9.4 Scoring Samples

This section provides the Scoring samples.

Event

This scoring rule defines various scoring criteria to be followed focusing on the event attributes. The Event Scoring is performed on the following event attributes:
  • Scenario
  • Total Transaction Amount and Risk Score
  • Aging
Scenario
  • Provide default scoring for each scenario. The total of events scored contributes to the pre-case score. The following are the default score for different scenarios:
    • ML –10
    • Fraud –5
    • Transaction/Sanctions Filtering –30
    • KYC –20
  • If a correlation is formed for three events (A, B, and C) by ML, TF, and KYC. The following is the pre-case score for correlation.
    • Event A – ML (Rapid Movement of Funds – All Activity (CU focus)) –10
    • Event B – TF –30
    • Event C – KYC – 30
    • Total pre-case score –70.
  • If a correlation is formed for 3 events (A, B, and C) all ML scenarios. The following is the pre-case score for correlation.
    • Event A – ML (Rapid Movement of Funds – All Activity (CU focus)) –10
    • Event B – ML (CIB - Previous Average Activity (AC focus)) – 10
    • Event C – ML (HR Trans – Focal HRE (CU focus)) – 10
    • Total pre-case score –30.
Total Transaction Amount and RiskScore

In this attribute, each event is scored. The total of the events scored contributes to the pre-case score.

  • When event has total transaction amount >= <Configurable amount> and risk score >= <configurable risk score>, give X score to event. Risk scores for amounts can be segregated into 3 buckets. For dollar amounts transactions between 50K and 100K should be given a score of 20, 100K to 500K should be given as 30 and anything above 500K should be 50.
  • Correlation is created for 2 events A and B by an ML and TF. Transaction amounts between 0 and 50000.99 get 10 points; Trxn amounts between 50001 and 100000 get 20 points; Trxn amounts > 100000 get 30 points. The Pre-case score should be calculated as below:
    • Event A – (Total amount of transactions - $ 80K) - 20
    • Event B – (Total transaction amount - $ 300K) - 30
    • Total pre-case score is 50 (A(20) + B(30) =50)
Aging

Scores of the events in the correlation are decreased if the correlation is not consolidated to a case after some time. After a certain duration event is completely dropped from the correlation and shall be archived. The score reduction is configurable by country, jurisdiction, scenario, and time period.

In this attribute, each event is scored. The total of the events scored contributes to the pre-case score.

The following is the scaling for aging events that are members of un-promoted correlations. Age scaling must be configurable and can be changed from the following sample:

  • Scenario Rapid Movement of Funds All Activity (all focal types) - When an event age reaches 3 months reduce the event score by 3
  • Scenario Rapid Movement of Funds All Activity (all focal types) - When an event age reaches 6 months reduce the event score by another 3
  • Scenario Rapid Movement of Funds All Activity (all focal types) - When an event age reaches 9 months reduce the event score by another 3
  • Scenario Rapid Movement of Funds All Activity (all focal types) - When an event age reaches 12 reduce the event score to equal 0
  • Drop and archive any event of correlation age for more than a year.

    Note:

    You need to determine the process that would remove the event with a score of 0 from the correlation and close it with a specific reason.

Correlation is created for event A by (ML) Rapid Movement of Funds All Activity CU.

  • The correlation creation date is 1st Jan 2016 and Event A with event creation date 1st Jan 2016 has an initial score of 10. So the pre-case score is 10.
  • On 1st of February event B by (ML) Rapid Movement of Funds, All Activity CU with creation date 1st February 2016 is added to correlation. Event B score is 10 and the total pre-case score now is 20. A(10)+ B(10) =20
  • On 1st April, event A age is now 3 months. Event A score will be reduced by 3 points to 7 andthe total pre-case score is now 17. A(7) + B(10) = 17
  • On1st May, event B age is now 3 months. Event B score will be reduced by 3 points to 7 and the total pre-case score is now 14. A(7) + B(7) = 14
  • On1st July, event A age is now 6 months. Event A score will be reduced by 3 points to 4 and now the total pre-case score will be 11. A(4) + B(7) = 11
  • On1st Aug, event B age is now 6 months. Event B score will be reduced by 3 points to 4 and now the total pre-case score will be 8. A(4) +B(4) = 8.
  • On1st Oct, event A age is now 9 months. Event A score will be reduced by 3 points to 1 and now the total pre-case score will be 5. A(1) + B(4) = 5
  • On 1st Nov, event B age is now 9 months. Event B score will be reduced by 3 points to 1 and now the total pre-case score will be 2. A(1) +B(1) = 2.
  • On the 2nd Jan 2017, event A age is now 12 months. The Score will be dropped to 0. And Event A will be closed and completely dropped from correlation. Event B is the only event in correlation and the total pre-case score will be now 1.
  • On 2nd Feb 2017, event B age is now 12 months. The Score will be dropped to 0. And Event B will be closed and completely dropped from correlation.

Entity

This scoring rule defines various scoring criteria to be followed focusing on the entity attributes. The Entity scoring is performed on the following entity attributes:
  • Watch List Screening
  • Effective Risk
Watch List Screening

If the correlated entity is matched against screening specified watchlist, give the distinct customer a score. The total of the customer score contributes to the pre-case score.

For example: Entity A (10 for ML event) and B (10 for ML event) are part of the correlation. The total pre-case score is 20. After some time Event C is added to the correlation. Event C involves entity C and entity C is matched to a specific WL (configurable). Matches to that WL receive a score of 60. The Event score for Event C is 10 for the ML event. The correlation also now has an entity score of 60 for Entity C.

Pre-case score = A(10) + B(10) + C(10) + Entity C (60) = 90

EffectiveRisk

If the correlated entity, effective risk >= Y then increase customer score. The scale should be configurable by effective risk and jurisdiction.

The total customer score contributes to the pre-case. For example:

  • Set up the rule to find the KDD CORR_LINK.BUS_NTITY_KEY_ID and KDD CORR_LINK.BUS_NTITY_ID for an in the correlation. Look at the respective business table (based on the BUS_NTITY_ID type) to find the Effective Risk.
  • Event A Rapid Movement of Funds All Activity CU focus – scenario score of 10; Customer XXX has CUST. CUST_EFCTV_RISK_NB = 8
  • Event B Rapid Movement of Funds All Activity CU focus - scenario score of 10; Same customer XXX has CUST. CUST_EFCTV_RISK_NB = 8
    • Customer Effective Risk >= 7 add 10 points
    • Pre-case score = A(10) + B(10) + Cust XXX(10) = 30. Dev Note – this is on distinct customer in correlation

Correlation

This scoring rule defines various scoring criteria to be followed while creating an entire correlation. The score generated by correlation scoring contributes to the pre-case score.

This is performed on the following criteria:
  • Number of events
  • Combination of Scenarios
  • Total Transaction Amount
  • Repeated Scenario Events
Number of events
  • If the number of events in the correlation is more than X, increase the correlation score.
  • Scaling of correlation by the number of events should be as below (scaling should be configurable by no. of events):
    • Numberof events greater than 3 and less than or equal to 5 should be given a correlation score of 30.
    • The number of events between 6 and less than or equal to 10 will be given 40.
    • Correlation with more than 10 events will be given 50.
  • The additional score has to be added to the pre-case score.
  • For example:
  • A correlation has 4 events A, B, C, and D by ML. Event scores for 4 events are as follow:
    • A –10
    • B –20
    • C –10
    • D –30

The pre-case score will be now 70 but an additional 30 correlation score will be added to the pre-case score as the number of events in the correlation is 4. And correlation is promoted to the case.

Combination of Scenarios
  • When correlation contains events from scenario X and Scenario Y at the same time consider correlation to add a score.
  • The total of the correlation score contributes to the pre-case score.
  • For example:
  • Event A Rapid Movement of Funds All Activity CU focus and Event B Deposit Withdrawal Same or Similar Amount AC focus are correlated in the same correlation add 50 points
    • Event A –10
    • Event B –10
    • Correlation –50
    • Pre-case score =70
Total Transaction Amount
  • If the total amount of transaction of the correlated events is greater than X amount, consider adding a score to correlation. Risk scores for amounts can be segregated into 3 buckets (configurable). For dollar amounts, the total of transactions across all correlated events is between 50K and 100K should give a score of 20, 100K to 500K should be given as 30 and anything above 500Kshould be 50. The transaction amount should be based on the matched binding for the total transaction amount (configurable to use a functional currency total transaction amount is scenario configured for it).
  • The total correlation score contributes to the pre-case score.
  • For example:
    • Event A ML scenario – total base transaction amount =15000
    • Event B ML scenario – total base transaction amount =40000
    • Event C ML scenario – total base transaction amount =45000
    • Total correlation transaction amount =100000
    • Scoreis A(10 for ML) + B(10 for ML) + C(10 for ML) + Correlation(30) = 60 for pre-case score
Repeated Scenario Events
  • Increase the score of the correlation if events are generated for the same customer/entity within a configurable time period.
  • Scaling for correlation by repeated scenario events should be as below:
    • Increase score by 30 if 2 events are created for the same entity/same scenario within look back period. The number of events and lookback are configurable.
    • Increase score by 50 if 3 or more events are created for the same entity/same scenario within look back period. The number of events and lookback are configurable.

For example:

  • Assume customer CU1 had an event A on Rapid Movement of Funds (RMF) on 1st July 2016 and which had a score of 50 to start with.
  • On 28th July 2016, the customer had another RMF event B with an Event score of 30. But since this a repeat event for the same scenario on the customer within a (Repeated scenario event lookback) 31 days, the correlation score could be increased by say 20 points. So overall the pre-case would tip over to 100 which is the score required to convert the pre-case to the case.
  • The total correlation score contributes to the pre-case score.