AML Real Time Screening

AML Real Time Screening matches your customer or business relation data with data in provided central sanction lists and/or lists containing Politically Exposed Persons (PEP) and Relatives and Close Associates (RCA). PEP and RCA commonly referred to as PEP from here on. All screening results which contain one or more matches must be properly evaluated and validated, before proceeding to business processes for how to handle either PEP or Sanction persons or companies.

AML RTS consists of 4 endpoints for screening purposes:

• AML Person Screening
• AML Person Screening with identification
• AML Company Screening
• AML Company Screening with identification

AML Company Screening with identification provides Beneficial Owner and Board Member lookup with simultaneous screening of all added entities.

The national id number of a person and registration id of a company are used for identification, therefore both endpoints with identification must be requested with those numbers, as well as the nationality of the id number. These values are validated for Norway, Sweden, Finland and Denmark. Requests without these numbers result in defaulting to name screening as with the endpoints without identification.

AML Person Screening with identification works currently only for SE individuals.

Authentication

To use the AML Real Time Screening API you need a client ID and a secret. Bisnode uses OAuth2 for authentication. More information https://www.dnb.com/en-gb/developers/authentication/#expand-anchor-first-anchor.

Get Started

You'll need 3 things to get started:

• API Key
• Client secret
• Dun & Bradstreet Direct Plus credentials

Contact api-support@bisnode.com if you don't have them yet.

For easy access we have provided you with a postman collection: https://github.com/Bisnode/rts-postman.

Match quality

Be aware that screening results with one or more matches may very well be a false match; it is not unlikely that some of your business relations or customers has the same or closely resembling name and same birth date as a listed person, without them actually being the same person as present in the list.

Name matching algorithm with Trapets screening

The following characters are allowed but ignored before the matching process is started: *()[]{}~.,'+?\"^\\<>

Two modes are currently supported: “exact mode” and a “fuzzy” matching mode.

For every keyword (name part) that is submitted the search result narrows. The search result will narrow additionally if you submit birthdate. Information about date of birth, year of birth etc may or may not be present in the sanction lists. There may also be information about more than one date (and year) in the same match. These are all returned in the field “birthDateInfo”. There may even be a non-specific date, but just a range of years or range of complete birth dates. The logic matches also against these ranges and adjusts the matchIndicator accordingly.

Exact mode

In the Exact mode of match logic (that is the same as the current system) all name parts in a query must match a post, i.e. person or entity, on the list for a match to be generated. The internal order of the keywords does not matter – the search algorithm checks that every name part matches. "Persson Göran" is a match for "Göran Persson".

The hit must match the exact spelling of the name that is sent in the query. Annie Löf will not generate a match for the list post Annie Lööf.

Hyphen is something the search algorithm ignores (a so-called noise word), so Anna-Karin will match both someone named Anna Karin and Anna-Karin, however Anna-Karin will not generate a match for someone named AnnaKarin.

When it comes to the letters: å, ä, ö – if you search on Goran Persson, you will receive a match for Göran Persson. It does not matter if you spell Göran with o or ö, both spellings will generate a match.

Fuzzy mode

A name similarity for the name in the request and the name or names in the lists. This is done in several steps:

1. Distance calculations are made for each searched name part and the closest corresponding name part in the list item. Distance is calculated using the Levenshtein distance algorithm.
2. The distances for each name part are weighted by the number of letters in each part. The distance for a longer name part has greater impact on the total distance than the distance for a shorter name part.
3. The total distance is reduced with a penalty of 5% if the list item contains one more part than the searched number of name parts.
4. The total distance is reduced further with a penalty of 3% if the list item contains two more parts than the searched number of name parts.
5. The total distance is reduced further with a penalty of 2% if the list item contains three more parts than the searched number of name parts.

Phonetic resemblance is determined by the Metaphone3 algorithm.
When “fuzzy” is enabled it is recommended you supply more than one keyword (name part). Otherwise the number of false positives might become overwhelming. To improve the results of your hits use a combination of name and birth date whenever possible.

All matches with nameMatchValue lower that 0.6 will be ignored (unless there is a match by national id only).

Name matching level (nameMatchLevel) and Name match value (nameMatchValue)

The name matching level (mode) may be set by the “nameMatchLevel” argument in the request. A lack of this argument, or the value set to 0, results in the default “exact” level. Any value above will enable the “fuzzy” mode. Be prepared for the future enhancements by using the value of 5.

The response will contain a “nameMatchValue” in which ranges from 0.0 to 1.0. The higher the value is, the better the match was.

Match by nationality is currently only available for Swedish persons.

DnB Europe performs the matching with ElasticSearch framework, using a combination of different algorithms to support fuzzy matching (recognizing typos, spelling variations, etc.) and phonetic similarity. DnB Europe applies own match indicator and similarity score to the matching results to help compare input and output and recognize false matches, i.e. persons or companies that resemble the listed ones, but are not the same.

Similarity score with international screening

DnB Europe Similarity score is a general match quality characteristic that indicates how good was the match.

It is a value from 0 to 1, the higher the value the more similar is the input to the name on the list. Values equal to 1 can be considered exact matches – the name in the input is spelled exactly as the name in the list. Values below 1 are matches with different degree of fuzziness. 

Similarity score is designed to be a combined value, consisting of several components. It is currently under development. As of now, nameScoring is available: 

• nameScoring – indicates how good the name matching is, comparing the name from the input and the name on the list.

The name matching level may be adjusted by the “minimalSimilarityScoreLevel” argument in the request. A lack of this argument, or the value set to 0, results in the default “exact” level. Any value above 0 or below 1.0 will enable the “fuzzy” matching mode.

It is provided with hits against info4c lists.

Setting minimalSimilarityScoreLevel

The minimalSimilarityScoreLevel is set by DnB Europe for Real Time Screening account during onboarding. In the case when minimalSimilarityScoreLevel is not specified, the default value of 1 (exact match) will be set.

The default value will be overwritten every time the minimalSimilarityScoreLevel argument is provided in the request for screening.

The default value for the account can be changed at any time through DnB Europe team upon request.

Source lists available

The following is an overview of lists that are available in our services.

Board member role codes and titles

Be aware that a person can be a board member in a company of another nationality.

For example: A Swedish company may have Finnish and Norwegian board members. As the company is Swedish, the role codes will be as specified for Sweden regardless of the nationality of the person.

The board member information and the role codes are information about the company, not information about the person.

This also means that a person who is a board member of several companies of different nationalities will get different role codes, even though they mean the same.

Norway

The directorships for Norway are delivered through the special structure in the field directorships with a specific label and role description. The Board Member can hold roles in more than one company and the relation with each company will be provided. There are no codes in this case.

Denmark

Currently these are only described in Danish.

The system only delivers Danish titles. The actual title (the text) may differ slightly from the below descriptions, so consider the below descriptions just as that; a description.

Use the codes to identify the roles. Remember these codes are for Danish companies only. Other countries may have the same role codes meaning a different role.

Sweden

The system only delivers Swedish titles. The actual title (the text) may differ slightly from the below descriptions, so consider the below descriptions just as that; a description.

Use the codes to identify the roles. Remember these codes are for Swedish companies only. Other countries may have the same role codes meaning a different role.

Finland

The system only delivers English titles. The actual title (the text) may differ slightly from the below descriptions, so consider the below descriptions just as that; a description.

Use the codes to identify the roles. Remember these codes are for Finnish companies only. Other countries may have the same role codes meaning a different role.