Super.com replaced a growing manual-review bottleneck with a decision workflow that evaluates identity, payment, and behavioral signals together—without giving up transparency or operational control.
The customer
Super.com is a consumer financial services platform that offers rewards and credits across travel and everyday purchases to millions of users.
The problem
Super.com relies on third-party risk scores and performance indicators to determine whether an account is eligible to receive credits. While these models perform well in aggregate, there are false positives—particularly for legitimate users whose behavior resembles automation.
Because these scores act as a hard gating mechanism, false positives have real consequences: legitimate users are delayed from receiving credits, user trust suffers, and Super.com's operations team is forced to intervene more often.
As Super.com grew, the volume of flagged accounts increased steadily. What had been an occasional exception became a recurring operational challenge. The risk team recognized that a third-party score alone was no longer enough to make accurate decisions at scale.
Why additional context became necessary
The challenge wasn't that machine learning failed, but that opaque scores were asked to make nuanced decisions in isolation. The models lacked access to critical context, including:
- Identity verification outcomes
- Transactional and payment history
- Historical behavior and platform-specific signals unique to Super.com
Without this broader context, the system had limitations in consistently distinguishing malicious automation from legitimate activity.
The manual workaround
To mitigate false positives, Super.com's risk operations team began manually reviewing flagged accounts. Analysts pulled together identity verification results, transaction histories, and behavioral signals from multiple systems, then weighed positive and negative indicators to determine whether credits should be released.
While this approach improved accuracy, every review required careful investigation and manual interpretation. What was intended as an exception-handling workflow became a bottleneck—delaying legitimate users and consuming increasing amounts of operational capacity.
The solution with Socratix
Using Socratix, Super.com transformed its manual review process into an AI Operating Procedure: a human-defined, AI-executed decision workflow that evaluates identity, payment, and behavioral signals together.
Rather than relying on rigid rules or a single model output, Socratix applies explainable, context-aware decisioning that mirrors expert judgment while remaining fully auditable and controllable.
Super.com's risk team designed the workflow collaboratively so it reflected their institutional knowledge and decision logic. Socratix now executes that workflow consistently at scale, without manual intervention for routine cases.
Results with Socratix
Before Socratix, the team could manually review about five accounts per hour, with each decision requiring deep, cross-system investigation and subjective judgment.
accounts reviewed
total review time
account review scale
- Manual account reviews are removed from routine cases
- False positives are reduced so legitimate users are not blocked
- Edge cases remain transparent, auditable, and under team control
What once required sustained manual intervention now runs automatically—improving user trust and operational confidence without sacrificing decision quality.