Scenario: Monitoring node performance

IBM® Sterling Intelligent Promising helps the fulfillment managers to monitor node performance with the help of artificial intelligence (AI).

Fulfillment managers can monitor node performance with the help of an AI-powered node risk model. For more information about the node performance at a node, see Node performance.

The node risk model first predicts the probabilities of order cancellation and reschedules. The predicted probabilities are then converted into cancellation cost and reschedule cost by using the AI-powered node risk model. The node risk model depends on the historical data and inventory levels for predicting cancellation and reschedules probabilities. For more information about the node risk model, see Node risk model.

The node risk model predicts the node cancellation and node reschedule probabilities for multiple combinations of SKUs, nodes, and inventory levels. The node risk model also considers few other parameters while it calculates cancellation and reschedule costs based on which cost it is calculating. You can understand the cancellation cost and reschedule cost calculations with the help of following examples.

Example 1: Calculating order cancellation cost at a node

In a scenario where a node 1 can operate at various inventory levels, the inventory levels are fetched from the Optimizer API. The AI-powered node risk model considers the product margin and inventory levels to calculate probability of cancellation and cost. The product margin is calculated as follows:

Product margin = Store price - Item cost

Assume that in this scenario, the product margin is predefined as 5 USD.

Note: The AI-powered node risk model calculates the probability of cancellation. The cancellation cost is calculated by using the following formula:

Cancellation cost = Probability of cancellation × Product margin

Inventory level	Probability of cancellation (%)	Cancellation cost (USD)
Low	25	1.25
High	0	0

Example 2: Calculating order reschedule cost at a node

In a scenario where a node 2 can operate at various inventory levels, the inventory levels are fetched from the Optimizer API. The extra packaging probability, extra packaging cost, package delay probability, and package delay cost are predicted by the AI model. The AI-powered node risk model considers the predefined factors along with the inventory levels to calculate probability of reschedule and cost.

Assume that the AI-predicted values are as follows:

Extra packaging probability = 0.049
Extra packaging cost = 8.5
Package delay probability = 0.15
Package delay cost = 3.5

Note: The AI-powered node risk model calculates the probability of reschedule. The reschedule cost is calculated by using the following formula:

Reschedule cost = Probability of reschedule × [(Extra packing probability × Extra packing cost) + (Delay probability × Delay cost)]

Inventory level	Percentage reschedule possibility	Reschedule cost in USD
Low	20	0.1883
High	1	0.009415

The cancellation cost and reschedule cost are then weighted according to the optimization profile. Sterling Intelligent Promising uses the weighted costs for the selection of winning node.

Note: The fulfillment manager must consider external events when they monitor node performance. For example, the fulfillment manager must consider seasonality, spikes in demand for certain types of items, back to school wear, weather, promotions, and the impact of social media influencers.