Risk builds up between processes, where no single system can see it. The payment looks fine. It is the link to the contract, the grant or the person who approved it that is wrong. By the time an audit finds it, the money is gone.
This is how Namuda shows where to look.
Why checks inside one system miss it
Most controls test one table at a time. Audit analytics tools test transactions table by table, not how they connect across processes. Audits sample, a year late. A dashboard shows totals, not the record or why.
Each of those can pass while the problem sits in the connection: an invoice that is correct in itself, paid before its contract was signed.
Rules that look across processes
Namuda reads data as objects and the events between them. A payment belongs to an invoice, a purchase order, a project, a grant and a vendor. A claim belongs to a policy, a customer and an adjuster.
Because the links are in the data, a rule can ask questions that span them:
- Was this invoice paid before its contract was signed?
- Was this credit hold released by the person who placed it?
- Was this claim paid on cover that had lapsed?
- Was this settlement approved by its own assessor?
Every rule runs on every record, at each data load.
A risk score per unit
Risk is a question about concentration. Which branch, sales office, country office or loan purpose breaks the rules more often than its peers, and what drives it?
Each unit gets a score from the weighted rules, measured against its peers. A limit, for example 40, raises a signal when a unit goes above it.
Ranked by what is at stake
A long list of signals helps nobody. Namuda ranks them by impact: money where the data holds amounts, otherwise the number of records affected, with confidence beside it. The top tenth is marked important.
Indicators, not verdicts
A risk signal says where to look and why, and shows the records. It does not say what happened. Namuda shows where irregularities are likely and which records to check. Your people decide what it means.
Risk and compliance ask different questions
They use the same rules. Compliance asks whether a rule was followed. Risk asks where the breaks concentrate, and what drives them.
What it looks like on data
Lending. On the public Dutch consumer-lending log, all 12 loan purposes score above the limit of 40. 4,323 applications worth 72.7 million euros trigger at least one rule. The largest driver is offers cancelled after 14 or more days without contact.
Sales orders. On a sales-order demo, 17 of 21 sales offices score at or above 40. The drivers are credit holds released by the person who placed them, and orders closed with a hold still open.
Insurance. On a synthetic log of a fictional Norwegian insurer (October 2025 to May 2026), 1.5 million kroner was paid on lapsed cover. All of it sits in one office, up from 0.3 million the period before.
Donor-funded organisations. For one donor-funded organisation, a suite of 17 rules runs across grants, projects, procurement and invoices. Each country office is scored per year on the same terms, matching the organisation's own risk model.
From signal to action
One click turns a signal into an issue with an owner, with the rule's checks as tasks. The analyst agent runs the checks and comes back with the real records. Records the data explains are cleared, with the evidence.
The issue is verified only when the signals are gone, or clearly smaller, at a later data load.
How to start
A fixed-price pilot on one process, on data you already have, through a connection or an extract.
Talk to us about where you suspect risk is building up.