Data and models
The sources and models behind Enin.
Enin connects public registers, accounts and licensed news, analyses them with its own models and turns the result into signals. You always see the source and the actual data, and the models show what they may indicate. Here are the sources, the models and the limitations.
From source to signal
The sources, Enin’s analysis and the signals you get.
The sources
Data from public registers is regularly updated against the source.
| Source | What we collect |
|---|---|
| Central Coordinating Register and Register of Business Enterprises | Roles, purpose, signatory rights, procuration, announcements, mergers and demergers |
| Register of Company Accounts | Annual accounts and auditor notes |
| Register of Beneficial Owners | Beneficial owners |
| Norwegian Tax Administration shareholder register | Shareholders and ownership shares |
| Kartverket and the Register of Movable Property | Property ownership, registered encumbrances and charges |
| Public sources on court cases | Court hearings and cases |
| Norwegian media houses | News articles mentioning the company, from media houses Enin has licence agreements with |
| ComplyAdvantage | PEP and sanctions lists |
| KAR register and Swift | Who an account number belongs to |
Property data, news and PEP and sanctions data are only available to customers with an agreement for them. We do not process credit information about private individuals or sole proprietorships.
The models
Enin’s own models turn the data into risk you can act on.
Bankruptcy risk
An in-house machine learning model estimates the probability that a company goes bankrupt within the next twelve months.
The result is shown as a probability and a level from very low to very high. The model is continuously updated, and you see how the risk has developed over time.
The estimate is model-based decision support and does not replace your own credit assessment.
Fraud risk
Fraud risk is one score from 0 to 100, built from several risk components, with a level from very low to very high.
The model is continually tuned and updated to catch new fraud trends.
A high score is a signal of what to look into, not proof of fraud.
Network risk
The network model maps the people and companies connected to the company through roles and ownership, up to three steps out.
The model weighs risk signals across the entire company network, not just in the company itself, and combines them into one network risk.
More about network risk →Company and person flags
Flags mark high-risk events, such as changes in the board, management, ownership and accounts, weak finances, unusual register events and people with roles in earlier bankruptcies.
You see them on the company and person pages in the web portal.
- Unusual changes in the registers
- Weak finances in the latest accounts
- Changes in ownership
- Irregularities in the accounts
Customizable credit risk model
Enin can set up a credit risk model configured for each customer, built on Enin data and tailored to the customer’s criteria.
Talk to us about the model →Event ranking
Machine learning links the most relevant events and news to the right company and ranks them by how important they are.
That way your alerts are about what matters.
- Payment remark: new debt collection casePayment remarksHigh
- Accounts re-registeredRegister of Annual AccountsMedium
- Change of company nameCentral Coordinating RegisterLow
- Major board changeRegister of Business EnterprisesHigh
- Approved annual accounts filed after the deadlineRegister of Annual AccountsMedium
Limitations and responsible use
Enin is decision support. This is how the signals should be used.
Decision support, not the answer
Risk scores are model-based estimates and do not replace your own assessments, in credit or in customer due diligence.
Facts and signals
You always see the source and the actual data. Flags and risk scores are the models’ assessment of what the data may indicate, and some signals are clearer than others. A signal is a starting point for looking closer, not in itself proof of fraud.
Name matches need review
A match in PEP or sanctions screening is a starting point for review, not a conclusion.
A charge is not outstanding debt
Registered mortgage amounts are not the same as outstanding debt, and on their own do not give grounds for calculating the loan-to-value ratio.
Errors are fixed at the source
Data from public registers is regularly updated against the source. If something is wrong at the source, it has to be corrected there, for example in the Brønnøysund Register Centre.
Businesses only
We do not process credit information about private individuals or sole proprietorships.
Questions about the data?
We are happy to explain the sources and models in more detail, also as part of a supplier review.
Kristoffer is happy to have a no-obligation chat, by email or phone.