Rebalancing stock between stores and online with AI
Rebalancing moves inventory to where it will sell. AI helps decide what to move and when, but transfer costs, stock accuracy and fulfilment rules decide whether it pays off.
KEY TAKEAWAYS Summary by the editors
- Stock rebalancing means moving inventory between stores, or between stores and an online fulfilment location, when demand differs from where stock sits; it is usually treated as a last resort because transfers cost money and time.
- AI contributes by forecasting demand at store and online level, ranking candidate transfers by expected gain, and netting off logistics cost, but the result depends on accurate stock data.
- Pacsun said in 2022 that it allocated inventory for omnichannel demand and used stores as online fulfilment locations, reporting that the number of ship completes doubled and shipping distance fell.
- In 2018 Retail Dive reported that Zara had RFID-enabled stock management in stores in 25 markets, giving real-time insight into what each store holds and allowing online orders to be filled from store inventory.
- Benefit figures from vendors, such as sell-through gains of up to 3%, are unaudited, so a retailer should test on a subset of stores first.
Rebalancing stock between stores and online means moving inventory to the place where it is more likely to sell, or using stores to fulfil web orders. AI helps by forecasting demand per location, ranking possible transfers by expected benefit after costs, and keeping the calculations current. It pays off only when stock data is accurate and the cost of moving goods is lower than the gain in sales or margin.
This how-to explains when rebalancing is worth doing, what AI contributes, the data it needs, and how to run a controlled rollout. It uses published company statements and notes where figures are vendor claims.
When is rebalancing worth doing?
A vendor article on inventory rebalancing in fashion describes the triggers: consumer behaviour changes, items end up in the wrong channel or store, market conditions shift, and complex distribution networks make moving stock between points of sale sometimes more sensible than warehouse transfers. The same article admits that rebalancing is traditionally treated as a last resort, with leaders advising against frequent moves except for unique or high-ticket items, and that calculating transfers manually is practically impossible at scale. It also says benefits should be calculated after logistics costs, which is the right test.
In plain terms, a transfer is worth making when the expected extra margin at the receiving location, less handling and shipping cost and the sales lost at the sending location, is positive and exceeds a threshold that covers forecast error.
| Option | What moves | Best for | Main cost or risk |
|---|---|---|---|
| Store to store transfer | Units between stores | Slow sellers at one store, strong demand at another | Handling, shipping time, store labour |
| Store to online fulfilment | Units shipped from stores to web orders | Online orders for items short in the web depot | Pick and pack effort in store, risk of empty shelves |
| Online depot to store | Units sent to stores | Items selling out in stores while sitting in the depot | Replenishment lead time |
| Hold and sell where it sits | Nothing | Low-margin items, small differences in demand | Possible markdowns later |
What does AI add to the decision?
- Demand by location: forecasts per store and for online, including the effect of local events and weather where data exists.
- Candidate generation: lists the transfers that would raise expected sales, with a size by size view, which a human cannot check at scale.
- Cost netting: compares the expected gain with transfer cost, so low-value moves are filtered out.
- Timing: recalculates as sales arrive, so a recommendation made at the start of the week is not followed blindly at the end.
Vendor material shows the type of claim in circulation. Nextail, a vendor, reports sell-through improvements of up to 3%, up to 80% less manual work, and up to 86.5% of end-of-season inventory sold through store transfers. These are vendor claims, not independently verified, and the article gives no transfer cost figures. They are useful for framing a pilot, not as targets.
What have retailers published about stores and online working together?
Pacsun said in a 2022 announcement that it added Antuit.ai's allocation technology to its allocation and fulfilment processes so that inventory is allocated for both store and online demand, and that it uses stores as online fulfilment locations. The announcement states that the companies doubled the number of ship completes, reduced shipping distance for orders, improved forecasting for store and online demand and balanced inventory between stores and the web depot. Pacsun's co-CEO is quoted as saying that the company allocated inventory for omnichannel demand and minimised split shipments. The only numeric result in the announcement is the doubling of ship completes, and it is a company and vendor statement.
Inditex offers an earlier example of store stock visibility. Retail Dive reported in September 2018 that Zara had RFID-enabled stock management in stores in 25 markets, giving the company real-time insight into what each store holds, and that the system lets Zara fill online orders from store inventory, with same-day delivery in cities with both stores and e-commerce and next-day shipping elsewhere. The article reported plans to roll the store-stock system out to all Inditex brands by 2020 in countries where they run physical stores. It describes the position in 2018, so current practice may differ.
What data and rules do you need first?
A rebalancing engine can only be as good as its inventory record. Stores that carry out regular counts or use item-level tagging can trust the numbers, and others cannot. Beyond stock, the engine needs rules that reflect how the business actually runs.
- Inventory accuracy by store: measure the gap between system stock and counted stock before automating anything.
- Transfer cost: handling, shipping and time, by lane, so the engine can net it off.
- Protected stock: minimum display quantities, new arrivals and campaign items that must not be moved.
- Fulfilment capacity: how many web orders a store can pick per day without hurting the shop floor.
- Channel rules: whether online has a pooled stock position or a reserved share for each store.
How do you roll it out?
- Start with a subset of categories and stores where stock accuracy is high.
- Run recommendations in advisory mode first and compare them with what planners would have done.
- Measure three things: sales and sell-through against control stores, transfer cost per unit moved, and the share of recommendations accepted.
- Add online fulfilment from stores in a few locations, watching split shipments and shipping distance, the measures Pacsun highlighted.
- Review the rules every season, as transfer costs and campaign calendars change.
What are the limits and risks?
- Forecast error: small differences between stores are within noise, so moves based on them lose money after transfer cost.
- Stock inaccuracy: a transfer based on stock that is not there creates cancellations and frustrated customers.
- Returns and in-transit stock: if these are not visible, the engine may move goods that are already moving.
- Sustainability: extra shipping has an emissions and cost footprint that should be part of the calculation.
- Wholesale-led brands: stock held by partners is outside the brand's control, so rebalancing applies only to owned retail and online.
The sensible approach is incremental: fix inventory accuracy, define transfer rules, test on a few stores, and expand only when the net gain after logistics cost is visible in the control comparison.
Frequently asked questions
What is stock rebalancing in retail?
It is moving inventory between stores, or between stores and online fulfilment locations, to match where demand is. It is often treated as a last resort because transfers cost money and time, so each move needs an expected gain that exceeds the cost.
How does AI help rebalance inventory between stores and online?
It forecasts demand per location, lists candidate transfers by expected gain, nets off logistics cost and recalculates as sales arrive. It depends on accurate stock data and clear rules for protected quantities.
Which retailers fulfil online orders from stores?
Published examples include Pacsun, which said in 2022 it uses stores as online fulfilment locations, and Zara, which Retail Dive reported in 2018 could fill online orders from store inventory using RFID-enabled stock management in 25 markets.
What is a realistic benefit from stock rebalancing?
Published figures come mostly from vendors, for example sell-through gains of up to 3%, and are not independently verified. Results depend on stock accuracy and transfer cost, so test on a subset of stores with matched controls.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
