Scoring Engine overview
Updated on 21.02.23
6 minutes to read
The Scoring Engine is SEON's brain. It's where you can find rules, machine learning tools, and automated decisioning. While you may feel like Cinderella when fighting fraud, the Scoring Engine is a flock of white doves, ready to help you separate lentils from ashes or good customers from bad.
What are rules?
When SEON analyzes a transaction, we enrich all the data points in your API request. We then take all that new information and assign each transaction a fraud score or state.
Rules underpin this scoring and categorization. Some rules will increase or decrease a transaction's fraud score based on specific data points. Others can compare various data points. Let's start with a few simple examples:
- "No online profiles were found" +7 Fraud Score
This simple rule asks whether the phone number provided in a transaction is registered on any social or digital platforms SEON checks. If not, the number is suspicious, so the rule adds 7 points to the transaction's fraud score.
- "Email is not deliverable" +4 Fraud Score
This rule asks whether the email address in the transaction can receive emails. Because a technical glitch may cause a negative result, we only add 4 points to the fraud score.
In SEON, these categories are called states and indicate what action SEON took with a transaction. There are three states: APPROVE, REVIEW, and DECLINE. Rules can also change a transaction's state:
- "User has disabled cookies" DECLINE
The customer trying to access your site or complete a purchase has turned off all cookies in their browser. Fraudsters often do this in an attempt to mask their identities. But beware, drastic rules like this one can also block privacy-conscious users.
SEON divides rules into three categories depending on which SEON solution they are connected to.
These are pre-added rules created by our team of fraud prevention experts based on the breadth of industry knowledge. You can turn them on or off and change how they affect transaction fraud scores. However, you can't delete them or add new default rules. Head to custom rules to set up lookalike rules.
Custom rules are your team's creative fraud-fighting powerhouse. Your team can get stuck in creating rules themselves or work with our stellar Customer Success team to create some.
Machine Learning Rules
Our Whitebox machine learning algorithm will recommend rules specific to your account. You can turn these rules on or off, but you can't create new ones manually. However, our algorithms are constantly generating new ways of keeping your business safe from fraudsters.
Rules can have one of three actions. A rule can either change a transaction's score, set the transaction's state, or add the content of any field to the blacklist/whitelist or any custom list.
Rule hierarchy will also affect the fraud score. If a transaction only triggers scoring rules, then the fraud score will be the sum of their values. In this case, the transaction's state is determined by the thresholds set in Settings.
State rules will set the state of a transaction to APPROVE, REVIEW or DECLINE regardless of its fraud score. The Applied Rules widget will still show you all rules triggered by the transaction, but the fraud score will jump to your organization's threshold for the given state.
You may add rules that do not modify the fraud score (set + 0) but trigger. Use these rules to observe behavior before committing to a decision or as a warning note on transactions.