How are the AML/PEP/Sanctions screening scores calculated?

This article explains the GlobalPass AML Screening Weighted Percentage Match scoring system, used to determine the percentage probability of a screened entity being a true match to a detected possible hit.

Refer to the table below for a quick overview. For more in-depth explanations, please refer to the sections below.

Once an AML Screening of an Individual is initiated, the entire database of watch-listed names is screened for name matches with 85% similarity and up, using fuzzy logic principles. 

Names of 85% and higher name match are listed in the GlobalPass AML reports as potential hits. Each of these hits are then further checked for full name, gender, date of birth and country matches, and assigned a final match score based on detected similarities, as described in the following section.

For more information about the screened database and fuzzy logic, please contact our support. 

For maximum matching accuracy, GlobalPass screens up to four pieces of personal data together: 

For each of the found potential hits, a percentage match score model is used to determine the final match score, which consists of individual match scores multiplied by the weights of all four of these data pieces. 

Full name match score is determined by the fuzzy match score, which will be 85% or higher. 

To get the input of the name match for the final match score, the name match is multiplied by the weight of 0.6. 

Gender match score is determined as follows: 

To get the input of the gender match for the final match score, the gender match percentage is multiplied by the weight of 0.1. 

Date of birth match score is determined as follows: 

To get the input of the gender match for the final match score, the date of birth match percentage is multiplied by the weight of 0.2. 

Country match score is determined as follows: 

To get the input of the nationality / country match for the final match score, the country match percentage is multiplied by the weight of 0.1. 

Having determined the percentage match score of each of the four data categories of the found possible match, the assigned percentages are used to determine the final percentage match score, which is built taking the following weights: 

The calculation for the final weighted percentage match score is the following: 

The closer the final percentage match score is to 100%, the higher the probability of the possible match being a true match.

Based on the final match score, GlobalPass also defines four match strength categories: 

In cases where not all four pieces of personal data are provided, the skipped data piece(s) are not included in the final calculation, and the final score is divided by the sum of weights of the provided data.

Example: only full name and date of birth are provided:

For further explanation of the weighted percentage match scoring system, the following hypothetical person with hypothetical matches will be used: 

Four name matches of 85% and higher strength are found, therefore they appear on the AML report and final match score calculations are run on them: 

1st example – 100% overall match:

Overall match = 100% × 0.6 + 100% × 0.1 + 100% × 0.2 + 100% × 0.1 = 100% 

2nd example – 91% overall match:

Overall match = 94% × 0.6 + 100% × 0.1 + 75% × 0.2 + 100% × 0.1 = 91% 

Overall match = 89% × 0.6 + 100% × 0.1 + 75% × 0.2 + 100% × 0.1 = 88% 

4th example – 80% overall match:

Overall match = 91% × 60% + 100% × 0.1 + 75% × 0.2 + 0% × 0.1 = 80% 

Once an AML Screening of a Business is initiated, the entire database of watch-listed company names is screened for:

Watch-listed companies found based on one or both criteria above are listed in GlobalPass AML reports as potential hits. Each of these hits are then checked for company name, registration number and country matches, and assigned a final match score based on the detected similarities, as described in the following section of this guide. 

For more information about the screened database and fuzzy logic, please contact our support.

For maximum matching accuracy, GlobalPass screens up to three pieces of business data together: 

For each of the found potential hits, a percentage match score model is used to determine the final match score, which consists of individual match scores multiplied by the weights of all three of these data pieces and further rules described below. 

Full name match score is determined by the fuzzy match score. It is important to note that it is possible that a potential hit will not have any name match, if the potential hit was found purely based on registration number match. 

To get the input of the name match for the final match score, the name match is multiplied by the weight of 0.45. 

Registration number, to be considered a match, must show a full match in correct digit order. 

To maximize the search and avoid false mismatches due to country and source data differences, screened registration number is being matched to any provided legal identifier in the source (e.g., if screened registration number is matching to TIN in the source – GlobalPass shows a match). 

Registration number match score is determined as follows: 

To get the input of the registration number match for the final match score, the registration number match percentage is multiplied by the weight of 0.1. 

❗ Important exception: if both registration number and country are matching, the final match score will be 100% no matter the company name match. 

To maximize the search and avoid false mismatches, the system is looking for country matches across any country where the company is known to hold a registered address. 

Country match score is determined as follows: 

To get the input of the country match for the final match score, the country match percentage is multiplied by the weight of 0.45. 

❗ Important exception: if both registration number and country are matching, the final match score will be 100% no matter the company name match. 

Having determined the percentage match score of each of the three data categories of the found possible match, the assigned percentages are used to determine the final percentage match score, which is built taking the following weights: 

The calculation for the final weighted percentage match score is the following: 

 The closer the final percentage match score is to 100%, the higher the probability of the possible match being a true match. 

Based on the final match score, GlobalPass also defines four match strength categories: 

In cases where both company number and country are matching, the match score will be 100% no matter the company name match.

In cases where not all three pieces of company data are provided, the skipped data piece(s) are not included in the final calculation, and the final score is divided by the sum of weights of the provided data. 

Example: only company name and country are provided:

Similarly, if a registration number is provided but no company identifiers are listed in the source, the registration number weight is not included in the final score calculation, as per example above. 

For further explanation of the weighted percentage match scoring system, the following hypothetical company with hypothetical matches will be used: 

Three name matches of 85% and higher strength and two matches purely for registration number are found, therefore they appear on the AML report and final match score calculations are run on them: 

Overall match = (100% × 0.45 + 100% × 0.45) / (0.45 + 0.45) = 100% 

2nd example – 100% overall match 

Overall match = 100% as both registration number and country are matching 

3rd example – 88% overall match 

Overall match = 96% × 0.45 + 0% × 0.1 + 100 × 0.45 = 88% 

Overall match = 89% × 0.45 + 100% × 0.1 + 0% × 0.45 = 50% 

Overall match = 0% × 0.45 + 100% × 0.1 + 0% × 0.45 = 10% 

Once an AML Screening of a Transaction is initiated, the entire database of watch-listed individual and business names is screened for name matches with 85% similarity and up, using fuzzy logic principles. This makes a transaction AML screening unique as it is not necessary to define entity type, and all three components of a transaction – sender, receiver, and financial institution – can be screened using the same parameters. 

Names of 85% and higher name match are listed in the GlobalPass AML reports as potential hits, and their final match score is based on their name match percentage. 

For more information about the screened database and fuzzy logic, please contact our support.