Business intelligence for risk teams: from raw company data to decisions

Business intelligence in a risk context is not reporting on what your own systems already know. It is combining external company data — registry records, financial filings, ownership, payment behaviour, sector and country context, sanctions and news — with your internal exposure, so that a decision about a counterparty can be made on evidence rather than on familiarity.

The four layers that matter

Identity: which legal entity is this, and is it the one on the invoice? Structure: who owns and controls it, and what else in the portfolio belongs to the same group? Capacity: what do the financials, the ratios and the payment record say about ability to pay? Context: what is happening in the sector, the country and the news that changes the answer? Most reporting stops at the first layer and calls it a customer view.

Entity resolution comes first

Nothing downstream works until records match the right entity. Trading names, transliterations, branch registrations and near-identical group companies routinely split one exposure into three, or merge two unrelated counterparties into one. Matching on registration number and jurisdiction rather than on name is the single highest-value data-quality investment a risk team can make.

Aggregate exposure to the group

A portfolio view that lists legal entities separately understates concentration. Rolling exposure up to the ultimate parent regularly reveals that the largest counterparty is not the one anyone named — and that is exactly the number a board wants before a downturn, not after.

Freshness beats sophistication

A simple score computed on current filings and last month's payment behaviour is more useful than an elaborate model on two-year-old accounts. Record when each field was last verified, show it in the interface, and let users see the age of the evidence they are acting on.

Design for the decision, not the dashboard

Every view should end in an action: approve, set a limit, request security, escalate, decline, or monitor. Analytics that do not change what someone does are a cost. The practical test of a risk BI programme is whether the number of avoidable losses falls, and whether decisions take less time.

— RisQo Research Team, Infocredit Group