the document from the regulator’s library or a web archive and say so. Where we have relied on a report of a judgment rather than the judgment itself, the page says that too, in the sentence that makes the claim rather than in a footnote.

The rule we hold ourselves to is simple: a number does not go on a page without a source and a date. Those become the numbered footnotes at the foot of every article, so you can check any figure against the document it came from. Fast-moving items — a pending case, a regulation in abeyance — carry short review dates so they get revisited while they still matter.

Corrections

We publish a corrections policy and a public log of every material correction we make. If you believe something here is wrong — especially if you are an insurer and we have misread your wording — tell us and we will check it against the source and act.

Write to [email protected], or read the full corrections policy and log.

How we compare policies (India)

Everything in this section is about IRDAI-regulated products sold in India — the pages under /in. The work is reading contracts, because in this market the contract is what varies and the price is set by rate ladders that have little to do with it.

What we rank an Indian policy on

We rank the contract, not the brand and not the price. Every verdict on these pages is built by reading the filed policy wording, prospectus and Customer Information Sheet for each product and comparing them on the clauses that decide what actually happens when a policyholder claims. In rough order of weight:

1. 1
   Does the published wording cover the core peril at all?

   The first question, and it disqualifies. If a doctors’ policy excludes bodily injury on its face, nothing else about it matters.

2. 2
   The compulsory excess — is the amount published, and does it also consume defence costs?

   This is the money a policyholder pays on every single claim. A carrier that will not publish the number before you buy is scored down for that alone.

3. 3
   The reporting window after cover ends

   Claims-made cover dies at the edges. The spread across the Indian market runs from 90 days to zero, and it decides whether a late-arriving claim is covered at all.

4. 4
   Who controls the defence, and who pays for it

   Hammer clauses, panel-lawyer restrictions, consent-before-you-instruct clauses, and whether defence costs sit inside the limit or on top of it.

5. 5
   Retroactive and continuity mechanics

   Whether the retroactive date survives a renewal, a lapse, or a switch of insurer — and what documentary proof the carrier demands before it honours past years.

6. 6
   Exclusions, and the carve-backs inside them

   Two carriers can both "exclude cosmetic surgery" and mean opposite things once you read whether burns and trauma reconstruction are carved back.

7. 7
   Eligibility

   Whether the carrier will write your specialty at all. This is separate from exclusions and is usually only discoverable on a product page, not in the wording.

8. 8
   Documentation transparency

   Scored openly, below. What a buyer — or an AI answer engine — can actually verify from primary documents before purchase.

9. 9
   Price

   Deliberately last. Premiums in this market are set by rate ladders rather than by the quality of the contract, so ranking on price would rank the contracts backwards.

Where a carrier publishes nothing on one of these axes, we record the absence rather than substituting a broker’s description of the product. An unverifiable term is itself a finding, and we score it as one.

The transparency score (India)

One axis is scored numerically, because it can be. Documentation transparency measures what a buyer can actually verify from primary documents before purchase, out of ten:

• Policy wording published4 points
• Prospectus published2 points
• Customer Information Sheet published2 points
• UIN published1 point
• Documents hosted on the insurer’s own domain1 point

A document readable only through a third party — the regulator’s own library, a bank distributor’s mirror, an intermediary’s site, or a web archive — counts as published but forfeits the own-domain point. We state which route we used in every case, so the score is reproducible by anyone willing to repeat the work.

The scores themselves are published on the verdict pages, carrier by carrier — see the transparency table in our doctors’ indemnity verdict.

What we cannot tell you about Indian policies

The most useful thing a comparison site can do is be clear about the edges of its own evidence. Ours:

• We never publish an indicative or estimated premium. If a figure is not sourced to a carrier document, a filed rate table or an attributed published quote, it does not appear. Ranges appear only with a source and an as-of date.
• We will not publish a claim-settlement ratio for professional indemnity, because none exists. No Indian regulator publishes one. Where we use liability-segment or ombudsman figures as context, we state precisely what they do and do not measure.
• Our judgment analysis is a litigated sample. Claims an insurer pays without a fight never generate a public record, so a catalogue of judgments over-represents conflict by construction. We report conduct patterns within litigation and never convert them into a settlement rate.
• We rank documents, not service. We can tell you what a contract obliges an insurer to do. We cannot tell you how quickly its claims team answers the phone, and we do not pretend a proxy for that exists.
• Our coverage is not exhaustive. We compare the products we could locate filings for, and we name the carriers whose products we could not verify rather than quietly leaving them out.

How we price plans (United States)

The US pages do something different, because the market is different. There is no contract to hunt down: every marketplace plan’s benefits and cost-sharing are filed with CMS in a standard form and published. What is missing is the arithmetic. The marketplace sorts plans by premium, and a premium is not what a plan costs you — so we compute the total, and this is how.

Where the US data comes from

One source, and it belongs to the public. Every figure on a /us/health page is derived from the CMS Public Use Files for plan year 2026 — no carrier summary, no broker feed, no quote engine.

Source

CMS Qualified Health Plan and Benefits & Cost Sharing Public Use Files, plan year 2026

Compiled

3 September 2026

Licence

Public domain (US Government Work)

Coverage

340 rating-area ledgers, 28 states, 23,544 plan rows

Published

Florida, Texas and Ohio — 409 counties, 11,105 plans

Reference buyer

Single non-smoking adult aged 40, household size 1, full price

We compiled 340 rating-area ledgers covering 28 states and 23,544 plan rows, and we publish three of those states. Florida, Texas and Ohio are 409 counties and 11,105 plans — plans a reference buyer can actually buy, in the three states we have checked county by county. The county tables behind every page are available as CSV from our data page, after a short form, so the totals we print can be rebuilt from the data we used.

Twenty-eight states is not a sample we chose. The Public Use Files cover the federally-facilitated marketplace only. Twenty-two states and the District of Columbia run their own exchange and publish no plan-level data through it, so for those there is no equivalent file to read — and we would rather say that than fill the gap with an issuer’s own description of its plans.

How we build a total annual cost

Unless a page says otherwise in its own heading, the buyer is a single non-smoking adult aged 40, household size 1, paying full price with no premium tax credit applied. A total for that buyer is three things:

1. 1
   Twelve months of the plan’s filed premium

   At the reference age, in that county’s rating area. This is the number the marketplace sorts on, and on its own it is the half of the answer everybody already has.

2. 2
   Plus the plan’s own filed cost-sharing, applied to a course of care

   Its deductible, copays and coinsurance exactly as filed — not a category average, and not another plan’s. This is the half that decides the ranking.

3. 3
   Capped at that plan’s out-of-pocket maximum

   ACA §1302(c) makes the out-of-pocket maximum a legal ceiling on in-network cost-sharing for the year, so no total we print can exceed premium plus that cap.

The care scenarios are CMS’s, not ours. The Public Use Files ship stylised courses of care — the agency’s own standard baskets — and we price every plan against each of them. They are a common yardstick, not a forecast of anybody’s year: nobody’s care looks exactly like one, and the plan that wins on a low-use year is frequently not the one that wins on a high-use one — which is most of what the US pages are about.

Catastrophic plans are excluded. ACA §1302(e) sells them only to buyers under 30 or holding a hardship exemption, so the reference buyer cannot purchase one. Leaving them in would put an unbuyable plan at the top of most tables.

Subsidised figures apply the PY2026 premium tax credit. Where a page gives subsidised numbers it states the household income they assume, because the credit now turns on it sharply: the enhanced credits expired on 31 December 2025, which restored the 400%-of-poverty cliff. Above that line the credit does not taper — it stops.

What we cannot tell you about US plans

• A filed figure is not a quote. Everything here is what the issuer filed with CMS for plan year 2026. What you pay depends on your own age, county, tobacco use and income — confirm it on HealthCare.gov or with the issuer before you enrol.
• We do not compare networks. The plan with the lowest total cost in a county may also have the narrowest network, and whether your own doctor or hospital is in it is not in these files. We rank what is filed; a provider directory is not.
• A $0 deductible can still have a deductible. A plan filing no medical deductible may file a separate drug deductible, which the filed scenario we price does not charge. If a headline deductible looks impossibly low, read the plan’s drug tiers before you assume it.