Choisissez votre pays ou votre région

Risk scoring in account opening, explained.

Factors, tiers and two-person decisions: how an explained risk score decides who gets an account online.

By Everbranch team · Published · Updated · 5 min read

When a customer applies for an account on their phone, someone – or something – decides how much checking they need. In a branch, that judgement sat with an experienced officer. Online, it sits in a risk score. If nobody can say why the score came out the way it did, your compliance team can’t defend it, your auditors can’t test it and your customers can’t challenge it.

This guide explains how an explained risk score works in digital account opening: what goes into it, how it routes applications, where people stay in charge and what happens after the account is open. The examples come from how onboarding and risk and compliance work in Everbranch, but the questions apply to any system you evaluate.

What a risk score is – and isn’t

A risk score is a number built from facts about the applicant and the checks you ran on them. It is not a verdict. Its job is to sort applications into tiers so that the low-risk ones move quickly and your team spends its time on the ones that need judgement.

The facts come from two places. The customer’s answers: who they are, where they live, which products they want. And the checks: identity verification (KYC), AML screening against sanctions lists and politically exposed persons, and for businesses, company registry checks (KYB). The score adds these up under your bank’s policy, the same way every time. Same facts, same policy, same answer.

Factors and weights: every point has a reason

An explained score is a list of factors, each with a description and its points. The points always add up to the score, so anyone reviewing the case can see exactly where it came from. Typical factors:

  • A base score for the kind of applicant, a person or a business
  • Points for the product applied for, if some products carry more risk
  • Points for each high-risk country connected to the applicant, through nationality, residence or tax residence, naming who it connects
  • Points for each possible sanctions, PEP or adverse-media match
  • For businesses: the legal form, the sector, and ownership that isn’t explained by the owners declared

The weights are yours. A bank in the Gulf serving expatriate workers and a microfinance bank in East Africa will weigh countries and products differently, and they should. What matters is that the weights live in a versioned policy, not in a vendor’s black box.

Four tiers, and where each one goes

Thresholds turn the score into one of four tiers. Each tier carries its own routing, document requirements, payment limits, review deadline and how often the customer is reviewed again.

Risk tiers in a typical published policy
TierWhat usually happens
LOWApproved automatically, if your policy allows it, every check came back clear and at least one check ran
MEDIUMA person reviews it; your policy can ask the customer for documents straight away
HIGHA person reviews it, and approval needs a second person – never automatic
PROHIBITEDA prohibited country or sector forces this tier; it goes to a person with a recommendation to decline

Some rules hold whatever the policy says. A possible sanctions match always goes to a person, even if a lax policy gives it no points. Any finding, any provider error and an application with no checks at all go to a person too. HIGH and PROHIBITED never approve automatically and always need four-eyes approval.

A provider outage is not a finding about your customer

Identity and screening providers go down. Networks time out. In markets where connectivity is uneven, this happens more than anyone likes. A system that treats a timeout as a failed check rejects good customers for reasons that have nothing to do with them.

So a provider error is neutral. It is shown on the case as “not a finding about the customer”, the application goes to a person, and nothing is declined because a provider didn’t answer. Results that arrive late, by webhook or by polling, are picked up when they come; a check that never answers times out to the same neutral outcome. Your providers – Sumsub, Smile ID, uqudo or a national scheme – plug in behind the same rules.

People decide: two of them for the big ones

The score recommends. Your team decides. In the operations portal, a reviewer sees the application, every check result and the full explanation: factors, points, the rules that applied and the policy version. They can ask the customer for more information, approve, or decline.

Approving a HIGH-risk application needs a second person, never the one who asked, and it is checked against the snapshot they both reviewed. The policy itself follows the same rule. You can dry-run a draft against sample applications to see how tiers would move, but publishing it needs a second person who didn’t edit it last, and a published version never changes. Every decision is in an audit trail nobody can edit. Our security page lists these controls.

Automatic decline, only if you turn it on

Some cases are clear-cut. A bank may switch on automatic decline in its four-eyes policy, for five rules only: a prohibited nationality, residence or country of incorporation, a prohibited sector, or an applicant under the bank’s minimum age. A match, a score, a provider error or missing data never declines anyone. The customer gets a message saying the decision was automatic and can ask once for a person to review it, and staff can reopen it within the period you set.

Businesses: the company and the people behind it

For a business, the score looks at the company and its people. The company is checked against the registry, and the applicant declares its owners and directors. Each ultimate beneficial owner is a person in their own right, screened like any customer – and one person can hold roles in many businesses without being entered twice. Ownership that the declared owners don’t explain, or an unusually long list of owners, adds points your reviewers can see.

The score keeps working after the account is open

Risk changes. A customer moves country, a screening list is updated, a document request is answered. Perpetual KYC re-assesses the customer when new information arrives, when you publish a new policy and on a schedule set by their tier. A move into a higher tier opens a review for your team; nothing here declines, closes or offboards anyone on its own.

If you are building the app side too, how to launch a mobile banking app covers the rest of the journey, and comparing building in-house shows what an explained risk engine takes to build from scratch.

Pilot programme

Bring your risk policy. We’ll score it.

We’ll set up your factors, weights and tiers in a demo and run sample applicants through them with your team.

  • Your name and colours in the demo
  • Straight answers for risk, compliance and IT
  • No commitment