How Customer Classifier Works
Customer Classifier is a machine learning model trained on real betting behaviour: staking patterns, selection choices, timing, and outcomes. It continuously scores every account and updates its classification as behaviour evolves, giving your risk team a live picture of who they're really dealing with.
Behavioural Signal Engine
Every bet, stake adjustment, and market choice feeds a growing behavioural profile for each customer, capturing patterns that manual review could never keep pace with.
Real-Time Classification
Customers are reclassified continuously as new bets arrive, so a change in behaviour, sharpening, staking up, or shifting markets, is reflected immediately, not at the next manual review cycle.
Confidence-Scored Segments
Every classification ships with a confidence score, so your team knows exactly how certain the model is and can prioritise manual review where it matters most.
Citadel Integration
Classifications feed directly into Citadel's Neural Risk Engine, giving your automated risk decisions the customer context they need without any manual handoff.
Why Customer Classifier
Smarter Risk from Day One
New accounts are classified from their very first bets, so risk decisions are informed from the outset, not after weeks of manual observation.
Fully Automated
No manual tagging or spreadsheet reviews, classifications update themselves as behaviour changes.
Built for Citadel
Designed to plug straight into Citadel for a fully automated, end-to-end risk assessment pipeline.
Protects Your Margin
Earlier, more accurate classification means exposure is managed before it becomes a problem, not after.
Smarter Risk From Day One
Most classification systems only become useful after weeks of accumulated history on an account. Customer Classifier starts scoring behaviour from the very first bet, using early signals, selection type, staking pattern, timing, to produce an initial classification that sharpens as more data arrives.
This means your risk team is never operating blind on new accounts. High-risk behaviour is flagged early, and low-risk customers are identified quickly too, so your onboarding experience isn’t unnecessarily cautious for the customers who don’t need it.
Fully Automated Classification
Manual customer review doesn’t scale, it relies on individual traders noticing patterns, and it lags real behaviour by days or weeks. Customer Classifier replaces this with continuous, automated scoring that never falls behind.
As a customer’s behaviour shifts, staking up, changing markets, adjusting timing, their classification updates automatically. There’s no queue of accounts waiting for manual review, and no inconsistency between how different staff members classify similar behaviour.
Built to Work With Citadel
Customer Classifier is designed from the ground up to integrate with Citadel’s Neural Risk Engine. Classifications flow directly into Citadel’s decision-making process, giving every automated risk decision the customer context it needs.
The result is a fully automated risk assessment pipeline: behaviour is observed, classified, and acted on without a single manual handoff between systems or teams.
Protecting Your Margin
Every day a high-risk customer goes unclassified is a day of unmanaged exposure. Customer Classifier’s speed and accuracy directly translate into earlier, better-informed risk decisions.
By catching behavioural shifts as they happen rather than after the fact, your team can act while exposure is still manageable, protecting margin without needing to blanket-restrict customers who aren’t actually a risk.