Meelo

The removal test leaves manual file review across several disconnected tools. More than 400 signals are cross analysed in one pass, combining documentary authenticity checks, biometric facial matching and passport chip reading, behavioural analysis of the digital journey including proxy and automation detection, and contextual evidence, with a separate double scoring model assessing company trustworthiness alongside the representative's identity across over 100 control points. The company describes its models as detecting weak signals invisible to manual checks, which is the definition of what the models add.

One sentence states the position more clearly than most vendors manage in a page: critical alerts do not dictate the procedure to be followed but help the analyst prioritise the angles of control. The product reflects it, presenting a summary score and indicating which additional checks would secure the file rather than returning a verdict, so the system directs attention and the analyst decides.

Complete explainability supports that, since an analyst can only exercise judgement over reasoning they can see. What is absent is any description of thresholds or of what proceeds automatically, particularly given the company states it can replace existing processes as well as complement them.

Explainability is claimed without qualification, described as complete even while the models detect signals a human reviewer would miss, which is the harder version of the claim and the one that matters for a regulated decision. Named supervision by a certified practitioner adds accountable ownership of model behaviour. False positive reduction is stated as an objective and speed is quantified at two to five seconds. What is absent is measurement: no accuracy, false positive rate, detection rate or validation result appears anywhere, so the explainability claim cannot be checked against performance.

No customer is named, and two institutional signals carry real weight in their place: the company states it is approved by the largest banking groups, which for a French vendor means passing the supplier assessment of institutions that examine data handling closely, and an external audit classified it as excellent on privacy compliance, which is a third party assessment with a stated result rather than a self declaration. Product cadence is published at new features every two months. Performance is quantified on speed, with most checks completing in two to five seconds and company analysis in under five, and not on outcomes.

No data boundary statement was located. The platform cross analyses more than 400 signals per subject across a client base of banks and lenders competing for the same customers, and behavioural and fraud detection systems typically improve by observing outcomes across that whole base. Nothing states whether a fraud pattern observed at one institution informs scoring at another, whether verification results are retained or reusable, or what a client contributes by participating. Sovereign hosting answers where data sits without answering what is done with it.

This is the strongest privacy position in the index and it rests on verifiable specifics rather than assertions. Infrastructure is stated to be hosted in France and fully sovereign, with European data protection compliance described as by design rather than retrofitted. The company explains the onward transfer chain rather than leaving it implicit, noting that France holds adequacy status for United Kingdom transfers so British customers' requirements are met.

Most significantly an external audit classified the company as excellent on privacy, which is an independent assessment with a published result, and few vendors anywhere in this index submit to one or disclose the outcome.

No information security attestation, certification or trust centre was located. A privacy audit exists and is credited elsewhere in this assessment, and approval by major banking groups implies security review has occurred, but neither is a published control set against a recognised framework.

For a platform handling identity documents, biometric data and bank account information, that is the disclosure a prospective institution would request alongside the privacy assessment already published.

Named instruments carry this grade. European data protection is cited throughout with strict compliance claimed and externally audited, the second payment services directive is named as the basis for bank data access, and customer due diligence and anti money laundering obligations are identified as the regime the platform is designed to satisfy. The adequacy position for United Kingdom transfers is explained specifically. What is missing is the supervisory layer: no financial regulator is named, and for a platform performing regulated due diligence for French institutions that would be the natural next disclosure.

Three elements together make this the most complete bias position in the identity category. The design goal is stated explicitly and addresses the central failure of this technology, that the platform distinguishes risky profiles from legitimate customers even atypical ones, which is precisely the population that document and behaviour based verification tends to reject: people with irregular paperwork, unusual digital footprints, recent arrivals or non standard circumstances.

Models are described as completely explainable, so a rejection can be reasoned about rather than accepted. And governance is named and credentialed, with the artificial intelligence described as ethical by design and supervised by a certified practitioner, which is the only named ethics qualification in this index. Held at B because no testing, outcome analysis or acceptance rate by population is published.

No guarantee, indemnity or correction process was located. The institution is well placed because explainable output lets an analyst reconstruct and justify any decision. The applicant is not addressed, and their exposure is concrete: someone rejected at onboarding because a document scanned poorly, a proxy connection was flagged or their circumstances read as atypical is unlikely to learn which signal caused it, and nothing describes a route to contest a determination or have corrected information reassessed.

The volume of inputs is stated at more than 400 signals and their categories are described as documentary, behavioural and contextual, with bank data attributed to regulated open banking access or statement parsing. Beyond that no individual data provider, bureau, watchlist source, biometric engine or model supplier is named, and no subprocessor list appears, which is a notable omission for a company whose privacy position is otherwise its strongest disclosure and for whom subprocessors are the usual weak point in a sovereignty claim.

Adoption is designed to be additive, with the platform connecting through an interface to enrich existing customer journeys rather than requiring them to be rebuilt, and the company states it works with or without technology involvement, which matters for institutions whose engineering capacity is committed elsewhere.

Data acquisition is handled two ways, through regulated open banking access where available and by parsing account statements where it is not, which is the right design for coverage. Passport chip reading implies device level integration. No named core banking, onboarding or bureau system appears.

This is the clearest residency disclosure in the index. The platform is hosted in France on infrastructure described as fully sovereign, which is a stronger claim than regional hosting because it addresses control of the underlying provider rather than only the location of the servers, and it is the answer European institutions increasingly require.

The company goes further by explaining the consequence for cross border customers, noting France's adequacy status means transfers satisfy United Kingdom requirements. Held at A despite no named provider because sovereignty is asserted as a property of the infrastructure rather than delegated to a hyperscaler region.

No pricing, packaging or basis of charge was located. Cost control appears as a stated benefit of consolidating checks that would otherwise be bought from several vendors, which is the commercial argument, and nothing indicates whether charge falls per check, per verified customer or by subscription, nor how the modular checks are priced against one another.

Functional coverage is the strength, spanning individual verification, business verification, anti money laundering screening, solvency assessment, account validation and documentary control in a single platform, which addresses the fragmentation the company identifies as the problem. Document types handled are unusually specific, covering identity documents, payslips, proof of address, social security certificates, tax notices and utility bills.

Both consumer and business onboarding are served. The limit is geographic: this is a French platform operating under European rules with an explicit bridge to the United Kingdom, and no wider footprint is described.