Stock Transfers Between Stores: When Moving Stock Beats a New Order

A product that has not sold in one store for weeks and keeps selling out in another looks like stock the chain already owns. Moving it means no new order, no supplier minimum and no wait for delivery. Sometimes it really is the cheapest fix. The research on transfers between stores, known in operations research as lateral transshipment, shows when it pays, when it costs more than it saves, and when the need for it is a sign of a different problem.

1. Transfers fix imbalance, not shortage

A transfer can only move stock the chain already holds. When there is too little of a product across the whole network, moving it between stores mostly moves the empty shelf. A study of 212 health facilities sharing medicines found exactly that: transfers made "little impact on the overall service level" when stock across the system was scarce, and paid off only when there was enough in total, sitting in the wrong places¹. So the first question is the chain total: if it is short, the answer is an order, not a transfer.

2. What a transfer costs

The freight is the visible part. At LC Waikiki, Turkey's largest apparel retailer, moving an item between stores costs roughly 1 to 5% of its price, and sending it back to the depot 3 to 7%². The hidden part is store labour. LC Waikiki sends the whole stock of a product to a single store rather than splitting it, because of "the considerable labor and time burden associated with sorting such items", and its stores can pack for, or unpack from, only a limited number of other stores each week².

Grocery is harder still. When researchers interviewed twelve grocery retailers in Germany about preventing waste, store-to-store transfers came out as low impact and rarely used; one retailer put it as "logistics costs eat up potential earnings", and only nearby stores were considered at all³. A transfer pays when the margin it saves is clearly larger than the freight plus the time two stores spend on it:

3. Not every empty shelf needs a transfer

Some of the demand for a missing item is not lost; it moves to the nearest substitute. In a sportswear chain of 217 stores, about half the unmet demand for an out-of-stock shoe went to the next size of the same style, a fifth to other styles, and the rest, a little over a quarter, was lost⁴:

The exact split is 51.7%, 20.2% and 28.1%, and it changes when stock-outs spread: when many items are missing at once, fewer customers find a substitute and lost sales rise sharply⁴. A transfer pays most for the item without a close substitute, or in a store where several of its substitutes are out at the same time. The out-of-stocks guide shows how to find those shelves first.

4. Which stores to connect

A chain of n stores has n² possible routes between them, against 2n if everything goes through a warehouse⁶. Few chains can manage hundreds of routes, and they do not need to. Using three years of grocery sales, Griffin and colleagues showed that grouping stores whose demand does not move together, so that one has spare stock when the other runs short, keeps most of the benefit of transfers at a fraction of the complexity⁵. In practice that means small groups of nearby stores, not one network where everything moves everywhere.

Whether to move stock directly or through the warehouse depends on freight. In a model built with the Spanish children's-wear brand Nanos, direct moves won unless shipments through the warehouse cost about half as much⁶; the authors note their figures come from generated instances, so the direction holds better than the number.

5. When transfers point to a different problem

In the health network above, a third or more of all demand was met by transfers, and the authors trace that to the first allocation: deliveries came in fixed boxes that did not match what each facility used¹. A chain that moves a large share of its stock between stores every week is usually allocating it wrongly in the first place, and the reorder point guide covers sizing each store's order to its own sales.

Transfers are also a cost of excess, not only a cure for it. Burlington Stores lists "transferring product to other stores" next to markdowns among the responses to excess inventory that "would further decrease our gross profits"⁷. Stock that sells nowhere is a dead-stock question: return, markdown or clearance, not a tour of the network.

6. A weekly transfer rule

Run it once a week, on the same day stores already receive deliveries, so the move travels with existing freight. Then check what each accepted transfer did: whether the product sold in the receiving store, or simply sat there instead.

7. Limitations

Most of the studies are models tested on real or realistic data rather than measured results, two are cited from their abstracts, and the network study is from public health rather than retail. The split of lost demand comes from footwear, where sizes substitute for one another more than most products do. Freight and labour costs differ widely by country and category.

How marql proposes transfers

marql reads sales and stock for every product in every store and proposes a transfer only where the rule above holds: one store holds well above its maximum, another sells the product and is at or below its reorder point, the sender keeps at least a week of sales, and the margin on the units covers the cost of moving them. Warehouses are left out. These are three of the moves it listed for Berezka Retail, a chain of 20 stores in Romania, in its success story:

Every transfer is a proposal that a person accepts or dismisses. In a chain on the e:corg ERP an accepted transfer becomes a draft there; with other systems it leaves marql as a file.