Choosing your KYB solution: the 7 criteria of real business verification

In short: choosing a KYB solution in 2026 comes down to seven criteria: the depth of local data, the ability to detect fraud (not just tick the compliance box), the identification of beneficial owners, response time, the explainability of the AI, data sovereignty and the ability to extend. The right tool depends on your volume, your markets and your exposure to fraud.

KYB (Know Your Business) is no longer a mere regulatory checkbox. Between the AMLD6 package applicable on 10 July 2027, the DSA requiring marketplaces to verify sellers since 2024, and organised fraud networks that set up shell companies in a matter of hours, business verification has become a strategic function. Here is how to evaluate a solution, and the exact questions to ask in a demo.

1. Data depth on your actual markets

"200 countries" coverage looks impressive on a brochure, but your risk is concentrated where your customers are. On the French market, for example, the difference comes down to access to official registers (INPI, Sirene, Infogreffe, RBE) but above all to historical depth: a solution able to compare a filed balance sheet against the full filing history detects a copied balance sheet, the classic signature of an organised network, where a simple existence check sees nothing.

Question to ask: "On my main market, which sources do you query, with what historical depth, and in real time or in batches?"

2. Fraud detection, not just compliance

Many KYB tools verify that a company exists. Yet professional fraudsters use companies that exist perfectly well: reactivated shell companies (several balance sheets filed the same day to simulate a track record), companies registered at addresses hosting hundreds of entities, recent changes of management just before a financing request. A serious KYB cross-references these weak signals and returns them as an actionable score, then completes them with dual scoring: the company on one side, its director on the other, because behind a shell company there is always a natural person whose identity deserves the same scrutiny.

Question to ask: "Show me how your tool reacts to a company reactivated three months ago with a change of management."

3. Beneficial owners (UBO), continuously

Identifying UBOs is an AML/CFT obligation, and the AMLD6 package tightens it: continuous rather than one-off verification, European interconnection of registers (BORIS), minimum penalties of €1,000,000 for failures. The differentiating criterion is no longer "can you identify UBOs?" but "do you detect a change in ownership or an entry onto a sanctions list after onboarding?" That is perpetual KYB (pKYC applied to companies).

Question to ask: "What happens if a UBO of an existing customer becomes a politically exposed person six months after onboarding?"

4. Response time and automation rate

A manual KYB check takes from a few hours to several days; at scale, that is a commercial bottleneck. The best solutions return a documented decision in a few seconds and escalate only ambiguous files to an analyst. Two figures to demand in a demo: the actual response time (not the "marketing average") and the share of files processed without human intervention at clients comparable to yours.

Question to ask: "For a client in my sector and at my volume, what percentage of files goes through automated decisioning?"

5. Explainability of the AI, now non-negotiable

If the scoring contributes to a granting decision (credit, financing, payment line), the AI Act and supervisors' expectations (ACPR) require you to be able to justify every decision. A "black box" score exposes you twice over: impossible to defend in an audit, impossible for the customer to contest. Demand explainable AI: every score must come with its reasons, and the audit trail must be exportable.

Question to ask: "For this score of 23/100, give me the exact list of signals that explain it, exportable for my audit."

6. Sovereignty and data hosting

KYB data includes personal data (directors, UBOs) subject to the GDPR. Hosting in France or in the EU, with no transfer outside the Union, simplifies the impact assessment, reassures your own compliance teams and is becoming a knockout criterion in most banking and public tenders.

Question to ask: "Where is the data hosted and processed, and through which legal entities does it pass?"

7. The ability to extend beyond KYB

The need almost never stops at KYB. The same file calls for an IBAN verification before the first transfer, a creditworthiness score before granting, a document check on the tax return package. Stacking one tool per need means multiplying integrations, contracts and decisions that are impossible to reconcile. Check whether the solution covers these adjacent cases in the same API, or, if you choose a specialist, how it orchestrates with the rest of your stack.

Question to ask: "If tomorrow I add IBAN verification and creditworthiness scoring, how many additional integrations and contracts?"

The decision grid at a glance

• Local data: the minimum, official registers; the differentiator, deep history and compared balance sheets.
• Fraud: the minimum, company existence; the differentiator, shell companies, networks, director score.
• UBO / AML: the minimum, identification at onboarding; the differentiator, continuous monitoring (pKYC).
• Speed: the minimum, response < 1 h; the differentiator, documented decision in seconds.
• AI: the minimum, a score; the differentiator, a score + its reasons, exportable (AI Act).
• Data: the minimum, EU hosting; the differentiator, sovereign hosting in France.
• Extension: the minimum, KYB alone; the differentiator, IBAN, creditworthiness, fraud in the same API.

How Meelo meets these 7 criteria

Meelo is a French customer onboarding platform that treats KYB as a fraud problem as much as a compliance one: from a single SIREN (French company number), the Company Trust Score cross-references financial statements (compared against the filing history), litigation, registered addresses and the mapping of the ecosystem (the director's other companies, insolvency proceedings, recent strike-offs), and returns in 2 to 5 seconds a score together with its reasons (explainable AI audited by IEEE, data hosted in France). In dual scoring, the Identity Fraud Score verifies the director themselves from their contact details: the two checks are always recommended together. The same contract covers IBAN analysis, open banking creditworthiness and AML screening with continuous monitoring. Documented result at a financing player: fraud divided by 12. To see these controls on your own cases, you can schedule a demo.

FAQ

What is the difference between KYB and KYC?

KYC verifies the identity of a natural person; KYB verifies a business: legal existence, directors, ownership structure and beneficial owners. In practice the two are inseparable: verifying a company without verifying its director leaves the door open to shell companies.

Is KYB mandatory?

Yes for entities subject to AML/CFT obligations (banks, insurers, fintechs, credit players), and since February 2024 for marketplaces via the DSA (verification of professional sellers). The AMLD6 package, applicable on 10 July 2027, harmonises and tightens these obligations across the EU.

How much does a KYB solution cost?

Most vendors charge per verification or via a SaaS subscription based on volume, with significant differences depending on the depth of the checks (simple existence vs full anti-fraud analysis). Always compare the cost per actionable decision, not per raw request.

How long does a KYB check take?

From several days manually to a few seconds with an automated solution. The right indicator is end-to-end time including cases escalated to an analyst, not just the API response time.

Does a KYB tool detect shell companies?

Not always: many stop at confirming legal existence. Detecting shell companies requires specific signals, such as identical balance sheets or bulk filings, mass registered addresses, recent changes of management, a director linked to struck-off companies, all cross-referenced in a score.

‍