Intelligent routing is an order routing mode in NewStore that scores each fulfillment location in a location group and routes an order to the locations with the highest scores. The score is based on the available-to-promise (ATP) inventory, which a location holds for the items in the order, and on the number of items already waiting to be fulfilled at that location.
NewStore supports two routing modes:
Priority-based routing (default): The routing engine tries locations in the order you define in your routing ruleset. You maintain this order manually.
Intelligent routing: The routing engine re-orders the locations within each location group for every routing attempt, based on a weighted score. The routing engine uses the order you define in the routing ruleset only to break ties.
Both modes use the same routes, shipping option priorities, location groups, and split shipment settings. Intelligent routing only changes the order in which the routing engine considers locations inside a location group. See About order routing for how routing works in general.
Important
NewStore enables intelligent routing per tenant on request. Intelligent routing requires the routing ruleset (fulfillment configuration v2) with location groups. See Enabling intelligent routing.
How intelligent routing compares with priority-based routing
Behavior | Priority-based routing | Intelligent routing |
|---|---|---|
Order of locations within a location group | The order defined in the routing ruleset. | Sorted by weighted score, highest first. The routing ruleset order breaks ties. |
Order of location groups in a route | The order defined in the route. | The order defined in the route. The routing engine tries every location in the first group before any location in the second group, regardless of score. |
Single shipment preferred | Yes | Yes |
Fallback when no single location can fulfill the order | Greedy fulfillment* over the ruleset order. | Greedy fulfillment* over the scored order. |
Maintenance | You update location priorities manually when inventory or workload shifts. | Scores follow ATP and workload automatically. You rarely need to update priorities. |
Optimizes for | Your defined priorities. | Locations with the deepest stock for the ordered items and the lightest fulfillment workload. |
*If no single location can fulfill the whole order, the routing engine reviews the ranked list again and assigns as many items as possible to each location in turn. NewStore calls this | ||
How intelligent routing selects a location
For each routing attempt, the routing engine:
Finds the route that matches the destination country, ZIP code region, and shipping option. The route lists one or more location groups in priority order.
Removes locations that are not eligible, for example locations that have already rejected the fulfillment request, or stores that have no remaining capacity.
Calculates a weighted score for every remaining location in the location group and sorts the group by score, highest first. If two locations have the same score, the location that appears first in the routing ruleset wins.
Concatenates the sorted groups in the order defined in the route. For a route with location groups
[A, B], every location inAranks above every location inB, regardless of score.Looks for the highest-ranked location that can fulfill the whole order in a single shipment. If one exists, the routing engine routes the order there. A location that can ship the whole order wins over locations with higher scores that can ship only part of it.
If no single location can fulfill the whole order, reviews the ranked list again and assigns as many items as possible to each location in turn. NewStore calls this greedy fulfillment. Your split shipment configuration still applies.
If no combination of locations can fulfill the order within the split shipment limits, places the order on hold.
Calculating the weighted score
ATP is the main contributor to the score. The fulfillment queue, which is the number of order items waiting to be fulfilled at the location, acts mainly as a tie-breaker between locations with similar ATP.
For a location L and the set of products P in the order:
Term | Description |
|---|---|
The ATP of product | |
The number of products in the order | |
The sum over each product | |
The highest ATP of product | |
The number of order items in the fulfillment queue of location | |
The total number of queued order items across all eligible locations. |
The queue term is 1 for a location with an empty queue. It never reaches 0 because of the + 1 in the denominator.
For example, a location with an ATP term of 0.5 and 3 queued items out of 9 in total has this score:
W = 0.5^0.9 × (1 − 3/10)^0.1
= 0.536 × 0.965
≈ 0.517The queue moves the score only from 0.536 to 0.517. This is why ATP dominates the ranking and the queue acts as a tie-breaker.
Example A: Single-item orders across three stores
Store A, Store B, and Store C are in one location group, listed in the routing ruleset in that order. Each store starts with an ATP of 5 for the ordered stock keeping unit (SKU) and an empty fulfillment queue. NewStore routes a series of single-item endless aisle orders for that SKU.
Order | ATP (A / B / C) | Queue (A / B / C) | Score (A / B / C) | Routed to | Reason why the store was selected |
|---|---|---|---|---|---|
1 | 5 / 5 / 5 | 0 / 0 / 0 | 1.00 / 1.00 / 1.00 | Store A | All scores are equal. Store A is first in the routing ruleset. |
2 | 4 / 5 / 5 | 1 / 0 / 0 | 0.76 / 1.00 / 1.00 | Store B | Store B and Store C tie. Store B is ahead in the routing ruleset. |
3 | 4 / 4 / 5 | 1 / 1 / 0 | 0.79 / 0.79 / 1.00 | Store C | Store C has the highest ATP and an empty queue. |
4 | 4 / 4 / 4 | 1 / 1 / 1 | 0.97 / 0.97 / 0.97 | Store A | All scores are equal again. Scores are relative, so they rise from 0.79 to 0.97 when every store reaches the same ATP. |
5 | 3 / 4 / 4 | 2 / 1 / 1 | 0.73 / 0.98 / 0.98 | Store B | Store B and Store C tie on ATP and queue. Store B is ahead in the routing ruleset. |
If Store C then ships the item in its queue, its queue drops to 0 while Store A and Store B still hold two queued items each. With an equal ATP of 3 at every store, Store C scores 1.00 against 0.95 for the other stores, so Store C receives the next order. The queue decides the result only because the ATP is equal.
Example B: Fewer shipments than priority-based routing
An order contains one unit each of SKU 001, SKU 002, and SKU 003. Four locations are eligible, listed in the routing ruleset in the order L1, L2, L3, L4. All fulfillment queues are empty, so the ranking follows the ATP term.
Location | SKU 001 | SKU 002 | SKU 003 | ATP term |
|---|---|---|---|---|
L1 | 1 | 0 | 0 | (0.1 + 0 + 0) / 3 = 0.033 |
L2 | 0 | 1 | 0 | (0 + 0.1 + 0) / 3 = 0.033 |
L3 | 0 | 0 | 1 | (0 + 0 + 1) / 3 = 0.333 |
L4 | 10 | 10 | 0 | (1 + 1 + 0) / 3 = 0.667 |
No single location can fulfill the whole order, so both routing modes fall back to greedy fulfillment.
Priority-based routing walks the ruleset order. L1 takes SKU 001, L2 takes SKU 002, and L3 takes SKU 003. The result is three separate shipments.
Intelligent routing sorts the locations by score into L4, L3, L1, L2. L4 takes SKU 001 and SKU 002, and L3 takes SKU 003. The result is two separate shipments.
Example C: More shipments than priority-based routing
Intelligent routing can also produce more shipments than priority-based routing. This happens when one location holds a small quantity of several items in the order, while other locations each hold a large quantity of a single item. The same order and ruleset order apply, with empty queues.
Location | SKU 001 | SKU 002 | SKU 003 | ATP term |
|---|---|---|---|---|
L1 | 1 | 1 | 0 | (0.01 + 0.01 + 0) / 3 = 0.007 |
L2 | 0 | 0 | 1 | (0 + 0 + 1) / 3 = 0.333 |
L3 | 100 | 0 | 0 | (1 + 0 + 0) / 3 = 0.333 |
L4 | 0 | 100 | 0 | (0 + 1 + 0) / 3 = 0.333 |
Priority-based routing visits L1 first. L1 takes SKU 001 and SKU 002, and L2 takes SKU 003. The result is two separate shipments.
Intelligent routing sorts the locations into L2, L3, L4, L1. L2 takes SKU 003, L3 takes SKU 001, and L4 takes SKU 002. The result is three shipments.
Example D: A multi-item order with two location groups
A route lists two location groups. Group 1 contains Location A to Location D and group 2 contains Location E to Location G. An order contains one unit each of SKU 001, SKU 002, and SKU 003.
Group 1 location | SKU 001 | SKU 002 | SKU 003 | Queue | Score |
|---|---|---|---|---|---|
Location A | 4 | 4 | 0 | 0 | 0.57 |
Location B | 0 | 0 | 5 | 2 | 0.36 |
Location C | 5 | 5 | 0 | 2 | 0.68 |
Location D | 3 | 3 | 3 | 4 | 0.60 |
Location C has the highest score, but it cannot ship SKU 003. Location D is the only location in group 1 that can ship the whole order in one shipment, so the routing engine routes the order to Location D. The routing engine does not consider group 2, because group 1 can fulfill the order.
If no location in group 1 can ship the whole order, the routing engine looks for a single-shipment option in group 2 before it splits the order across locations.
Example E: A Distribution center in the same location group as stores
A location group contains a distribution center (DC) and three stores. The DC starts with an ATP of 50 for a SKU, and each store starts with an ATP of 5. The DC has the highest ATP, so it wins the first orders. Each order the DC wins adds an item to its fulfillment queue, which slowly lowers its score. After the DC has won 43 orders, with an ATP of 7 and 43 queued items, its score drops to 0.68 while each store scores 0.74. From that point, the stores start winning orders.
If the routing engine must always try your DC before stores, place the DC in its own location group ahead of the store groups. Intelligent routing never ranks a location in a later group above a location in an earlier group.
Recommended location group structure
Use one location group for each set of locations that you want to score against each other. Put locations that the routing engine must always try first, such as a DC, in a separate group at the start of the route.
The following routing ruleset routes orders in two regions of a country to a DC first and then to the stores in that region. Intelligent routing re-orders the stores within Region East and Region West for every routing attempt. The DC is in its own group, so it is always the highest-priority location, regardless of its score.
{
"routes": [
{
"fulfilled_by": {
"2DAY": ["DC", "Region East"],
"STANDARD": ["DC", "Region East"]
},
"destination_region": {
"country": "US",
"zip_code_region": "Region East"
}
},
{
"fulfilled_by": {
"2DAY": ["DC", "Region West"],
"STANDARD": ["DC", "Region West"]
},
"destination_region": {
"country": "US",
"zip_code_region": "Region West"
}
}
],
"sl_levels_priority": ["2DAY", "STANDARD"],
"location_groups": {
"DC": ["DC01"],
"Region East": ["store_101", "store_102", "store_103"],
"Region West": ["store_201", "store_202", "store_203"]
},
"zip_code_regions": {
"Region East": ["0*", "1*", "2*"],
"Region West": ["8*", "9*"]
}
}See Tutorial: Setting up a fulfillment configuration for the full routing ruleset format. See Managing fulfillment location groups to manage location groups via Omnichannel Manager.
How rejected fulfillment requests affect scoring
When a location rejects a fulfillment request, for example because of a mispick, the rejection affects intelligent routing in two ways:
A mispick sets the ATP of the item at that location to
0. This lowers the ATP term of the location for every later order that contains the item, until the ATP changes again.Before scoring, the routing engine removes the location that rejected the request from the list of eligible locations, so it does not route the request back to that location. If you enable the
reroute_to_rejected_fulfillment_nodesproperty, the routing engine instead moves rejected locations to the end of the list. They can receive the request again when no other location can fulfill it.
See Fulfillment rejection rerouting and Configuring order routing settings.
Intelligent routing with preferred source routing
If you also use preferred source routing, the routing engine routes in two passes: first to the preferred locations in the location group, then to the remaining locations. Intelligent routing scores the locations within each pass separately. The routing engine always tries preferred locations first, regardless of their score. See the preferred source routing section in About order routing.
Enabling intelligent routing
Before you enable intelligent routing, ensure that your tenant meets these requirements:
Your tenant uses the routing ruleset (fulfillment configuration v2) with location groups, not the legacy fulfillment configuration v1. The legacy configuration does not define location groups, so the routing engine cannot guarantee the order of locations, for example that a DC is always tried first. See Migrating from the legacy fulfillment configuration to the new fulfillment configuration.
Each location group contains only locations that you want to score against each other. Locations that the routing engine must always try first are in a separate group at the start of the route.
Intelligent routing has no additional limitations compared with priority-based routing. All existing routing features, such as split shipment limits, store capacity, and on-hold handling, continue to apply.
To enable and validate intelligent routing:
Create a support request with your NewStore contact point to enable intelligent routing for your sandbox or staging tenant.
Set up the intended routing ruleset in the sandbox or staging tenant.
Test the routing behavior by injecting test orders, or by simulating orders in NewStore Omnichannel Manager. See Simulating routing options.
Compare the resulting fulfillment locations and shipment counts with what you expect for your inventory distribution.
When you are satisfied with the results, create a support request to enable intelligent routing for your production tenant.