Running the solution

The Requisition Anomaly solution applies advanced, AI-driven analytics to continuously evaluate purchasing behavior at the requisition line level. By leveraging multi-temporal analysis, the solution examines both short-term and long-term trends in quantity and price across historical transactions, enabling it to establish expected baselines for normal purchasing activity.

The solution differentiates price evaluation based on item sourcing type.

  • Contracted items are evaluated for Contract Price Deviation.

    The system compares the requested unit price against the contracted unit price, any deviation beyond acceptable tolerance is flagged as Contract Price Deviation.

  • Non-contracted items are evaluated for Price Anomaly.

    The system evaluates the requested unit price through a weighted analysis of historical and recent pricing behavior to determine an expected range, flagging any deviations as a price anomaly.

Quantity anomaly is calculated through a weighted analysis of historical and recent purchasing behavior to determine an expected range, flagging any deviations as a quantity anomaly.

Pattern anomaly uses historical patterns to find deviations in ordering frequency and total quantity for the frequency.

How the solution works

  • Requisition Patterns Analysis: This feature analyses historical requisition data to identify patterns.
  • Anomaly Detection: This feature flags requisitions that significantly deviate from expected patterns and notify in context widget.
  • Notification Mechanism: This feature displays alerts through a widget to ensure prompt action.
Note: The Requisition Anomaly Context Widget displays relevant anomaly insights even when the requester is using RSS, ensuring consistent visibility of data across different request creation modes.

When a user clicks, selects, or creates a requisition line, the Requisition Anomaly Widget in the context viewer panel shows relevant information.

The Contracted or Non-Contracted Item indicator provides users with clear information on whether the selected item is part of an existing contract or not, helping them make more informed and compliant purchasing decisions.

Note: The Requisition Anomaly widget displays data for backdated requisitions up to the past year only.

Anomaly messages for Price section

This table shows the anomaly messages in the Price section.

Anomaly message Description
Contract Price Deviation Detected Requisition Price is above the contract price.
Contract Price Deviation Detected Requisition Price is below the contract price.
Price Anomaly Detected Current Price is above the expected range.
Price Anomaly Detected Current Price is below the expected range.
No anomalies detected Requisition is within the expected range.
No Contract Price Deviation Detected Requisition price matches the contract price.
No Price Anomaly Detected Current Price is within the expected range.

In the Price section, this information is displayed:

  • Requisition Price: Shows the unit cost listed on the Requisition Line.
  • Expected Range: Using the IQR method, AI provides an acceptable cost range that is based on historical data.

Anomaly messages for Quantity section

This table shows the anomaly messages in the Quantity section.

Anomaly message Description
Irregular Pattern Detected Ordering pattern is irregular, hence unable to fetch the expected range.
No Quantity Anomaly Detected Requested quantity is within the expected range.
Quantity Anomaly Detected Requested quantity is above the expected range.
Quantity Anomaly Detected Requested quantity is below the expected range.

Historical data is insufficient to predict the range or trend. This condition explains the distinction between an API error and insufficient historical data.

In the Quantity section, this information is displayed:

  • Requested Quantity: Shows the Quantity specified on the Requisition Line.
  • Total Requested Quantity: Shows the total sum of requested quantity for the calculated ordering Frequency.
  • Expected Range: Using the IQR method, AI estimates an acceptable range that is based on historical data.
  • Ordering Frequency: Using the IQR method, AI provides forecasts regarding how frequently an item is ordered.

An example of the Total Requested by Frequency is when a requester creates a requisition line and orders five units of an item. Behind the scenes, the system reviews all other orders placed on the same day for that item at the same requesting location and finds another requisition with six units. Instead of treating them separately, it combines both and shows a total requested quantity of 11 for that day. This aggregated view provides better visibility into ordering patterns, and the total requested quantity plays a critical role in identifying and flagging potential anomalies.