Matt Kelly says brownfield packing can pay back under two years

A warehouse adds robotic picking, then automated packaging, and still stalls between the two. Orders move faster upstream, but the handoff to pack-out remains manual and the promised throughput never shows up. Speaking in a recent webinar, Matt Kelly and Joe Bradley used that kind of partial upgrade to make a broader point: brownfield projects often spend first on picking and sortation, then discover the real constraint at packing.

The timing has shifted in 2026. SupplyChainBrain reported in June 2026 that Kelly now sees packaging automation as a growing focus in brownfield modernization, with some service-model projects paying back in as little as one month and capital projects in under two years. On Mondi's site in September 2026, an article on e-commerce fulfillment carried a similar choke-point argument from Ranpak chief revenue officer Bryan Boatner, a sign that packaging has moved from a side topic to a current operating decision.

• 1. SupplyChainBrain reported in June 2026 that brownfield packaging automation can pay back in as little as one month on service models and under two years on capital buys.
• 2. Bradley and Kelly argue BoxLast lowers brownfield software risk because the machine can work from a license plate ID instead of preloading weights and dimensions.
• 3. Sparck's published benchmarks, up to 20 manual stations replaced and one box every 3 seconds, show why operators should compare throughput claims by workflow, staffing and machine scope.

Packing is where upstream gains often stall

In the webinar, Kelly and Bradley described the same pattern in operating terms: a site improves picking, reduces walking and raises upstream flow, then runs into pack benches where cartons still have to be selected, erected, filled, taped, labeled and checked by hand. The result is not a failed picking project, but a system that shifted labor and delay to the last manual step before shipping.

Packsize is not alone in that view. On Mondi's site in September 2026, an article about packaging automation said earlier spending on picking, sorting and transport software has pushed packaging into a larger role in fulfillment performance. Packaging repeatedly shows up as one of the next constraints. On its site, DEVELOP describes a manufacturer whose manual packing station stopped the rest of the line when staffing ran thin, a different setting but the same failure mode.

That matters because brownfield projects are judged on total flow, not on the local success of one machine. A faster pick module can prove its own labor case and still leave the building short of the promised throughput if orders back up before sealing and shipping. No widely available industry study ranks packaging against put walls, sortation or deconsolidation as the top post-pick bottleneck, so the firmer read is narrower: once upstream automation improves, packaging repeatedly reappears as a limiting step.

Box Last changes the interface, not just the machine

Bradley says BoxLast needs only a license plate ID. In warehouse terms, that license plate is a unique order identifier, the barcode or number that tells software which order is at the machine. Bradley's description is simple: the system receives that identifier, prints the shipping label, applies it to the carton and automates what a manual packer already does when they scan an order at a bench.

That is why Bradley and Kelly frame BoxLast as a brownfield fit. A warehouse management system, or WMS, holds the order record; a warehouse control system, or WCS, directs conveyors and equipment; a warehouse execution system, or WES, coordinates work across people and machines. Bradley's argument is that BoxLast does not need those systems rebuilt around product weights and dimensions before the order reaches the pack line, because the carton is built after the order is already picked.

The contrast with Box First is mechanical as much as software driven. In Box First, the carton is made before picking starts, based on what the system expects the order to contain, so order sequencing and data quality matter earlier in the flow. In Box Last, Kelly said, the order is already on the belt and the machine builds the box around what is actually there in real time.

Packsize's own product descriptions help explain the workflow tradeoff. On its site, the company says the X5 supports both Box First and Box Last methods and can produce up to 600 boxes per hour, or one every six seconds. That does not make every Box Last project simple, but it shows why the integration question is different from the usual upstream automation problem of predicting carton size before the order is complete.

Packsize does not publish a BoxLast machine price in the material reviewed, and the available coverage does not map the exact interface into common WMS, WCS or WES platforms. No reviewed coverage offers a comparable benchmark for Packiro. What the market does show is the design principle: BoxLast is trying to minimize the amount of new information that has to move into the packaging subsystem before the order arrives. For a brownfield operator, that can mean a smaller software project even when the conveyor, induction and exception paths still need detailed planning.

The business case rests on labor, material and floor space

Kelly put capital payback at under two years in June 2026. SupplyChainBrain reported that he said some service-model projects can pay back in as little as one month, while capital purchases typically pay off in under two years. Those are broad benchmarks, not a site-level pro forma, but they tell operators where to look first: labor at the benches, corrugate use, shipping volume and the space tied up in box inventory.

Bradley's floor-space argument is more specific. In the webinar, he described one particular warehouse that holds six months of box stock on site. That stock is not just cardboard. It is racking, replenishment labor, working capital and obsolescence spread across multiple box SKUs, with SKU meaning a distinct stock keeping unit, in this case one box size or style that has to be stored and managed separately.

On-demand packaging changes that inventory structure. Instead of storing a long list of finished cartons in many sizes, the site stores a smaller set of raw corrugate inputs and makes the shipping box at the point of need. Kelly and Bradley treat that as a space-release strategy for brownfield buildings that say they have no room left, because the first place to look is often the area now consumed by pre-made cartons and the aisles required to replenish them.

Kelly's 10-station figure came as a ballpark in speech, so it works best as a planning estimate, not as a universal benchmark. Even so, it clarifies the labor comparison buyers need to make. The real choice is rarely machine versus one packer. It is machine, floor space and raw-material storage versus a bank of manual benches, their staffing across shifts, the box inventory that feeds them and the downstream cost of shipping more air when carton sizes are fixed in advance.

Other fit-to-size systems set a harder comparison

Supply Chain 24/7 reported that Sparck says it can make a box every 3 seconds with two operators. The March 2023 article said Sparck Technologies' CVP Everest and CVP Impack can replace up to 20 manual packing stations, reduce packing labor by up to 88 percent and cut shipping volume by up to 50 percent. In a product announcement on Material Handling 24/7 in October 2022, Sparck said CVP Everest measures, constructs, seals, weighs and labels single- and multi-item orders in one flow.

That process description matters because it answers the reader's first comparison question: compared with what? Sparck's machine scope includes an inline scale and a 3D scan of the order to determine the minimum box size, then label creation that the company describes as carrier compliant, meaning the printed label meets the parcel carrier's formatting and scanning requirements for shipment. The headline throughput number, one box every 3 seconds, sits inside that wider machine sequence.

Packsize's site benchmark is slower and different in emphasis. The company says the X5 can produce up to 600 boxes per hour, or one every six seconds, within a fully loaded footprint of 52 feet 4 inches by 20 feet 5 inches and support both Box First and Box Last methods. The comparison is useful, but it is not a like-for-like test, because Sparck describes one machine scope and staffing model while Kelly and Bradley are making a brownfield fit argument that puts more weight on phased installation, upstream compatibility and removal of finished box inventory.

That is why throughput alone is not enough. A site choosing between fit-to-size systems still has to ask where the box is formed, what data the machine needs before induction, how many operators the benchmark assumes and whether the bottleneck sits at packing itself or at the transfer step that feeds it. Vendor benchmarks can frame those questions, but the site still has to answer them.

The handoff is the brownfield failure point

DEVELOP sized one packaging cell for a 4.2-second production rhythm. On its site, the integrator describes a manufacturer whose manual packing station was stopping the rest of the line when staffing was short, so it built an end-of-line system inside a 21.5 by 17.5 foot safety cage with robotic case packing, case erecting, sealing, labeling and palletizing. The operational lesson was plain: packaging had to match upstream pace or the rest of the investment would idle behind it.

Bradley's UK example from the webinar makes the same point in warehouse terms. He described a site that added robotic picking and automated packaging but never solved the middle step between them. That middle step can be deconsolidation, which breaks a tote or batch back into order-level units, or a put wall, a bank of cubbies used to sort picked items into complete customer orders before they move to packing.

Exotec's September 2026 article on its site shows why that handoff becomes a project risk. The vendor said brownfield projects slip when success criteria stay open, interface ownership is unclear and staffing or cutover decisions are left too late. Cutover is the moment live work moves from the old process to the new one, and Exotec's point is that brownfield teams need one owner for each interface and a dated transition plan before they assume a modular machine will be easy to install.

That is the real complication in Kelly and Bradley's thesis. A modular packaging machine can reduce the size of the packaging project, but it does not erase the need to decide who owns conveyor logic, message timing, exception handling and labor during ramp-up. Brownfield packaging works best when the software boundary is small and the handoff design is explicit.

Peak planning decides whether phasing works

Kelly framed peak as three to five times average volume in posing the peak-season problem. Bradley added that some customers see items per order rise from 2.2 to 4.5 during peak, a shift that changes carton choice, induction rate and labor load at the end of the line. Those two numbers explain why packaging can look manageable in an average week and still fail first when promotions or holiday volume change the order profile.

Bradley's answer was to build for peak capability, then use off-peak periods to reduce reliance on labor or extra shifts. In the webinar, he also described sites that use rules in their WMS to divert some orders to automated packing, some to manual benches and some to Box First when that path fits the order better, which lets the same building manage different order shapes without rebuilding the whole flow at once.

The urgency behind that argument has sharpened over the last year. SupplyChainBrain's June 2026 report framed packaging automation as a live brownfield investment question, and Mondi's September 2026 article treated packaging as a determining part of fulfillment performance after earlier automation waves. Bradley's point about Brexit reducing available fulfillment sites in the UK still explains why operators stay in older buildings, but the nearer-term signal is simpler: more teams are trying to pull another growth cycle out of the sites they already have.

That changes the next decision for an operations leader planning phase two. The first question is no longer whether packaging can be automated. It is whether the site has mapped the transfer step, the label trigger, the box inventory burden and the peak order profile well enough to choose the right packaging path before the next upstream speed gain simply moves the bottleneck again.