Skip to main content
To provide a safer environment for our merchants, we scan and intercept transfers that may lead to a potential loss of money because of fraudulent activities or security breaches. The scanned transfers are either blocked or put in review waiting for approval.

​
Proprietary machine learning model

Payouts has an anomaly detection machine learning model designed specifically to identify patterns that deviate from the expected behaviour and flags the initiated transfer as a potential fraudulent activity. Without any technical integration required from your end, you can easily detect and block individual fraud transactions in real-time.
Integration
Talk to your account manager or Fill out the Support Form to integration the machine learning model and find anomalies in your Payouts account.

​
How does it work?

  1. Learning normal
The model first studies a bunch of data points to understand what normal transfers looks like. It analyses the data itself to identify patterns and commonalities.
  1. Spotting the strange
Once the model learns what normal looks like, it keeps an eye on new data points. If the data is way off, it flags the particular transfer. Following are a few examples of anomalies:
  • Unusual transfer amount - The model detects any high value transfer happening on an account which deviates from normal behaviour of the transfer that took place in the past.
  • Spike in transfer frequency - The model detects any instance where the number of transfers to a specific bank account deviates significantly from the expected behaviour within the given time frame.
  • Temporal patterns - The model isn’t just interested in if a single day has high transactions, but also in how the transfer volume changes over time. By combining individual transfers, cumulative sums, and looking across different time windows, the model can uncover interesting patterns. For example, it might detect a sudden rise in transfers over 30 days, followed by an even steeper increase in hourly transfers within that period. This could signal unusual activity.
  1. Investigate the anomaly
Once the transfer is flagged for review, look into the details of the transfer and then you can approve or block it accordingly.
Whitelist a beneficiary
If you want the model to know a transfer is genuine for a beneficiary, you can whitelist the beneficiary in the dashboard.

​
Receive daily notifications

Once you integrate the model to your Payouts account, you will receive daily notifications regarding blocked/flagged transfers via email, SMS, Whatsapp, and you can also review it in the dashboard.