Anomalies

BTI uses artificial intelligence and machine learning to detect anomalies in your supply chain document transactions.

What data is used?

Only your company data is used to identify your anomalies. Historical data is analyzed to establish day-to-day document trends between you and your trading partners. This data is called the 'expected range of values'. You see an anomaly when your daily transaction totals lie outside of the expected range. No separate training data is used for this AI, and this analysis applies only to your actual data.

Data tracking begins when BTI is enabled for your company's account, so the relevance of your anomalies increases as more data, across days of the week, months, and seasons are gathered. For example, you might increase order volumes in certain months of the year. The first year these transactions are flowing in BTI, they are anomalies. When they recur in following years, Watson recognizes these transactions as a normal pattern in your data.

Why did the AI identify this data as an anomaly?

BTI evaluates your daily data against the expected range of values. An anomaly is identified when your daily transaction totals lie outside of the expected range. The trading partners that most contributed to the anomaly are provided. The most influential dates of the expected range are provided so that you can understand any seasonal adjustments that are considered. Anomalies are detected for the following values:
  • Order value
  • Invoice value
  • Payment value
  • Order count for document types 850, 860, and 875
  • Invoice count for document type 810
  • Shipment count for document type 856
  • Document volume
Remember:
  • Document count is the number of documents.
  • Document volume is the total file size.
  • All dates that are displayed on the Anomalies page are based on Coordinated Universal Time (Coordinated Universal Time).

View of the anomalies pane


Image of Anomalies tab
Click here for full size image
  • 1 View anomalies by clicking the Alerts icon in the header of the user interface.
  • 2 Filter results with the search bar. You can use keywords such as 'Orders', 'document volume', or a document number.
  • 3 Select an anomaly from the Anomalies panel to view it on the graph.
  • 4 The graph is a visual representation of the document trend. The timeline can display up to the most recent 90 days.
    • The green shaded area represents the expected range.
    • The blue line indicates where your data lies.
    • Blue circles represent anomalies.
    • Dashed lines indicate a time period during which no data is present.
  • 5 Click a circle on the graph to view more details.

How can I react to anomalies?

Anomalies can be opportunities. Depending on whether the anomaly is higher or lower than the expected range, you might want to look into the factors and plan accordingly to account for the anomaly. For example, on December 12, BTI calculates that you expect to receive between $600,000 and $800,000 in orders. But you received over $900,000. Your business might be doing better than expected! If you are the supplier, you might want to track behavior and ensure that you have enough inventory for future orders.

If you do not receive the usual volume of orders from one of your trading partners, a connectivity or system problem might be interrupting processing their data. Confirm that no outages or interruptions to be corrected.