Welcome to the 6th newsletter for 2026!
A quick note before we get started: If you are new here, welcome. I use this space to write through what I am seeing and learning across warehouse design and automation as the industry evolves.
The Automation Didn’t Fail. The Strategy Did.
I’ve been following these automated warehouse closures closely, not just reading the press releases, but talking to a few people who were in the room when the decisions got made.
What keeps surfacing isn’t that the automation failed.
It’s that the assumptions underneath it didn’t hold.
Over the past 18 months, we’ve watched a wave of automated fulfillment resets across grocery and retail: Kroger closing multiple Ocado CFCs, Sobeys shuttering Calgary, GIANT stepping away from an AutoStore hub, American Eagle winding down most Quiet Logistics sites, Zalando announcing the closure of a highly automated facility in Erfurt.
The headline version is simple: automation failed.
The closer-to-true version is harder:
The technology largely did what it was designed to do. The strategy wrapped around it didn’t hold.
When I look across these cases, I see four repeatable failure modes.
1) The Geography Problem: density assumptions collapse
Centralized automated fulfillment can be powerful if your market behaves like the market the model was designed for.
Ocado’s CFC model was engineered in an environment with:
high population density
compact delivery routes
scheduled delivery windows
predictable order patterns
Many North American deployments landed in a different reality:
sprawling geography
lower stop density per route
car-centric households
rising same-day expectations
At that point, the robots can be efficient and the economics can still break. No amount of robotic efficiency overcomes last-mile physics.
The tell is consistent: the CFCs that survive tend to be in the densest markets. The rest get pushed back toward store-based fulfillment and third-party delivery.
2) The Architecture Problem: storage density ≠ retrieval speed
Cube-based systems optimize a specific variable: storage density.
That tradeoff can bite when your operating requirement is fast retrieval under volatile demand especially when:
order timing is continuous (not batched)
baskets are smaller and more frequent
multi-temp consolidation has to happen quickly
peaks arrive as spikes, not predictable waves
In a cube/grid architecture, you don't always "add capacity" by adding robots the way people assume. Certain constraints are structural: grid access, dig depth, consolidation mechanics, and how quickly slotting can adapt to a shifting demand profile.
I've seen what happens when order patterns shift faster than the slotting algorithm can adapt. A SKU that was low-velocity yesterday becomes high-demand today because of a promotion or weather event. The system will move it to better positions but not instantly. That repositioning happens as background work, and during peak hours, there's tension between serving orders and optimizing for tomorrow's demand profile. Batched systems designed for predictable patterns can handle this. Real-time e-commerce doesn't always give you that luxury.
That's not a bug. It's architecture.
3) The Orchestration Problem: you own the asset, not the steering wheel
One of the least discussed risks is workflow adaptability.
When strategy shifts - channel mix, service promise, network priorities - you need the automation to behave differently.
If the orchestration layer is vendor-controlled, those changes often require professional services, timelines, and approvals. That creates drag exactly when flexibility matters most.
The robots can “work” while the business model becomes incompatible with how the system is allowed to run.
That’s how functioning automation becomes strategically brittle.
4) The M&A Problem: networks merge, automation gets stranded
Automation is rarely as portable as marketing suggests.
Even “mobile” robotics becomes sticky because it’s deeply tied to:
facility layout and storage design
WES/WMS integrations
workflows and labor models
infrastructure and safety constraints
When networks consolidate after acquisition, redundancy appears fast, often faster than the ROI horizon assumed when the automation was approved.
That’s how working automation becomes a stranded asset.
What this reveals
What’s striking is who this happened to.
These weren’t small operators experimenting at the edges. These were some of the most sophisticated retailers in the world - Kroger, Ahold Delhaize, Zalando - placing billion-dollar bets and discovering that model-market fit matters more than technical sophistication.
Even RaaS structures didn’t eliminate the risk. They shifted capital exposure, but they couldn’t hedge against density assumptions that didn’t materialize, demand volatility that exceeded design parameters, or network changes that stranded assets.
Financial engineering doesn’t fix strategic misalignment.
Three questions before your next automation bet
Does your market already have the order density to justify this model or are you betting it will appear?
Is the architecture optimized for the demand profile you actually have, not the one you planned for?
When strategy changes, can system behavior change without a vendor dependency cycle?
Bottom line
The lesson isn’t “don’t automate.”
It’s this:
Don’t automate based on what should work. Automate based on what already works in your operational reality.
Because the robots can do their job perfectly and you can still lose the bet - if that bet was placed on assumptions that looked solid on paper but couldn’t survive contact with the floor.
If you’re navigating one of these decisions right now, I’m always happy to compare notes.
-Parth
Sources: This analysis draws from public financial disclosures and corporate announcements including Kroger Q3 2024 earnings, Empire Q3 FY2026 results, Ahold Delhaize Q4 2024 report, American Eagle strategic update, and Zalando logistics announcement.
News
Thoughts? Questions? Feedback? Reply to this email.