Store closures used to sit at the edge of supply chain planning, filed somewhere between real estate, finance, and store operations. That is no longer good enough. In 2025, the U.S. market saw 8,270 store closures, with about 7,900 more projected for 2026; by mid-2025 alone, 123.7 million square feet had been vacated. Five bankrupt brands — Rite Aid, Joann, Party City, Big Lots, and Forever 21 — accounted for more than half of the 2025 closures, turning many exits into liquidation events rather than orderly network adjustments.[1]
That scale changes the supply chain question. A store closure is not just the last day of sales at a location. It is a demand signal disappearing from one node, displaced customer demand moving unevenly to nearby stores and digital channels, inventory stranded in the wrong building, inbound purchase orders still pointed at the old forecast, and transportation capacity suddenly asked to do work it was never scheduled to do.
This is why AI supply chain planning for retail store closures is a useful stress test. The closure decision itself may come from the boardroom, a bankruptcy court, a lease calendar, or a portfolio review. The supply chain consequence lands elsewhere: planners have to decide what inventory still has value, which stores can absorb it, which products should be marked down in place, which routes need to change, and which demand forecasts should stop treating the closing store as if it will keep behaving normally.
The hardest part is not that there are many tasks. Retail has always had many tasks. The hard part is that closure compresses them. Decisions that might normally unfold across seasons, reset calendars, replenishment cycles, and markdown waves have to be made while associates, landlords, vendors, carriers, and customers are all reacting at once.
Closure Turns Ordinary Weaknesses Into One Short-Window Failure
A retailer can carry a surprising amount of planning weakness during stable weeks. A store can have inaccurate on-hands, slow allocation logic, a rough markdown calendar, and a distribution center lane that everyone knows is awkward, yet the system still appears to work because sales history keeps repeating enough to hide the seams. Closure removes that cover.
The closing location is no longer a normal demand node. Nearby stores are no longer normal comparison stores. E-commerce may pick up demand for some categories and miss others entirely. Inventory that looked clean on a weekly report becomes a value-recovery problem: move it, mark it down, liquidate it, or leave it to decay while the decision queue grows.
That is where maturity becomes visible. A retailer with weak inventory accuracy does not suddenly become accurate because an AI dashboard is present. A retailer without credible transfer logic does not become nimble because a model can rank stores by theoretical demand. A retailer whose logistics systems cannot support multi-node fulfillment does not become closure-ready because it has run a pilot in one region.
The practical closure workflow is less elegant than most transformation decks imply. First, the retailer has to re-forecast demand around the disappearing node. Then it has to value the inventory alternatives: transfer, markdown, hold for online fulfillment, return to vendor where possible, or liquidate. Then transportation has to move goods without consuming so much capacity that the rest of the network suffers. Finance has to see the margin consequence quickly enough to approve the better option before the cheaper-looking option becomes the only option.
| Closure decision area | What planning maturity changes |
|---|---|
| Demand re-forecasting | Whether nearby stores and channels receive realistic demand shifts instead of stale history |
| Inventory transfer | Whether goods move to nodes that can sell them profitably, not merely to the closest open store |
| Markdown timing | Whether price action preserves margin before liquidation becomes the default |
| Logistics rerouting | Whether transportation capacity can support transfers without breaking replenishment elsewhere |
| Financial review | Whether planners can compare recovery options fast enough for the answer to matter |
The inventory transfer and markdown decisions deserve the most attention because they are where value is visibly preserved or lost. A product sitting in a closing store may be a liability in that building and a margin opportunity two towns over. Another product may not justify a transfer at all once handling, freight, remaining seasonality, and local demand are considered. Rule-based systems can move goods by distance, store tier, or weeks of supply. Better planning has to ask whether the receiving node can actually convert that unit into cash at a better recovery rate than the alternatives.
The Maturity Gap Is Already Measurable
The closure shock would be easier to absorb if most retailers already had strong everyday execution. The evidence points the other way. In a 2024 Simbe/Coresight survey of 150 retail decision-makers, 70% of retailers reported losing at least 4.5% of operating margin to in-store inefficiencies, including out-of-stocks, poor allocation, and planogram non-compliance.[2]
Those are not closure-specific losses, and they should not be presented as if they are. Their value is diagnostic. Out-of-stocks, poor allocation, and planogram non-compliance are exactly the kinds of issues that become more expensive when a store exits the network. If a retailer cannot trust what is in the store, cannot allocate cleanly under normal demand, and cannot keep execution aligned to the plan, then closure does not create the weakness. Closure accelerates the bill.
The fulfillment infrastructure gap is just as blunt. A 2025 Kase/TrendCandy survey of 250 retail supply chain leaders found that 84% struggle to align IT infrastructure for multi-node fulfillment.[3] The survey was commissioned by a 3PL technology provider, so it should be read with that context. Still, the problem it identifies is the exact one a closure event stresses: inventory, demand, and fulfillment decisions have to move across nodes quickly, not stay trapped inside channel or location silos.
The AI strategy gap sits above both of those operational problems. Gartner reported in 2025 that only 23% of supply chain organizations had a formal AI strategy, while 74% of CEOs believed AI would have the most significant impact on their industry.[4] That mismatch matters because closure planning is not a place where informal enthusiasm helps much. Someone has to decide what data feeds the forecast, who can override a transfer recommendation, how financial thresholds are set, and whether the model is trusted enough to change execution before the liquidation clock runs out.
This is where maturity theater becomes expensive. A retailer can have pilots, dashboards, exception reports, and automation rules and still fail the closure test. If planners have to export store inventory, rebuild demand assumptions manually, ask transportation for capacity by email, and wait for finance to approve markdowns after the best transfer window has passed, the AI environment is not production-grade for this use case. It may still be useful. It may even be a good start. But it is not yet a system that can hold the network together when a node disappears.
What the Closure Diagnostic Actually Tests
A closure scenario tests four planning behaviors, not one software feature. The first is sensing: can the retailer see the inventory and demand change early enough? The second is valuation: can it compare recovery paths for inventory with enough financial realism? The third is orchestration: can recommendations move across stores, DCs, carriers, and channels? The fourth is governance: can people approve or override decisions without turning the process back into spreadsheet triage?
Reactive liquidation is the lowest-maturity version. Inventory is sold down in place, moved in bulk, or handed to liquidators because the retailer cannot generate better alternatives quickly enough. This may be unavoidable in some bankruptcies or hard shutdowns. The point is not that liquidation is always irrational. The point is that it often becomes the default when planning time, data confidence, and execution capacity run out together.
Rule-based redistribution is one step better. The retailer can push goods to nearby stores, designated clearance locations, or e-commerce nodes based on predefined logic. This can reduce chaos, but it has obvious limits. A rule that looks reasonable in a normal allocation cycle can be wrong during closure if local demand has shifted, store labor is constrained, receiving capacity is full, or the item is too late in its lifecycle to justify a transfer.
AI-enabled optimization is different when it is connected to real execution. It can estimate where demand is likely to move, compare transfer and markdown options, surface exceptions, and recommend actions by item, store cluster, and time window. The useful version does not simply say, “move product to better stores.” It helps decide which product, how much, when, by which lane, and at what expected recovery value.
More autonomous planning is still a narrower claim than many sales materials suggest. In a closure window, autonomy would mean that routine transfer and markdown decisions can execute within approved guardrails while planners focus on exceptions. It does not mean the model resolves lease obligations, workforce displacement, bankruptcy constraints, or brand damage. Those are real closure costs, but they sit outside the supply chain planning question.
Bankruptcy Liquidations Show the Cost of Running Out of Options
The 2025 closure wave included several full-liquidation cases large enough to show what happens when the exit is no longer a tidy network redesign. Rite Aid vacated 15.5 million square feet and carried $2.16 billion in funded debt; Joann involved 800 stores and 18 million square feet; Party City involved 700 stores and 8.4 million square feet. Each moved into full liquidation.[1]
Those cases should not be flattened into a simple claim that better AI would have saved the companies. That would be careless. Debt structures, leases, merchandising relevance, vendor relationships, and court timelines all matter. The narrower supply chain question is still worth asking: before liquidation becomes the only practical path, how much inventory value could be preserved if the retailer can identify transferable goods, forecast receiving-store demand, time markdowns, and route inventory before the selling window collapses?
That counterfactual is difficult to quantify from public information, but the operating logic is clear. Fire-sale markdowns treat inventory as a problem to remove. AI-optimized transfer and markdown planning treats inventory as a set of choices with different recovery values. The difference is not academic to the planner deciding whether a pallet should go to a nearby store, an online fulfillment node, a clearance event, a vendor return process, or a liquidator.
Capability Markers From the Better End of the Spectrum
The better end of the market is not imaginary, but the evidence needs careful labels. Target has described an AI Inventory Ledger covering more than 40% of its product assortment and predicting stockouts before they are visible to store teams.[5] That is not a public proof point about store closure profitability. It is a demand and inventory visibility marker, and that matters because closure planning starts with knowing what is where and where demand is likely to surface next.
o9 Solutions describes flexible network modeling that can “seamlessly adjust for new stores, stores closing, new channels.”[6] That phrasing is important because many planning systems can model stable networks better than changing ones. A closure-ready planning environment needs store openings, closures, channel shifts, and capacity changes to be ordinary planning objects, not custom workarounds built after the disruption begins.
Invent.ai has reported a 4.8% sales uplift from AI-optimized transfer optimization and a 6% reduction in lost sales from AI-powered allocation at a 270-store fashion retailer.[7] This is vendor-sourced and not closure-specific, so it should be treated as an indicator of adjacent capability rather than proof that closure events are solved. Still, transfer optimization and allocation improvement are exactly the muscles a retailer would need when inventory has to be redeployed before markdown value evaporates.
Walmart has said its route optimization AI saved 30 million driver miles.[8] Again, that does not translate directly into closure economics. It does show that logistics AI can operate at a scale relevant to network disruption. In a closure event, the planning recommendation is only useful if freight can actually move; otherwise, a high-scoring transfer plan is just another clean report waiting behind a dock door.
Taken together, these examples point to the shape of mature planning rather than a single answer. The retailer sees inventory earlier, models the network as changeable, optimizes transfers with demand and margin in view, and has logistics intelligence strong enough to execute the plan. None of those capabilities removes the pain of closing stores. They change how much value is lost while the company is making the exit.
The Question to Ask Before the Next Cluster Closes
A meaningful store cluster closure this quarter would be a cleaner maturity test than another AI steering committee update. Could the planning environment sense demand shifting to nearby stores and channels? Could it value transfer, markdown, online fulfillment, vendor return, and liquidation paths at item level? Could it recommend movements that transportation can actually support? Could finance see the margin trade-off before the easy liquidation answer becomes the only answer left?
If the answer depends on planners rebuilding the network in spreadsheets after the announcement, the retailer may have AI activity, but it does not yet have closure-grade AI planning. If the answer is a governed workflow that senses, values, recommends, and reroutes fast enough to preserve margin, the system is closer to production maturity.
AI will not make store closures painless. It will not erase debt, leases, workforce consequences, or brand damage. But closure events reveal whether planning AI is strong enough for the moments when normal demand history disappears, inventory is suddenly in the wrong place, and the network has to be redrawn before Friday’s exception report becomes next month’s write-off.
References
- The Great American Store Closure Tracker, 2026 Edition, MMCG Invest.
- Simbe/Coresight 2024 survey, Simbe and Coresight Research, April 2024.
- Kase/TrendCandy 2025 survey, Kase and TrendCandy, 2025.
- Gartner 2025 supply chain AI strategy data, Gartner, 2025.
- AI Inventory Ledger, Target.
- Flexible network modeling for retail planning, o9 Solutions.
- AI-optimized transfer optimization and allocation case study, Invent.ai.
- Route optimization AI, Walmart.
Comments
Join the discussion with an anonymous comment.