The Complex Challenges of Designing an Automated Distribution Center

Considering a new automated distribution center? There can be many operational and financial rewards at the end, but it's smart to well understand the many complexities involved in facility design and approval that logistics managers and executives may face in getting there.

That was clearly one of the takeways in an excellent presentation this week by Dalen Mathys-Cook of on-line grocer Peapod and Paul Huppertz, a consultant at The Progress Group, which assisted the retailer with its design for a new DC to support the Northeast market. The two told their story this week at the Material Handling and Logistics Conference in Park City, UT.

Somewhat under the radar, Peapod has been successfully expanding its business in terms of both top line revenue and number of markets served. Peapod has well penetrated its home market of Chicago, and now also operates in the New York City and New Jersey areas as well as some areas of the Mid-Atlantic.

Its distribution and logistics requirements are challenging and complex. Customers place on-line orders across a very wide number of SKUs, and are given two-hour delivery windows for the following day. Delivery trucks of course have to be loaded by order and route, with a small picking and staging window (about four hours) to get the first wave of trucks loaded each night. (Mathys-Cook says Peapod is also experimenting with same-day delivery.)

That core complexity is exacerbated by the storage handling needs of different classes of products. Peapod carries larger, more bulky products such as pet food or charcoal all the way to frozen foods that must are subject to "cold chain" requirements from DC to customer delivery. It actually has several different areas for refrigerated space, because some produce stores better at 41 degrees while others are  best say at 36 degrees.

In fact, a Peapod DC might have some nine different storage and picking zones from which product must be merged to consolidate a customer order, each with its own handling characteristics, volumes and SKU profiles.

At its flagship Chicago distribution center, picking and replenishment processes are manual, with downstream sortation of picked cartons and totes to enable consolidation for loading.

In the New York-area market, however, picking is primarily today performed in the back of local Stop & Shop stores, which like Peabody are owned by Ahold, in an area Peapod calls "WareRooms."

When the company decided a few years ago that its strategy would be to significantly expand the number of SKUs it carried in additon to tthe overall growth expected in in the region, it became clear that the current fulfillment processes in the Northeast would no longer cut it. A new dedicated distribution center would be needed.

But how large should the facility be? Peapod logistics executives wanted to consider automating what they could in the new operation to improve throughput capabilities and reduce cost per unit in this notoriously low margin sector, but what was the right level of such automation? What would be the best places to invest assuming the capital budget was fixed, as it almost always is?

These of course are the types of questions every distribution automation project faces, but perhaps made even more challenging to answer given all of Peapod's requirements and constraints.

Peapod and The Progress Group decided to evaluate each pick area using a high/medium/low approach to material handling automation, exploring what the options were and how those would intersect with throughput goals, capital expenditures required, and ROI. The effort would also identify the pros and cons of each level of automation, and its impact on DC complexity, flexibility for the future, and ergonomics.

"As we added levels of automation, the question was can we gain enough incremental benefit to justify the investment," Huppertz said. 

One early challenge was that there was no real forecast for growth in the thousands of new SKUs expected to be added over time, so Huppertz said they just had to extrapolate from the data they did have on existing SKUs as best they could. He noted that they did have pretty good numbers on productivity rates for the more manual approach from both Peapod's Chicago DC as well as some Ahold facilities in Europe.

Obviously, estimates of rates, equipment costs, implementation costs and other data had to be developed for the medium and high options - no small task, especially when the effort involved multiple picking areas.

At a high level, the "low" option involved basically what Peapod was currently doing in its Chicago DC, picking into cartons and totes using a "pick and pass" type approach in any individual pick area. When complete, the carton/tote is taken by conveyor to a sorter for delivery to the proper staging area.

The medium level approach selected for more detailed analysis was a use of automated replenishment of pick faces using "multi-shuttle" technology from Dematic. Picking would be semi-automated using what Huppertz called a "put wall," where pickers place items into a series "cubby holes" by order, which when an order is complete are then manually moved into cartons or totes for delivery to the sorter.

Put Wall System from Dematic

The "high" option involved a full good-to-picker (GTP) approach, in which multi-shuttles would deliver SKUs for an order to pickers, where the picker would then place items into cartons by order, after which the multi-shuttle then whisks away to go to another picker station or back to storage, and the process continues.

Also considered in the high option was an automated buffer between picking and staging/loading to enable Peapod to pick further ahead of loading times.

The initial concept was to pick very fast movers from pallets, medium movers would get the medium level solution of auto-replenished and manual picking, and slow movers would use some goods-to-person and some of the put wall concept.

Budget Reality Sets In

Somewhere along the way, the initial budget for the figure was reduced, causing the team to go a bit back to the drawing board. Clearly, some of the automation would need to be cut, and the team had to look for areas where changes could be made that would have the least impact on productivity and throughput.

To accomplish that, they went back to the original ABC velocity analysis across zones and redid the same analysis for the big block of B movers, which were then broken into their own ABC sorting by volumes.

A simulation tool was also used to see test the full system in action.

As Huppertz said, "We had all the individual data elements for each area, but we needed to see it all together." 

The simulation looked, for example, at how long it would take a carton or tote to make it through the system to the loading area across various scenarios.

This analysis also allowed Peapod to see at what point in looking at the full spectrum of fast to slow movers the cost per unit for manual methods would cross over and be higher versus a more automated system (it was around the top 40% of total B mover SKUs).

Some things just had to be cut. One of those was using automation in the freezer area, which Mathys-Cook said was desired primarily to make the work easier for employees there and therefore improve retention. But in the end, with the constrained budget, freezer automation just couldn't make the ROI cut.

Then came another wrinkle - early estimates of implementation costs proved too low, as they were based more on national averages versus the specifics of the New Jersey market, where labor costs turned out to be much higher. That required still more "value engineering" to find other ways to reduce costs to fit the downsized CapEx budget.

In the end, the design chosen for fast movers somewhat more automated the process, using a more sophisticated conveyor routing of cartons/totes to individual pick zones, with a combination of pick-to-light and RF scanning, depending on the SKU.

The medium movers would go in traditional "pick tunnels," but those pick slots would be auto-replenished using the multi-shuttle system.

Hupperz said that means there is very little chance that a replenishment will be late and a picker would need to wait to select needed items for an order. This approach, however, means items have to be "decanted" from shipping boxes at receiving into the multi-shuttle totes.

The slow movers would not use the full GTP approach, but would use the semi-automated put wall system.

The new facility just broke ground in New Jersey, the concrete floor has been poured, and Mathys-Cook expects go-live in about nine months.

One important lesson learned, she added, was that it is critical to gain a deep understanding of how the budgeting and approval process works within your company - from both a capital appropriations perspective as well as the functional design of the system.

We hope to give you an update on the Peapod system a year from now to see how the reality has met the design goals.

Source: http://www.logistics-gr.com/