How QuoteSweep rates insurtech companies

This is an assessment of what a company publicly offers and documents, as of a stated date, judged against the rubric below and comparable only to other companies of the same entity class.

It is explicitly not:

• a measure of financial strength or claims-paying ability
• a customer satisfaction or complaint score
• a prediction of how a claim of yours would be handled
• a judgement on the company as a business or an investment

We hold no performance data on any company listed. We collect no claims outcomes and run no customer surveys. Any rating that implied otherwise would be inventing an authority we do not have.

Profile language

The profile sections labeled “What stands out” and “What to consider” are not reputation verdicts. They summarize sourced product features, purchase constraints, and questions a buyer should resolve before choosing. Company age, funding stage, and gaps in QuoteSweep’s own research are not treated as evidence of quality.

We do not characterize a company as unproven, unreliable, weak, or difficult to work with without directly attributable, appropriate evidence. A missing or inconclusive source is omitted or converted into a neutral verification question. Claims about service, complaints, financial condition, or claims handling must identify the underlying source, scope, and as-of date.

What a source is allowed to support depends on what kind of source it is.

The rule: no dimension may be scored on Tier C evidence alone, except Disclosure Quality, where a company’s own published material is not merely acceptable evidence but the thing being measured. Everything else needs at least one directly linked Tier A or Tier B source that supports that specific dimension.

We classify the origin of the information, not the logo on the page carrying it. A company release remains Tier C when Business Wire or PR Newswire distributes it. An investor or accelerator profile is also Tier C. A database such as Crunchbase, PitchBook or G2 is not independent evidence for a corporate fact unless the entry links to an eligible underlying record; verified customer reviews are handled separately below. Syndicated copies of one release count once.

Before any dimension is scored, a profile must satisfy all of these conditions: its legal role and entity class are identified; every material claim has a direct, dated source URL; every scored dimension has at least one eligible source; and the record includes at least two independent Tier A or Tier B publisher families overall. The two-source floor is a corroboration minimum, not a substitute for dimension-level coverage. A profile that does not clear every test keeps its write-up, facts, sources and disclosures. It loses only the number.

The structure is borrowed from how rating agencies already work. AM Best’s financial strength scale runs from A++ down to D, and beside it sits a separate group the agency calls non-rating designations, one of which is NR: “status assigned to insurance companies that are not rated; may include previously rated insurance companies or insurance companies that have never been rated by AM Best.” The important part is structural. A non-rating is not the bottom of the scale. It is not on the scale at all.

Morningstar’s fund-rating research draws the same line: its star rating requires a three-year record, so a younger fund is ineligible rather than badly rated. NerdWallet’s small-business insurance methodology applies it to insurers directly, stating that a company without an AM Best financial strength rating “isn’t eligible for NerdWallet’s overall star rating.”

What we add is the reason code. AM Best’s NR does not say why, and a company reading it cannot tell whether the agency lacks data or simply never looked. Ours says which, and what would change it.

Every unrated profile carries one more line the old system could not produce: what specifically would make it ratable. That line describes the work or evidence QuoteSweep still needs; it must not imply that the company failed to publish something unless an eligible source establishes that fact.

Five dimensions. Four are scored; specialisation is recorded as a descriptor rather than a score, because a niche specialist is not better or worse than a generalist. It is built for a different risk, and scoring that as quality is a category error.

• Counterparty strength: who actually carries the risk, and how strong are they. For carriers and MGAs this is the AM Best rating of the issuing paper, not the brand. For intermediaries it is licence standing, carrier panel and E&O. For a fully insured health plan it is the underwriting carrier; for a level-funded or self-funded arrangement it is the stop-loss carrier and the administrator’s documented role. A hybrid offering is assessed product by product, not forced into one counterparty model.
• Disclosure quality: can a buyer determine coverage, exclusions, eligibility, cost mechanics and claims process before handing over their data.
• Coverage & access: lines written, states available, admitted versus surplus, and who is actually eligible to buy.
• Buying experience: the quote path, time to bind, self-service, and whether a human is reachable when it matters.
• Specialisation: descriptor only. What risk class this company is genuinely built for.

Weighting by entity class

Weights differ by class, and the class is shown next to every rating. Counterparty strength weighs heaviest where the company carries risk and lightest where it does not; disclosure weighs heaviest for intermediaries, because for someone who places your coverage rather than writing it, what they tell you before you buy is most of the product.

Ratings are comparable only within a class. A brokerage, a level-funded health plan and an ICHRA administrator do different jobs for different buyers, and a number that ranked them against each other would be meaningless.

A publishable rubric needs a written, observable definition for every point on every scored dimension, so that two researchers can score the same evidence independently and reach substantially the same result. Disclosure quality is the completed worked anchor:

5Complete

Coverage terms and named exclusions are published; the claims process is documented step by step; licence and issuing paper are disclosed; appetite or eligibility is specific; and the buyer can understand the cost mechanics, fees and quote path that apply to this entity class.

4Strong

Coverage terms, claims process, licence or paper, eligibility, and class-appropriate cost mechanics are substantially disclosed. One material element is missing.

3Adequate

Lines and eligibility are clear enough to assess fit. The quote path and any potential fees are identified, but important terms or the claims process are available only in outline.

2Thin

Lines are listed without enough terms, exclusions or eligibility detail to determine fit. The buyer cannot tell who charges what, whether additional fees may apply, or how to obtain a firm quote without contacting sales.

1Opaque

Positioning copy only. Coverage claims not substantiated by any published term, form or eligibility statement.

Cost transparency is assessed by entity class. Commercial insurance premiums and group health-plan costs are risk-specific, so the absence of a flat public price is not a drawback by itself. For a carrier or MGA, we look for a clear quote path and disclosure of payment plans and policy fees where applicable. For a brokerage or benefits broker, we look for disclosure of compensation and any broker fees, while recognizing that the carrier sets the premium. For an HRA administrator or software platform, published administrative or platform fees are relevant because those are service charges rather than risk-rated premiums.

The class-specific anchors for counterparty strength, coverage and access, and buying experience are still being calibrated. No v2 score will be released until those anchors are published, a blind double-score test has been run, and disagreements have been resolved without reference to company identity or referral status. That is a release control, not an invitation to improvise missing points.

Evidence grade: the second axis

Alongside the assessment we publish an evidence grade (A, B or C) recording how much independent material sat behind it. It never changes the score. A thin public record lowers our stated confidence and, below the threshold, routes the profile to a non-rating. It does not quietly drag a company’s number down, which is precisely the failure this standard was written to fix.

Public reviews: a separate signal

We look for public customer evidence on platforms appropriate to the product, including regulator complaint data, BBB complaint records, Trustpilot, G2 or Capterra, and app stores. We record the direct URL, platform, aggregate rating, review count and retrieval date. A company-selected testimonial, employer-review site, search snippet, or rating copied onto the company’s own site is not customer evidence for this purpose.

We display an aggregate only when the platform exposes the count and there are at least 20 reviews. We label small or old samples and do not merge unlike populations. Public reviews may inform a separately attributed customer-experience note, but they do not prove licensing, financial strength, coverage terms or claims-paying ability, and they do not clear the ratability gate by themselves.

QuoteSweep editorial scores are not emitted as Review structured data. Google’s review-snippet rules require organization ratings to be sourced directly from users; an editor-created score is a different construct. Structured data must describe the page accurately, not be used to manufacture search-result stars.

No company has paid QuoteSweep for a listing, a score, or a change to one. Some outbound links include measurement tags. Tracking alone creates no compensation. If a paid referral relationship becomes active, the affected link and profile will say so.

That is a claim about us, made by us, which is worth very little on its own. Two things make it checkable. The first is an admission: the structural rules below were written in August 2026, and most ratings in this directory predate them. Those ratings were made in good faith, but they were not made under controls, and we are not going to claim otherwise about work we cannot now audit. They are being re-scored. The second is that from this version onward:

1. Companies are scored against the rubric, with the score recorded and dated, before any commercial conversation. Where a partnership discussion is already live, the profile is scored by someone not party to it, or it is not rated at all.
2. We publish the correlation between rating and referral status annually. If there is no relationship, that table is the proof no paragraph can be.
3. A rating may only move on new eligible evidence cited in the changelog: Tier A or B for most dimensions, and Tier C only where the rubric expressly measures the company’s disclosure. No score changes after a commercial conversation without a logged evidentiary reason.
4. Referral status is disclosed in a dedicated section on every profile, not buried in a footer.

And a standing rule that matters most when someone objects: the rubric changes, or nothing changes. If an objection reveals a genuine flaw we fix the rubric, bump this page’s version, and re-score every profile at once. We do not adjust a single company’s number because that company asked.

Right of reply

Any company listed here may send us corrections and additional evidence. Tier A and Tier B evidence is assessed against the rubric above and the rating is re-derived; the change and its basis are logged. This is deliberately not an invitation to tell us what you disagree with. It is an invitation to give us evidence the rubric can consume. A company may also ask not to be rated at all, in which case the profile carries QS-NR4 and says so. Write to editorial@quotesweep.com.

Staleness

Volatile profile facts are reviewed monthly. This includes company headcount, customer and policy counts, product lines, state availability, licences, issuing carriers, leadership, acquisitions, funding, and commercial relationships. A precise volatile number must carry a source and an as-of date. If we cannot reverify it, we remove the number or retain it only as a clearly dated historical claim.

Employee counts are estimates on many public platforms and change quickly. We do not present an estimated headcount as a current fact unless the source, date, and estimated status are shown together.

Every rating carries an as-of date. Past twelve months without review it moves to QS-NR5 rather than sitting silently stale. Ratings we cannot maintain should not be displayed.

Versioning

This methodology is versioned. Any change to the rubric bumps the version, triggers a re-score of every profile, and is recorded below with what changed and why.

The directory currently contains 42 informational profiles that have not yet been rated under v2.1. That is a publication state, not a negative judgement or a non-rating decision about those companies.

Re-scoring against the v2.1 dimensions is a separate job and is not done. Legacy values remain in the content repository as audit history but are suppressed from profiles, comparisons and structured data. There are currently no published QuoteSweep scores in this directory.

A profile returns to rated status only after it has a claim-level source record, clears every gate condition, is double-scored against the published class anchors, and passes editorial review.