Anomaly Detection
Table of Contents
Use Case:
Organizations need to ensure that customer revenue transactions align with company policies and flag any unusual activity that could indicate fraud, policy violations, or errors. Transactions in categories such as invoices, credits, or payments can occasionally show unexpected variations, warranting closer attention.
The Anomaly Detection risk analytic identifies transactions that deviate from typical patterns using either Z-Score analysis, which measures how far a transaction’s value is from the average of similar transactions, or machine learning-based Clustering, which identifies transactions that do not fit within established groups of similar activity. The analysis groups transactions based on both transaction method or source system (e.g., ERP, billing platform, or order management system) and transaction type (e.g., invoice, credit, payment), establishing an expected range or "norm" for each group. Transactions that exceed two standard deviations from this average are flagged for review, helping organizations detect potential issues and ensure policy compliance.
| Description | Identification of anomalous transactions according to the transaction's value relative to a historical period of related transactions (1 year). Related transactions are those of the same transaction source or channel (e.g., ERP, billing platform, order management system) and the item's nature (transaction type, e.g., invoice, credit, payment). |
| Domain(s) | Customer Revenue |
| Analysis Type | Context Flag |
| Focus Area | Value |
| Score Methodology |
This depends on the algorithm selected for the rule:
|
Default Scoring Criteria
Importance: 5 (default)
Enabled: True (default)
| Risk Result | Default Value | Notation |
|---|---|---|
| Weak | Rules are configured within the application | |
| Moderate | ||
| Strong | ||
| Auto | ||
Rule Configuration
To configure a rule for this risk analysis:
1. From the Admin page, navigate to Risk Settings > Risk Analysis Settings > Anomaly Detection (Customer Domain). Then, click Add A Rule.
2. Complete the following required rule fields:
| Rule Field | Description | Settings Options |
|---|---|---|
| Name | Name for the specific rule | Free text |
| Description | Description of the rule | Free text |
| Algorithm | The desired algorithm for scoring transactions | Clustering or Z-Score |
| Context Period in Day | The period of time from the spend date for which transactions are considered by the rule | 1 to 365 Days |
3. Click Add A Rule Component.
4. To set a component, select a Field and Search Type from the dropdowns.
The Value field will appear after you select a Field and Search Type.
See a description of Field, Search Type and Value below:
| Field | Search Type | Value Type |
|---|---|---|
| Company | Includes OR Excludes | List is populated from customer data* |
| Customer (Payer) | Includes OR Excludes | List is populated from customer data* |
| Customer Billing Line Item Amount USD |
Greater Than or Equal To OR Less Than or Equal To |
Dollar Amount |
| Customer Billing Line Cost USD |
Greater Than or Equal To OR Less Than or Equal To |
Dollar Amount |
| Customer Billing Line Item Text | Contains OR Does Not Contain | Free text |
| Has Free of Charge Item | Includes OR Excludes | Yes/No |
| Is Return | Includes OR Excludes | Yes/No |
| Material Type | Includes OR Excludes | List is populated from customer data* |
| Organization Unit | Includes OR Excludes | List is populated from customer data* |
| Original Transaction Amount USD |
Greater Than or Equal To OR Less Than or Equal To |
Dollar Amount |
| Subject (Payer Country) | Includes OR Excludes | Preset List |
| Subject Account Group | Includes OR Excludes | List is populated from customer data* |
| Subject Classification | Includes OR Excludes | List is populated from customer data* |
| Transaction Reason | Includes OR Excludes | List is populated from customer data* |
| Transaction Type | Includes OR Excludes | Preset List |
| Transaction Type (Detailed) | Includes OR Excludes | List is populated from customer data* |
*For values populated from customer data, value lists will not populate if data is not available.
Note: Where multiple rule components are set, they operate only in an “and” manner.
5. Click the Save Changes button at the bottom of the page to save your rule.
Exclusions
None
Unique Configuration
This analysis utilizes rules configured by the user, for which they can configure the following:
- The historical period for anomaly context up to 365 days
- Configuration of cohort to detect anomalies
- Default dimensions configuration: Payment Method, Nature Detailed
- Scoring Criteria based on the transaction value (Strong, Moderate, and Weak values can be set)
- Automatic Criteria - Transactions that meet these criteria will appear in Monitoring Review results regardless of the Risk Engine Score
Available dimensions:
- Transaction Type (e.g., Sale, Receivable Increase, Receivable Decrease, Write-Off)
- Transaction Reason (e.g., Goods, Services, Financial, Return, Correction, Giveaway/Donation, Customer Issues (Quality, Warranty, etc.), Rebate, Other)
- All values are log-transformed, which allows us to mitigate the undesirable effects of a “long tail” distribution. This enhances the occurrence of our anomaly detection.
Exclusions
- Items less than $40 cannot receive a risk result