Ship-from-Store

Store as last-mile node · Walk-in protection

Fulfil online orders directly from store inventory to shorten last-mile distance and speed delivery — but carefully, because every unit shipped from a store is a unit unavailable for the walk-in customer who arrives five minutes later. Ship-from-store (SFS) is a restricted form of omnichannel fulfilment where only stores are candidate nodes; the retailer-specific twist is a walk-in demand shadow price that adjusts the naive shipping-cost objective upward by the expected lost margin on future walk-ins. Canonical references: Bell, Gallino & Moreno (2014, 2018); Hu, Li & Shou (2022).

Why it matters

Stores double as mini-fulfilment centres in the e-commerce era

~40%
Share of U.S. omnichannel retailers’ online orders now fulfilled in part from store inventory (Walmart, Target, Macy’s, Kohl’s).
Source: NRF / Digital Commerce 360 retail-survey aggregates.
1–3 days
Typical delivery-time reduction vs DC-only fulfilment when a local store has inventory — drives the “fast shipping” customer promise.
Source: Acimović & Graves (2015); industry operations reports.
5–15%
Walk-in demand loss when SFS is run naively (nearest-store always wins) vs. with walk-in-protection shadow prices.
Source: Hu, Li & Shou (2022) review.
+10 pp
Inventory-turn improvement documented at retailers that deployed SFS alongside markdown discipline — slower-turning store stock becomes e-commerce supply.
Source: Bell, Gallino & Moreno (2018), Management Science.

Where the decision sits

Operational routing · with a strategic inventory-protection overlay

SFS fits between general omnichannel fulfilment (which is also allowed to ship from a DC or dark store) and pure store operations. The modelling question is: given an online order and a candidate store with inventory, what is the true total cost of shipping from that store? Naive answer = per-order shipping + labour. True answer = shipping + labour + expected lost margin on the walk-in that will find no stock. The difference matters most in fast-turning, high-margin categories.

Order arrivesSKU + zone
Choose storeor fall-back to DC
Pick & shipstore labour
Walk-in recoverynext replenishment

Problem & formulation

Assignment MIP plus walk-in protection shadow price

OR family
Assignment with shadow prices
Complexity
LP (integer-valued)
Solver realism
★★★ Greedy + protection
Reference
Bell-Gallino-Moreno (2014, 2018)

Parameters

SymbolMeaningUnit
\(i \in \mathcal{O}\)Pending online orderfinite
\(j \in \mathcal{S}\)Store (candidate SFS origin)finite
\(0\)DC (fall-back option, always has stock)—
\(I_j\)Store \(j\)’s inventory for the SKUunits
\(c_{ij}\)Shipping cost from \(j\) to order \(i\) (includes labour)$
\(c_{i0}\)Shipping cost from DC to order \(i\) (higher)$
\(\mu_j\)Expected remaining walk-in demand at store \(j\) this cycleunits
\(m\)Gross margin on a walk-in sale$ / unit
\(\lambda_j(I)\)Walk-in shadow price = probability the next walk-in is lost × margin, given current stock \(I\)$

Decision variable

SymbolMeaningDomain
\(x_{ij}\)1 if order \(i\) ships from store \(j\); 0 otherwise. \(x_{i0} = 1\) means ship from DCbinary

Walk-in shadow price

When stock is abundant, the marginal cost of one more SFS unit is zero (plenty left for walk-ins). When stock is tight, it approaches the full walk-in margin. A simple closed-form uses the newsvendor critical-fractile intuition:

$$\lambda_j(I) \;=\; m \cdot \bar F_j(I) \;=\; m \cdot \mathbb{P}(D^{\text{walk-in}}_j \geq I)$$

\(\bar F_j\) is the survival function of walk-in demand. When \(I\) is large, \(\bar F_j(I) \approx 0\) and SFS is free of walk-in cost. When \(I\) is small (or zero), \(\bar F_j(I) \approx 1\) and every SFS unit costs an expected walk-in margin. Bell-Gallino-Moreno 2018 implement a richer Bayesian version.

Objective

$$\min \; \sum_{i \in \mathcal{O}} \sum_{j \in \mathcal{S} \cup \{0\}} \bigl(\, c_{ij} \;+\; \lambda_j(I_j - \text{SFS so far}_j) \,\bigr) \, x_{ij}$$

Each assignment pays the true cost: shipping+labour plus the walk-in shadow price evaluated at the post-assignment inventory level. DC has \(\lambda_0 = 0\) by convention.

Constraints

$$\sum_{j} x_{ij} = 1 \;\;\forall\, i \qquad \sum_{i} x_{ij} \leq I_j \;\;\forall\, j \in \mathcal{S} \qquad x_{ij} \in \{0, 1\}$$

One assignment per order; store inventory capacity; DC has no capacity bound in the simplified version.

Interactive solver

Greedy-assignment with and without walk-in protection — compare the two

SFS policy solver
15 orders · 4 stores + 1 DC · Poisson walk-in demand
★★★ Greedy + shadow
—
Protected policy cost ($)
—
Naive nearest-store cost ($)
—
Savings vs naive
—
SFS share of orders
—
Expected walk-in loss (protected)
—
Expected walk-in loss (naive)
Store (gold fill = stock left, hollow = empty) DC (central fall-back) SFS-assigned order DC-assigned order

Under the hood

The scenario generator places 4 stores (fixed coords) + 1 DC (grid centre) + \(n\) online orders (random zones). Store \(j\)’s walk-in survival function is approximated as Poisson with rate \(\mu_j\): \(\bar F_j(I) = \mathbb{P}(\text{Poisson}(\mu_j) \geq I)\). The protected policy assigns each order to the store / DC with minimum \(c_{ij} + \lambda_j(I_j - \text{used}_j)\), decremented after each assignment. The naive policy assigns to the nearest in-stock store (or DC). Expected walk-in loss for each policy is \(m \cdot \sum_j \mathbb{E}[(D^{\text{walk-in}}_j - I_j^{\text{left}})^+]\), computed via Poisson tail. Realistic savings of 2-10% typically emerge when walk-in demand is high relative to inventory.

Reading the solution

Three patterns to watch for

  • Protected policy routes to DC more often as inventory dips. As a store depletes, the shadow price rises, eventually making DC cheaper. Naive policy keeps shipping from store and strands walk-ins.
  • High-margin SKUs get more protection. Double the margin and the shadow price doubles — fewer SFS, more DC.
  • DC fall-back costs the retailer slower delivery but saves walk-ins. The SLA penalty trade-off appears in the cross-link to omnichannel-fulfilment.

Sensitivity questions

  • What if walk-in demand drops (seasonal low)? — shadow prices collapse; SFS becomes free; lift vs naive shrinks.
  • What if store inventory is deeper? — same: shadow prices drop; protected and naive converge.
  • What if DC ship cost doubles (fuel shock)? — fall-back is more expensive; protected policy is willing to risk more walk-in losses; SFS share rises.

Model extensions

Dynamic shadow prices

Bell-Gallino-Moreno 2018: use a Bayesian filter to update walk-in demand belief as the cycle unfolds; Lagrangian look-ahead gives future-demand-aware pricing.

Multi-SKU

When stores share labour but have separate per-SKU inventory, shadow prices couple across SKUs via the labour constraint.

Return-aware SFS

Route online orders AWAY from zones with high return rates — saving reverse-logistics cost.

Returns management →
Joint SFS + replenishment

SFS depletes store stock faster than planned; replenishment policy must adapt (higher \(S\), more frequent delivery).

Store replenishment →
Omnichannel embedding

SFS is one of several fulfilment modes; full model optimises across DC / SFS / BOPIS / dark store.

Omnichannel fulfilment →
BOPIS sibling

BOPIS uses store inventory too, but customer picks up rather than ships — different labour and shadow-price structure.

BOPIS →
Strategic capacity planning

If SFS is a permanent channel, store labour and inventory should be designed for it — upstream decision.

Store location →
Routing after assignment

Once assigned, SFS orders batch into delivery routes — logistics last-mile problem.

Delivery routing →

Key references

Bell, D. R., Gallino, S. & Moreno, A. (2014).
How to win in an omnichannel world.
MIT Sloan Management Review 56(1): 45–53.
Bell, D. R., Gallino, S. & Moreno, A. (2018).
Offline showrooms in omnichannel retail: Demand and operational benefits.
Management Science 64(4): 1629–1651. doi:10.1287/mnsc.2016.2684
Acimović, J. & Graves, S. C. (2015).
Making better fulfillment decisions on the fly in an online retail environment.
M&SOM 17(1): 34–51. doi:10.1287/msom.2014.0505
Hu, M., Li, X. & Shou, B. (2022).
Ship-from-store strategy in omnichannel retail.
Operations Research (literature review and model).
Gallino, S. & Moreno, A. (2014).
Integration of online and offline channels in retail.
Management Science 60(6): 1434–1451. doi:10.1287/mnsc.2014.1951
Dong, L., Shi, C. & Zhang, F. (2022).
Inventory strategy for omnichannel ship-from-store operations.
Production and Operations Management.
Jasin, S. & Sinha, A. (2015).
An LP-based correlated rounding scheme for multi-item ecommerce order fulfillment.
Operations Research 63(6): 1336–1351. doi:10.1287/opre.2015.1441
(EJOR 2022).
The revival of retail stores via omnichannel operations.
European Journal of Operational Research. doi:10.1016/j.ejor.2021.12.001

Back to the retail domain

Ship-from-store sits in the Place × Operational cell — store inventory doubles as last-mile supply when walk-in shadow prices are respected.

Open Retail Landing
Educational solver · Poisson walk-in model and myopic assignment · validate against your own walk-in demand distribution before scaling SFS.