Requisition pattern analysis and anomaly detection

This use case uses AI to improve the challenge of stock replenishment. You can perform anomaly analysis on requisition data by evaluating both quantity and frequency based on historical trends. This feature provides insights into deviations that may indicate anomalies, supporting more informed decision-making.
  • Requisition patterns analysis: This feature analyzes historical requisition data to identify patterns in quantity, timing (for example, weekly, monthly, quarterly), and item pricing
  • Anomaly detection: The system flags requisitions that significantly deviate from expected quantity and cost. There are two types of anomalies: cost anomalies and quantity anomalies.

Cost anomaly detection calculation

To calculate for cost anomaly detection: normal range = average ± 1.5 × standard deviation

For each item in the data, the system calculates:
  • The average cost of the item.
  • The standard deviation, indicating how much the cost typically varies.
  • A range of acceptable values based on the average and standard deviation. For example, lower and upper bounds are determined.
  • Any cost that falls outside this range is treated as an anomaly.

Quantity anomaly detection calculation

The process analyzes historical order data and transforms it into Item–Requesting Location pairs. Based on this data, the system uses AI to calculate the lower and upper bounds for normal quantity ranges.

The normal range is defined as: normal range = average ± 1.5 × standard deviation

Depending on the ordered quantity, the widget displays one of these widgets:
  • Quantity is within the normal range
  • Quantity is below the normal range
  • Quantity is above the normal range

Pattern detection calculation

This process analyzes historical order data to automatically determine the typical ordering frequency of each item at each requesting location such as daily, weekly, monthly, quarterly, or irregular. By grouping data by item and location, it assesses the consistency of order quantities across various time intervals and identifies the pattern with the least variation. If no consistent pattern is found, the item is classified as irregular. Anomalies are displayed through a widget using a notification mechanism to ensure prompt action.

When a requisition line is selected, the Anomaly Exists widget in the context viewer panel displays details these information:
  • Frequency: Daily, weekly, or monthly, quarterly, or yearly depending on the AI-analyzed pattern.
  • Messages to the user:
    • Quantity is within the normal range
    • Quantity is below the normal range
    • Quantity is above the normal range
    • Irregular pattern if no pattern is found for ordering
  • Expected cost: Displays the expected cost for the requisition line.
  • Ordered: Shows the quantity specified on the requisition line.
Note: The widget displays the base unit of measure (UOM) for an item. If the item has mixed UOMs, for example, BOX, CASE, and PALLET, any anomaly is shown in its base UOM, which is defined for that item.