Why warehouse automation projects fail

A documented, named case: Kroger and Ocado

The clearest, best-documented recent failure is US grocery retailer Kroger's partnership with Ocado. Kroger closed three Ocado-technology automated fulfilment centres, in Pleasant Prairie, Wisconsin; Frederick, Maryland; and Groveland, Florida, and disclosed, as reported by trade press in November 2025, that it expects roughly US$2.6 billion in impairments related to these closures and other underperforming automated warehouses. Ocado, for its part, is reported to be losing out on roughly US$50 million of revenue in the affected financial year, and is receiving US$250 million in compensation from Kroger under their agreement.

A named industry consultant, quoted in that same reporting, attributed the failure to poor site selection, insufficient order density to justify the fixed cost of the facilities, and a lack of clear strategic direction from Kroger leadership. That's one consultant's interpretation, worth stating as such, but the dollar figures themselves are company-disclosed facts, not speculation.

The pattern behind that case: order density and site selection

Kroger's case fits a broader, plausible pattern rather than standing alone: an automated facility is a fixed-cost bet on a level of order density and volume that has to actually materialise for the investment to pay back. Get the site wrong, or the demand forecast wrong, and the fixed cost is still there whether or not the volume shows up. That's a substantially different risk profile to a conventional warehouse, where under-utilised space is a much cheaper problem to carry than an under-utilised automated system, and it's a harder problem to grow out of through ordinary expansion planning once the structure and the automation are already committed.

When automation fails physically, not just financially

Kroger's case is a financial write-down, but automation failure isn't always confined to a balance sheet. Ocado's own Andover, UK warehouse suffered a major fire in February 2019, caused by a robot collision on its automated storage grid — the company's second robot-caused fire, with the earlier incident estimated to have cost over £100 million in damages and up to 400 jobs reported at risk. A further 2021 incident saw three robots collide on the same type of automated grid, again causing a fire and forcing order cancellations. Those incidents are part of why fire protection for automated storage is treated as its own engineering discipline rather than an afterthought.

What the evidence doesn't support

It's worth being honest about what wasn't found. Management consultancy McKinsey states that many warehouse automation projects fail to deliver expected results, citing lack of a cohesive vision, poor leadership understanding of the technology being bought, and misaligned incentives within the buying organisation, illustrated by one unnamed consumer-goods company that reportedly spent over US$150 million on a failed consolidation project. That's a useful qualitative pattern from a firm that also sells automation-strategy advisory work, so treat it as informed opinion rather than disinterested data — the same scepticism worth applying to any single vendor's labour-productivity uplift claim. No credible, sourced industry-wide automation failure rate was found in this research, and a page claiming one would be inventing precision that doesn't exist.

Claims circulating in integrator and vendor commentary about specific rework-cost percentages or per-hour downtime costs tied to vendor lock-in trace to single blog posts with no cited underlying study. They're plausible in direction, but this page won't repeat specific figures it can't stand behind.

What this means before committing to automation

The recurring, evidenced pattern isn't "automation doesn't work" — Australia already has genuine working examples, covered in Australian automated distribution centres — it's that automation converts a labour and capacity problem into a fixed-cost bet on demand actually showing up as forecast, in the specific location chosen. A fixed system also has to be sized against the demand it will actually see, including seasonal peaks, not just an average day. Getting the throughput, SKU and layout-stability questions right before committing, the same questions covered in when ASRS actually makes sense, matters more than any vendor's technology comparison. Whether a project that gets this wrong actually reverts to manual operation, or simply closes instead, is its own question — see whether de-automation is even a real thing for the honest answer.

Common questions

What's the most documented case of a warehouse automation project failing?

Kroger's closure of three Ocado-technology automated fulfilment centres in the US, with roughly US$2.6 billion in disclosed impairments reported in November 2025, is the clearest, best-documented recent case found in this research.

What actually caused the Kroger-Ocado failure?

A named industry consultant, quoted in trade reporting, pointed to poor site selection, insufficient order density to justify the facilities' fixed cost, and a lack of clear strategic direction from Kroger leadership. That's the consultant's interpretation of the cause; the financial figures are company-disclosed facts.

Is there an industry-wide failure rate for automation projects?

No credible, sourced industry-wide failure-rate figure was found in this research. Treat any specific percentage offered elsewhere with real scepticism.

Have automation failures ever caused physical safety incidents, not just financial losses?

Yes — Ocado's Andover, UK warehouse suffered fires in 2019 and again in 2021, both caused by robot collisions on its automated storage grid, with the earlier incident estimated at over £100 million in damages.

Read next

• when does ASRS make sense
• whether operators ever un-automate
• rack-supported clad-rack structure
• Superflat floor standard
• autonomous mobile robots
• goods-to-person picking