FRISS

Nothing survives the removal test and nothing else was ever built. The company states a hundred percent dedication to fraud, risk and compliance for non life insurers, and has held that focus since 2006, so there is no policy administration system, no claims workflow engine and no billing platform underneath to persist if the scoring were taken away.

Every product is a score: an application assessed for misrepresentation in the seconds it takes to complete a form, a claim scored at first notice of loss and again across its lifecycle, a renewal reassessed. The investigation tooling exists to work the cases the models flag, so it is downstream of the models rather than independent of them.

The company's own framing depends entirely on automated assessment, since clearing trustworthy business quickly at scale is only possible if something is judging trustworthiness automatically, and screening every interaction within seconds rules out manual review as the mechanism.

The structural position is flag and hand off, confirmed independently, with one customer statement that complicates it. An independent 2026 assessment of this category states the position plainly for every vendor in it including this one: they flag suspicious claims and hand them to a human investigator, and none occupies the investigation layer downstream of the flag, which remains manual.

So on the claims side a person adjudicates by construction, and the company's investigation tooling exists to support that person rather than replace them. The complication sits at underwriting. A published customer states that the platform is automatically consulted for all new policy applications and that the results of its screenings are decisive in helping decide what is accepted, reviewed or rejected, and screening happens in the seconds an applicant takes to complete a form. Decisive is the customer's word, not the vendor's, and it indicates a score materially driving an accept or reject on a live application. Nothing published states a floor beneath that.

Strong operational outcomes, no detection measurement, and one structural limitation disclosed only by a third party. The operational figures are among the better constructed in this index, particularly savings per investigator rising from 550,000 dollars to 2 million, which measures analyst effectiveness rather than gross recovery and is harder to flatter. What is missing is everything about whether the scores are right.

Across two passes no detection rate, false positive rate, precision or recall measure, validation methodology, sample or observation period was located, and for a product screening every application and claim the false positive rate is the number that determines how many honest customers are delayed or refused.

The limitation a buyer most needs is disclosed by an independent assessment rather than by the vendor: the cross carrier contributory data network behind the models is smaller than the market leader's, so the shared industry signal feeding detection is thinner, which bears directly on performance.

Deep deployment history and one outstanding outcome figure, held below the top band by the absence of a named customer. The deployment record is substantial and specific: more than 300 implementations across more than 45 countries over twenty years, roughly 223 staff, and 81 million dollars raised including a 65 million dollar Series B led by a named technology investor.

The outcome figure is better constructed than most in this index because it is expressed per capita rather than as a gross total: a customer reports 21 million dollars of total fraud savings within two years of going live, and separately that savings per investigator rose from 550,000 dollars to 2 million, which measures whether the tool made people more effective rather than only whether it found things.

Independent placement is current, with a 2026 buyer's assessment naming it the mid market detection vendor most commonly shortlisted by United States carriers alongside the industry data utility. Every customer statement published is unattributed to any named institution.

A published responsible artificial intelligence position, which is rarer in this index than it should be. The company sets out four principles in its own material: transparency, meaning systems should explain how they reach decisions; fairness, meaning artificial intelligence should treat everyone equally and not be biased; accountability, meaning developers are responsible for their systems; and governance, meaning clear rules and oversight over development.

It engages the European artificial intelligence regulation by name and identifies its enforcement timeline rather than mentioning it in passing. Around that sits a stated purpose that shapes the product rather than decorating it, namely that honest customers should not pay more because others commit fraud, which is why the platform is framed around clearing trustworthy business quickly rather than around volume of fraud caught. Against that, principles are not practice. Across two passes no model card, evaluation methodology, red team result, incident disclosure or acceptable use boundary was located.

A Dutch company operating under European data protection law with nothing published about how it does so. Across two passes no privacy policy content, data processing description, retention schedule, subprocessor list or lawful basis statement was located on any surface.

The data in question is not incidental to the product but is the product: every application and claim screened produces an assessment of an identified individual's trustworthiness, retained in a system the insurer consults again at renewal, and the company describes sharing learnings between underwriting and claims, which means an assessment formed in one context can follow a person into another.

Nothing published describes how long a risk assessment persists, whether a flag raised at claim time affects a later application, or what a policyholder is told or can request about a score attached to them. For a vendor headquartered in a jurisdiction with strong automated decision making protections, the silence is conspicuous.

A disclosure gap rather than a security one, with the inference resting on deployment history. Across two passes no trust centre, named certification, attestation report, penetration test summary or subprocessor list was located on any public surface.

The controls behind that silence are very likely substantial: a company operating since 2006 with more than 300 implementations at insurers across more than 45 countries has passed each of those carriers' third party risk assessments, several of them under European supervisory expectations, and a private equity led Series B involved technical diligence.

That places this alongside the established vendors in this index whose empty security page can be discounted, rather than alongside the young ones where nobody demanding has demonstrably looked. It remains unestablishable from outside, and the practical cost falls on the mid market carriers the company targets, who have the least procurement leverage to compel documents and the least capacity to assess them.

An unregulated software vendor that engages the regulation governing its own product more openly than most of its peers. Published material addresses the European artificial intelligence regulation by name, describes it as creating a legal framework promoting trustworthy development, and identifies when it becomes fully enforceable, which is more than several larger vendors in this index manage anywhere on their public surfaces.

That engagement matters because insurance risk assessment and pricing fall within the regulation's high risk designation, so this company's own products sit inside the regime it is writing about. What is absent is any statement of its own position under it.

Across two passes no conformity assessment declaration, no technical documentation reference, no statement of role as provider or deployer, and no account of how its systems meet the human oversight, accuracy or record keeping obligations was located, and no financial services authorisation or supervisory outcome exists.

The company has published the standard it claims to meet and no evidence that it meets it, which is a more specific gap than the usual silence. Two published statements bear directly. Its responsible artificial intelligence material defines fairness as artificial intelligence treating everyone equally and not being biased. And its underwriting product is marketed as enabling faster, consistent and unbiased policy writing by identifying high risk applicants through real time analytics.

Unbiased is an affirmative claim about outcomes, made about a system that scores individual applicants for trustworthiness in seconds and whose results a customer describes as decisive in accept, review and reject decisions. Across two passes no bias testing, disparate impact analysis, fairness metric, model card or conformity assessment was located. The exposure is direct: an applicant scored high risk may be refused cover, and a claimant flagged faces investigation, with neither told a model made the assessment.

No commercial instrument is published. Across two passes no terms of service, master agreement, warranty, indemnity, liability cap or service level was located, and nothing states what an insurer is owed if screening misfires in either direction. The uncovered exposure divides between two parties who are affected differently. The carrier bears a commercial loss from missed fraud and an operational cost from excessive flagging, neither addressed.

The applicant or claimant bears something worse and has no relationship with this vendor at all: an application scored high risk may be refused or repriced within seconds of submission, on a determination a customer describes as decisive, and a flagged claimant faces investigation and delay. Neither is told a model produced the assessment, neither can see it, and nothing published describes a correction route. For a company whose stated purpose is that honest customers should not pay for others' fraud, the honest customer wrongly caught has no remedy on the public record.

Almost nothing is disclosed by the vendor, and the one substantive fact available comes from a third party. Across two passes no model provider, technique, architecture, framework or named data supplier was located on any vendor surface, so a buyer cannot establish whose models score their applicants, what external data enriches the assessment, or which of their existing data subscriptions might be duplicated inside the platform.

The single meaningful supply chain observation available is an independent one, that the company's cross carrier contributory data network is smaller than the market leader's and the shared industry signal is therefore thinner. That is exactly the kind of dependency disclosure this axis asks for, since contributory data is the input a detection vendor cannot manufacture alone, and it is notable that a buyer learns it from an analyst rather than from the vendor. Nothing published describes what happens to the contributory pool if a large carrier leaves it.

Distribution through the core system most of the target market already runs, plus an interface people evidently want to use. An independent 2026 assessment attributes the company's growing United States footprint specifically to integrations with the dominant claims core platform, which is the route that matters in this market because it puts screening inside the system where claims are actually handled rather than beside it.

The company states that carriers can expect seamless integration and quick time to value, and the deployment record of more than 300 implementations across more than 45 countries indicates that has held across very different technology estates. Real time performance supports embedding rather than batch use, with screening completing within the seconds an applicant spends filling in a form.

The same independent assessment credits the workflow interface as one investigation directors consistently find easier to live with than enterprise alternatives. Across two passes no public interface documentation, developer portal or connector catalogue was located.

Hosted software with the placement questions unaddressed. The platform is described as a software as a service solution and one customer is identified as the first online insurer to adopt it in that form, so the delivery model is clear enough. Beyond that, across two passes no hosting provider, region list, tenancy description or residency commitment was located, and no customer hosted option is described.

The gap carries more weight for this vendor than its size suggests because of where it operates. It is a Dutch company running screening for insurers across more than 45 countries, including markets inside the European Economic Area where insurance regulators and data protection authorities both take an interest in where policyholder data is processed, and its United States presence in Ohio implies transatlantic processing arrangements that are nowhere described. What is screened is not incidental data but every application and claim an insurer receives.

No price, unit or tier appears on any vendor surface across two passes, and the published entry route is a contact request. What is available comes from outside the company and is recorded here as third party estimate rather than disclosure: an independent 2026 buyer's assessment describes the commercial model as subscription based, custom quoted, with mid market deals typically falling between roughly 300,000 and 1.5 million dollars annually depending on volume.

That is a genuine order of magnitude a prospective buyer can plan against, and it is more than most vendors in this lane leave available anywhere, but it is an analyst characterisation and neither the vendor nor a customer confirms it. What remains unknown from any source is the unit.

Volume based could mean policies screened, claims screened, premium under management or investigators seated, and those scale very differently for an insurer weighing whether to screen every application or only a segment.

Wide geographically, deliberately narrow by line, and concentrated in the middle of the market. Geographic reach is real rather than aspirational, with more than 300 implementations across more than 45 countries and offices spanning the Netherlands, the United States, the United Kingdom, France, Spain, the German speaking markets and Latin America, which is broader physical presence than most vendors of this size maintain.

Lifecycle coverage is complete within its scope, addressing underwriting at application and renewal, claims at first notice of loss and beyond, and the investigation function that works the output. The narrowing is deliberate and stated: property and casualty and non life only, with no life or health lines, and an independent assessment identifies automotive as the strongest line and positions the company in the mid market rather than at the top of it. No individual insurer is named anywhere, in any market, which leaves the depth of any single relationship unestablished.