Replenishment risk score calculation

The replenishment risk score evaluates and categorizes product replenishment risk levels based on five factors, calculated per warehouse:

  • Line Hits score: Represents the number of order lines that included the product during the past 12 months and is compared with the busiest product in the same warehouse.
  • Usage Quantity score: Represents the total quantity of the products sold or used during the past 12 months and is compared with the maximum in the warehouse.
  • Average Inventory Value score: Represents the average on-hand inventory dollar value of the product over the past 12 months and is compared with the maximum in the warehouse.
  • COGS score: Represents the total cost of goods sold for the product during the past 12 months and is compared with the maximum in the warehouse.
  • Unique Customer score: Represents the number of distinct customers who purchased the product during the past 12 months and is compared with the maximum in the warehouse.

Each factor score is calculated using the formula: (Product value / Warehouse maximum for the factor) × 160. This ensures that each factor receives a score between 0 and 160. The Weighted Score is then calculated by multiplying each factor score by the assigned weightage and adding the results. The final Weighted Score ranges from 0 to 160, where a higher score indicates greater replenishment urgency.

The replenishment risk score is the weighted sum of these five factors, with a maximum of 160:

Factor Weightage Duration Score range
Line Hits score 50% 12 Months 0-160
Usage Quantity score 20% 12 Months 0-160
Average Inventory Value score 10% 12 Months 0-160
COGS score 10% 12 Months 0-160
Unique Customer score 10% 12 Months 0-160

The replenishment risk levels are categorized into Low, Medium, High, or Critical. The categorization can be performed using either fixed score cutoffs or the K-Means Machine Learning algorithm. With fixed cutoffs, the final replenishment score is classified as follows:

  • Critical: 601-800
  • High: 401-600
  • Medium: 201-400
  • Low: 0-200

The final replenishment risk score is calculated as follows:

  • Step 1 - Normalize each factor score:

    The score of each factor is normalized within the warehouse to a range of 0-160, using the formula:

    Metric Score = (Factor Value/Warehouse Maximum for the Factor) * 160

    The score is calculated relative to the warehouse, and each product is measured against the product with the highest value for the same factor in the warehouse.

  • Step 2 - Calculate the weighted score:

    The weighted score is calculated by applying the defined weightage to each factor, using the formula:

    Weighted Score = (Line Hits * 0.50) + (Usage Quantity * 0.20) + (Average Inventory Value * 0.10) + (COGS*0.10) + (Unique Customers * 0.10)

    The weighted score ranges from 0 to 160.

  • Step 3 - Calculate the final replenishment score:

    The final replenishment score is calculated, using the formula:

    Final Replenishment Risk Score = Round (Weighted Score * 5, 2)

    The final replenishment risk score ranges from 0 to 800, where a higher score indicates greater replenishment urgency.

The K-Means approach clusters replenishment risk based on the normalized factor scores and ranks the resulting clusters according to the average scores. The highest-scoring cluster is classified as Critical, while the lowest-scoring cluster is classified as Low, with the remaining clusters categorized as High and Medium. This approach helps reduce variation within each risk group.

A demand-aware overlay is then applied to the calculated risk level to prevent products with little or no genuine demand from receiving an artificially high-risk rating:

  • Demand gate: Products with zero COGS and zero unique customers, or products with near-zero COGS and fewer than two customers based on the applicable tenant threshold, are forced to Low risk regardless of the raw score.
  • Activity floor damping: The score is adjusted using min(1, Line Hits ÷ 12). This prevents a product from receiving a maximum score simply because of being the only active product in a warehouse with very little overall activity.
  • Concentration cap: Products with fewer than two unique customers cannot be classified as Critical and are capped at High.
  • Frozen band cutoffs: Risk-level boundaries are calibrated against reference data and remain fixed between runs, ensuring that the meaning of each risk band remains consistent.

These guardrails address situations where a product could otherwise receive a high raw score despite having negligible real demand. For example, a product with only three-line hits, zero COGS, and zero customers could receive a score of 128 and be classified as Critical simply because of being the most active product in an otherwise inactive warehouse. The demand-aware overlay prevents such cases from being incorrectly classified as Critical.