Skip to main content
This guide outlines how we determine which segments of your data to monitor for anomalies, and what constitutes an anomalous change in a given metric for a monitored segment.

​
Segment Selection Criteria

Pagos doesn’t automatically monitor every segment of your business, only those that meet our defined criteria. To ensure your alerts are based on reliable data, we only automatically monitor segments that meet the following thresholds:
  • Has a history of 45 days of transaction data
  • Has a minimum of 100 transactions per day on at least 75% of those days
  • Accounts for at least 5% of your total transaction volume
To ensure we’re always keeping up with ongoing changes in your own operations, we re-evaluate which of your business segments we automatically monitor on a weekly basis.

​
Alert Criteria

Pagos’ Alerts surfaces unexpected changes in key payment metrics, such as sudden drops in approval rate or unusual spikes in chargebacks, so customers can quickly investigate issues. To ensure alerts are meaningful and not overly sensitive, we combine statistical models with strict volume safeguards when reviewing each monitored metric.

​
Attempted Transaction Count

An Attempted Transaction Count alert indicates the monitored segment’s daily transaction count dropped significantly. An alert is triggered for a given segment if all the following are true:
  • Today’s transaction count is at least one standard deviation below the minimum transaction count seen in the last 90 days AND
  • Today’s transaction count is less than 60% of the lowest transaction count seen in the last 90 days
When reviewing Attempted Transaction Count for anomalies, we review data from the last 90 days to capture and account for seasonal patterns.

​
Approval Rate

An Approval Rate alert indicates the monitored segment’s daily approval rate dropped significantly more than your typical daily swings. An alert is triggered for a given segment if all the following are true:
  • Today’s approval rate is at least two standard deviations below the minimum approval rate seen in the last 45 days AND
  • The approval rate on a given day is less than 75% of the lowest approval rate seen in the last 90 days
When reviewing Approval Rate for anomalies, we review data from two different lookback periods—45 and 90 days—to balance recent data with more stable trends over time.

​
Decline Rate

A Decline Rate alert indicates a specific decline code (e.g. insufficient_funds) is spiking for a given monitored segment. By monitoring decline rate per decline code, Pagos gives you more insight into where you may have an issue in your current setup. The segment selection criteria for decline rate is the same as all other metrics with one additional caveat: the number of transactions declined with the specific decline code is 1% of total attempted transactions for that segment. An alert is triggered for a given segment if all the following are true:
  • Today’s decline rate is at least 1.5 standard deviations higher than the highest decline rate seen in the last 31 days AND
  • Today’s decline rate is more than 150% of the highest decline rate seen in the last 31 days AND
  • There were at least 25 declines today with the given decline code
When reviewing Decline Rate for anomalies, we review data from the last 31 days to match the monthly billing cycles of most credit cards.

​
Chargeback Count

A Chargeback Count alert identifies an unusually high volume of chargebacks, indicating you may be facing a fraud attack or operational issue. An alert is triggered for a given segment if all the following are true:
  • Today’s chargeback count is at least two standard deviations above the mean chargeback count from the last 31 days AND
  • Today’s chargeback count is more than three times higher than the highest chargeback count seen in the last 31 days AND
  • You saw at least 20 chargebacks today
When reviewing Chargeback Count for anomalies, we review data from the last 31 days to match the monthly billing cycles of most credit cards.

​
Refund Count

A Refund Count alert indicates an unusually high number of refunds. An alert is triggered for a given segment if all the following are true:
  • Today’s refund count is at least two standard deviations above the mean refund count from the last 31 days AND
  • Today’s refund count is is more than three times higher than the highest chargeback count seen in the last 31 days AND
  • You saw at least 20 refunds today
When reviewing Refund Count for anomalies, we review data from the last 31 days to match the monthly billing cycles of most credit cards.

​
Historical Data Requirements

Before Pagos can start sending you alerts, we need to ingest at least seven days worth of your transaction data. Alerts will start working as soon as you have a week of data. Pagos anomaly detection will improve with time, as additional historical context helps us calculate your baseline and identify expected trends; more data ensures a more accurate detection of truly unusual patterns. Our optimal lookback window is 4-13 weeks (31-90 days), depending on the alert metric.

​
Why Didn’t I Get an Alert?

Common reasons you might see concerning numbers without receiving an alert:

Not enough volume

You need to meet the following thresholds for each metric:- Approval Rate - 100 transactions- Chargeback Count - 20 chargebacks- Refund Count - 20 refunds- Decline Rate - 25 declines with the specific decline code

Only some checks passed

All conditions must be met. A lower than usual approval rate might not be low enough compared to your worst recent days

You’ve already been alerted

Pagos doesn’t repeat alerts for ongoing issues. Check your previous alerts.

Not enough history yet

You need at least seven days of ingested payments data before we can trigger alerts. To ensure optimal accuracy, Pagos requires 2-13 weeks of data depending on the alert type

This is your normal pattern

Pagos Alerts learns expected trends and patterns for different days or times (e.g. transaction volume on Sundays is naturally lower)

Not enough data in this segment

The segment was not large enough to be monitored by Pagos Alerts. Create a custom segment to define the exact market segment you want us to monitor for anomalies.