Scenario: Node balancing cost optimization

The Optimization service provides you with the capability to optimize your orders to consider node balancing. You can prioritize node balancing over shipping costs when, for example, you want to satisfy high demand during peak shopping times.

E-commerce is in higher demand each year, especially during peak shopping seasons. During the holidays last year, your business had a high backlog in its stores because not all fulfillment dimensions were considered when an order was allocated to a specific store.

While e-commerce business grows, the Optimization service optimizes orders by considering all nodes (stores and distribution centers) for fulfillment. Especially during peak times, you might want to prioritize network balancing and customer SLA over shipping costs. The Optimization service helps optimize orders to find the lowest cost to serve while still meeting customer expectations.

By using existing nodes and distribution centers to optimize order fulfillment, your business can meet the high e-commerce demand during peak shopping times. You can set business priorities to avoid backlog, and the Optimization service optimizes each order to fulfill the order while avoiding backlog.

Capacity cost calculation is function of CapacityConsumed,

Node
Balancing Values

, Daily Capacity (Processing Plan), and

quantity
(input of Optimizer Call)

being optimized at ship nodes.

For more information, see the following definitions:

Let us suppose the configuration for Store1 and Store2 as follows:

Node ID	Backlog days	Tie breaker cost	Overcapacity penalty	Capacity (Processing plan)
Store1	0.9	0.2 USD	3.5 USD	30 units per day
Store2	1.1	0.2 USD	3 USD	40 units per day

Let us consider the other costs such as shipping and processing are same for Store1 and store2; this means that the capacity cost is driving the Optimization decisions.

When threshold is not reached

The balancing of demand takes place based on tie-breaker cost. Please refer the following table for calculation.

The orders are allocated to Store1, then Store2, and again to Store1 until the threshold is reached. The Store1 and Store2 are of similar sizes and have the same tie-breaker cost. Suppose the same instance at network level where the demand is balanced across network.

Note: The balancing is controlled by node balancing calendar configuration. The threshold for regular days can be less than one day and you can set to 2, 3, or even higher for holidays and peak season to see the same effect of demand balancing across network.

Time series	Store1	Store2
T(1)	1 qty	1 qty
T(2)	1 qty	2 qty
T(3)	2 qty	2 qty
T(4)	2 qty	3 qty
T(N)	Reached threshold first as its daily capacity is lower	Might be still below threshold

T(N) is greater than T(2) and T(2) is greater than T(1).
At time T, any order has 2 assignments for a given order line.

Time series	Store1	Store2	Optimization result	Comments
T(1)	`CapacityConsumed` = 1 CCU = 0.03333 `CapacityCost` = 0.000222	`CapacityConsumed` = 1 CCU = 0.025 `CapacityCost` = 0.000125	Store2	Only tiebreaker cost is contributing towards capacity cost. Since it is same at nodes, CCU becomes the driving factor.
T(2)	`CapacityConsumed` = 1 CCU = 0.03333 `CapacityCost` = 0.000222	`CapacityConsumed` = 1 CCU = 0.050 `CapacityCost` = 0.0005	Store1	Since demand was allocated to Store2, OMS should give CapacityConsumed as 2 for Store2
T(3)	`CapacityConsumed` = 2 CCU = 0.0666 `CapacityCost` = 0.000222	`CapacityConsumed` = 2 CCU = 0.050 `CapacityCost` = 0.0005	Store2	Observe how demand is balancing, at this time demand was allocated to Store1. This cycle of balancing continues as the average utilization across Stores increase.

This demand balancing continues until one of the ship nodes reach its threshold.

If tiebreaker of Store2 is higher, say, 0.5 USD, then initial allocations happen to Store1 only until its full utilization is comparable to Store2 cost. The pattern of demand allocation between the stores is different in such cases.
For more information, see <>.

When threshold value has reached

For the same example in Example, CapacityConsumed values are put to depict the effect of load balancing. The value may not represent the actual demand allocation based on node balancing calendar values configured.

Time series	Store1	Store2	Optimization result	Comments
T(1)	`CapacityConsumed` = 33 CCU = 1.1 `CapacityCost` = 4.2605	`CapacityConsumed` = 42 CCU = 1.05 `CapacityCost` = 0.055125	Store2	As Store2 is below threshold and the Store1 has crossed threshold, the penalty is applied at Store1.
T(2)	`CapacityConsumed` = 65 CCU = 2.2 `CapacityCost` = 8.157	`CapacityConsumed` = 48 CCU = 1.2 `CapacityCost` = 3.372	Store2	Both have crossed threshold, however, Store1 has more backlog.

Note: Store1 has CapacityCost of 8.157 USD as compared to 4.2605 USD for CCU of 2.2 and in T(1) and T(2) that is the penalty is applied per unit of backlog.