Testing supply-chain changes before they touch the real one
Retailers that want to change how goods flow through their supply chain usually face a blunt choice: model the change on a spreadsheet and hope, or roll it out live and find out. Target has spent the past period building a third option in-house: a piece of software called Proxima, a digital twin of the "middle mile" of its supply chain, the leg between distribution facilities and stores where inventory positioning decisions get made.
Target described the system in a company newsroom post on August 13, 2026, later covered by Supply Chain Dive and Retail Dive. A digital twin, in this context, is a simulation built on the same underlying data and logic as the live operational system it mirrors, so a test run through it behaves the way the real network would. Target's supply chain, product, and engineering teams built Proxima together, purpose-built for the scale of Target's own network rather than adapted from off-the-shelf software.
What it replaces
The middle mile is the part of a retail supply chain most exposed to compounding error: a distribution decision made incorrectly there does not surface until it shows up as an empty or overstocked shelf days later, by which point tracing the cause back through the network is slow and expensive. That lag is what makes testing changes before deployment valuable, and it is what a spreadsheet model or a smaller-scale trial run struggles to replicate, since neither carries the full complexity of the live network with it.
The pitch is straightforward: before Target changes how inventory moves, a team can run that change through Proxima first. If the simulation surfaces a problem, the team can fix it or choose a different approach before anything touches the live network. Target put it this way in its announcement: "If a digital twin emulation through Proxima surfaces a potential problem, teams can correct the issue or take another path."
Two examples from Target's own account of the tool give a sense of what that looks like in practice. In a small pilot covering 63 fresh food items, the company used Proxima to simulate changes to inventory flow, catch issues before they reached stores, and improve on-shelf availability for those items by 2.5%. Separately, before opening a new 1.2-million-square-foot facility in Houston, Target used Proxima to simulate how inventory would move in and out of the building, reaching roughly 98% accuracy against what actually happened once the facility opened, and used that to reduce the disruptions that typically follow a large facility launch.
Sousan Ortega, Target's senior vice president of field replenishment, said: "Proxima creates faster feedback loops and helps us see around corners so we can modernize complex supply chain processes with confidence." Siobhán McFeeny, Target's senior vice president of technology, added: "Proxima is a powerful example of that vision in action. The result is a more resilient operation and an even more reliable experience for guests."
Where this fits against the rest of the industry
Most of the recent supply-chain technology news cycle has been about startups selling AI agents to freight brokers and logistics operators, funded by venture rounds. Proxima is a different kind of story: a large retailer building simulation infrastructure internally, aimed at its own operations rather than at a product it plans to sell. Target says it intends to extend Proxima beyond middle-mile inventory to cover more parts of its network, and frames the tool's simulation output as a foundation it could eventually connect to AI-driven, agentic decision support, systems that would evaluate options and help prioritize or automate operational responses, though Target has not given a timeline for that.
What it means in practice
For supply-chain and information-systems teams elsewhere, the more transferable point isn't the specific software, which is proprietary to Target, but the pattern: reusing your live system's real data and logic inside a simulation environment, rather than building a separate model that can drift out of sync with reality, is what let Target trust the Houston facility's simulated numbers enough to act on them before the building opened. That is a higher bar than most internal simulation tools clear, and it is the detail worth asking about before trusting any "digital twin" claim, whether it is built in-house or bought from a vendor: does the twin share the operational system's actual data and logic, or does it just resemble it.
