What Cutter & Buck's second AutoStore warehouse in Texas tells cube storage buyers

Angeline Fernandes of ARC Advisory Group argues the number that matters most is two. A company's first automation project is built on simulations, reference customers and a projected ROI. Its second is built on its own numbers.

That difference matters to anyone reviewing a cube storage bid. A vendor's newest installation shows what a system should do. A repeat customer shows what it actually did after commissioning, through peak season and with real SKU growth, and then decided to buy again.

The building is a new distribution center in Grand Prairie, Texas. It houses Cutter & Buck's second AutoStore goods-to-person system, and Kardex implemented it, as it did the first. Kardex says the two companies took what they learned at the first site in Renton, Washington, including several years of operating experience and performance data, and used it to design the new facility around Cutter & Buck's changing fulfillment needs.

Fernandes describes the change as a move from one question to another. The first asks whether to automate at all. The second asks how the next building should be automated, given everything learned from three years of running automation. She sees the second question as a much better sign of automation maturity.

The Renton problem the robots solved

Renton's story starts with people, not robots. During the pandemic, adding peak-season labor at the Washington fulfillment center was a real struggle. In 2020, and again at the late-2021 holiday peak, surging ecommerce sales ran into that constraint, and sales representatives, the CFO and CEO Joel Peter Freet ended up picking orders in the three-level pick module themselves, Modern Materials Handling reported in July 2024.

That order of operations is worth copying. Cutter & Buck weighed process fixes first, a new layout or a slimmer SKU range, and concluded neither would keep pace with growth. Several types of automation were then on the table, including mobile robots, before the team decided to replace the pick module with high-density robotic storage.

The business behind that decision is not a simple one. The Seattle-based company sells active sportswear and lifestyle apparel across channels, holds fan-apparel partnerships or licenses with the NFL, MLB and PGA of America, and customizes apparel with logos for corporate clients and organizations. It runs three distribution centers in North America, but Renton does the custom work, and embroidery jobs drive close to 50% of the picking there.

That gives Renton two picking streams, according to Scott Sumpter, Cutter & Buck's executive director for distribution, enhancement and logistics. One fulfills non-customized apparel shipped to online buyers or to retail outlets such as fan and pro shops. The other keeps the embroidery area supplied with the right SKUs and sizes, close to just in time.

The labor figure may deserve the closest look. Sumpter said the aim was for a smaller team to handle order picking so more workers could be trained and shifted to embroidery or value-added services. Kardex has also described the 60% cut as a reduction in peak-season labor hours, in a June 2026 white paper. Either way, this is the vendor's own reporting. Still, it is the right kind of number for a buyer to ask about.

Fernandes makes the same point as an analyst: a warehouse can look highly automated and still not be highly productive. In her view, the business case depends on how automation changes cost per order, service levels, capacity, accuracy, order cycle time and labor requirements. Renton's published results address labor, cycle time and throughput. The other items on her list make a ready agenda for a reference call.

More bins, fewer robots

Put the two sites side by side and the second clearly isn't a photocopy of the first. Texas starts with more storage and more ports than Renton launched with, but with fewer robots.

Renton later added another 9,962 storage bins, and that expansion also brought more ports and infrastructure. The published material doesn't explain Grand Prairie's lower robot count, so buyers shouldn't read it as a rule for their own sites.

For operations teams weighing cube storage proposals that cite total system throughput, dividing that throughput by port count could give a more useful basis for comparison. At Grand Prairie's design rate, that works out to roughly 97 bin presentations per port per hour.

From automated storage to automated flow

The design also covers what happens once items leave the port. At Grand Prairie, conveyors carry AutoStore output to embroidery, order routing, Print & Apply, and shipping. Fernandes describes this as a shift from automated storage to automated flow. For a company that customizes apparel, the handoff to embroidery is where time is either lost or saved.

That handoff is where the second site looks most different on paper. At Renton, as MMH photographed it in 2024, apparel headed for customization was picked to totes and moved by cart to the embroidery department, while conveyor carried outbound orders to shipping. Texas puts the embroidery leg on conveyor as well. With close to half of Renton's picks feeding embroidery, the change suggests the lessons Kardex cites included where the manual moves sat.

Fernandes made a related argument in a separate November 2026 analysis of robotic truck unloading. She expects the next battleground in warehouse automation to be the physical handoffs between automated systems, such as feeding conveyors and transferring goods from one process to another. Grand Prairie's link from storage to embroidery is a small, concrete version of that idea.

Software holds that flow together. Kardex introduced FulfillX at MODEX 2024 as a warehouse execution system built specifically for AutoStore, and claimed it lets new systems ramp up and meet or exceed their business cases in as little as six months. That's a vendor claim, but a testable one. A reference customer can tell you how long its own ramp actually took.

Planning for demand nobody can forecast

Renton shows why the ability to expand matters. Its 2023 configuration didn't stay fixed. The bin count grew by nearly 10,000 as Cutter & Buck needed more capacity. Fernandes's point is that forecasting volume, SKU count and peak shape over a full investment horizon is the hardest part of an automation business case. A modular system turns the question from "did we forecast right?" into "can we add capacity when the forecast turns out wrong?"

Grand Prairie is built with that second question in mind, since its robots and ports can each be expanded on their own as demand changes. Fernandes writes that a system sized exactly for current demand could turn out to be the wrong size five years later. If capacity can be added later, the starting configuration may be a smaller bet than its numbers suggest.

The best reference a cube storage vendor can offer is a customer who ran the first system for several years and then ordered a second one.

There's a limit to how far one repeat order travels. Cutter & Buck's profile is unusual, built around custom embroidery work, so its reasons to buy again may not apply to a pure ecommerce or spare-parts operation. Broader survey data is mixed, too. In one 2026 study, 74% of organizations already using robots said they met their business goals, but cost and time to value were less clear-cut. One buyer coming back proves the system worked for that buyer. It does not prove it will work for yours.

That study, the 2026 Intralogistics Robotics Study, collected responses from 166 professionals in warehousing, distribution and manufacturing who take part in decisions about robotic automation. Fernandes puts the distinction this way: successful automation is not just automation that works, but automation that keeps making economic sense as the operation changes. Meeting goals and meeting them at an acceptable cost are two separate tests.

Questions for the reference customer

For operators whose SKU count or peak volume is growing faster than their planning cycle, the Renton-to-Texas sequence suggests a better way to run reference checks. Fernandes recommends tracking how many of a supplier's existing customers expand, add robots or ports, or deploy in another building, because those decisions come after the sales pitch. In practice, that means asking a reference customer:

• What did you add after go-live: bins, ports, robots, and how much disruption did each addition cause?
• How did peak-season staffing compare with the headcount in the original business case?
• How long did it take to reach the rated throughput, and what was maintenance really costing by year two?
• Could the system absorb additional SKUs, and could throughput grow without redesigning the facility?
• If you built a second site, what did you change in the spec, and why?

Grand Prairie has a design rate of 1,256 bin presentations per hour on paper and 1,284 in simulation. The number that will matter is how close the building gets once real orders and a real peak run through it. That's the figure Cutter & Buck will use to judge the second system.