Vocollect Archives

Voice Recognition Basics

Logistics Technologies | By Steve Banker • 06/16/2010

Joe Pajer, COO of Vocollect, and Tom Murray, Vice President of Product Management, recently briefed me on their idea that logistics is entering the “voice-enabled warehouse era.” According to a study they commissioned, voice solutions have only penetrated about 10-15 percent of the warehouses it is well suited for, a percentage that I can’t verify but find very plausible. In light of the low penetration, I realized that providing logistics professionals with some basic information about voice recognition technology might be beneficial.

What is Voice?

Voice Recognition is a form of Automatic Identification (AutoID). Voice can accomplish the same things that a traditional Warehouse Management System (WMS) can using Radio Frequency (RF) scanning. So, for example, a Voice system might tell a worker to go to location “XYZ” in the warehouse. The worker confirms the location by saying “XYZ” instead of scanning a location barcode. The Voice system might then say “Pick “3” and the selector then confirms the pick by saying “Put 3” instead of scanning the product bar codes.

Multi-modal solutions combine different forms of AutoID. For example, you might use Voice to verify that the correct number of items was picked from the right slot and an RF barcode scan to identify which lots those SKUs came from. In these cases, the warehouse worker would likely wear a gladiator style scanner on their forearm or an RF ring scanner to verify the lot identification number.

Why would you use Voice?

For many warehouse processes, Voice has a very good ROI. In a survey we conducted two and a half years ago, more than 83 percent of the respondents who had implemented Voice reported that their investment met their financial hurdle rate goals. The ROI is almost certainly better today, as the price of the hardware has fallen and the software has improved. This is also reflected by the fact that five years ago when I heard about Voice implementations, it was usually in warehouses where 50 or more order selectors worked per shift. More recently, I have heard of Voice being successfully implemented in warehouses with as few as 10 order selectors per shift.

The primary payback buckets of Voice are improved productivity and accuracy. Both are based on the ability of Voice users to keep their eyes on the task at hand without having to look down at a computer screen or reach for a holstered scanner. Most of the companies we surveyed were grossly underestimating how much mistakes were costing them. There is a much greater ROI associated with accuracy than most companies recognize.

While the picking accuracy of RF scanning is very high, it is even higher for Voice. When using a RF scanner for each picking, for example, a worker would do a scan to verify they are in the right location and then glance down at the computer screen to check how many items to pick from that slot. When they bring their eyes back up to do the pick, it is possible for them to pick from the adjoining slot.

Using Voice provides other benefits, as our survey found:

Benefits of Voice Recognition (Source: ARC Advisory Group; click to enlarge)

Which processes use Voice most often?

Voice is best suited for picking. All of the voice-enabled warehouses that I know of are using Voice for picking. Once Voice has been implemented for picking, some companies begin exploring using Voice for put-away, value added services, receiving, cycle counting, and truck building. In general, processes that are highly structured and repeatable are well suited for Voice; processes that cannot be rigidly defined are less well suited.

Units of measure (e.g., pallets, cases, or each) and the processes associated with them (e.g., receiving, put-away) also come into play. In our survey, we found double-digit productivity gains associated with each picking to a cart, each picking to a tote on a conveyor, or building pallets on a pallet jack or fork lift with mixed SKU cases. We found single-digit productivity gains from non-picking tasks (e.g., receiving, put-away). Voice did not have a strong value proposition for pallet in, pallet out warehouses.

When it comes to Voice solutions, there are interesting developments in the core technology, as well as the partner and vendor landscape. These are topics that I’ll discuss in a future posting.

Goldilocks and Warehouse Technologies for Food Distribution

Food & Beverage Industry | By Steve Banker • 11/04/2009

We all remember the story of Goldilocks and the 3 Bears, where one bed was too soft, one was too hard, and third was just right. When it comes to warehousing technologies for food distribution, I see the same thing.

In grocery, most store shipments are composed of mixed-SKU pallets. The layout of a grocery distribution center (DC) generally mirrors a store’s layout. So, if fruits and vegetables are found in the first aisle of a store, they would also be in the first few aisles of a DC. Food distribution centers are typically very large (half a million square feet at the low end, with most facilities at over 1 million square feet). Pallets at the DC are stacked in reserve storage above a pick location in vertical racks, and cases are picked from pallets on the floor.

In a warehouse that is “too soft,” worker tasks are directed by a warehouse management system (if a warehouse operation does not use a WMS and RF terminals, the company is not soft—it is in the Stone Age!). Selectors use RF guns to confirm picks, and they typically walk up an aisle, putting cases on a pallet jack, then walk down the next aisle (aisle traffic is generally one way). After snaking up and down a few aisles to build a pallet, they make the long journey to the shipping dock.

A warehouse that is “too hard” is almost totally automated. For example, Kroger implemented an automation system that can receive and put-away full pallets, as well as break down and rebuild pallets into store-ready mixed pallets according to how the products will be put away on store shelves. This all happens with almost no human intervention. Very cool, but also very, very risky. My guess is that the payback for this type of setup can’t be less than 10 years. Each DC replenishes a number of stores. However, stores are always closing and opening, mergers are common, and it is very conceivable, perhaps even likely, that the warehouse cost savings across the network are eaten up by high transportation costs. Further, this type of material handling is almost totally unsalvageable, as opposed to forklifts and other mobile solutions that can be used in new facilities when older ones are closed.

So, what is “just right?”

Currently, at the very least, it is WMS plus RF terminals plus Voice. Voice recognition from vendors like Vocollect for picking has good payback, often less than a year. “Just right” may also include Automatic Storage and Retrieval Systems (AS/RS) for full pallet put-away and pallet let down to the floor location for picking. I’ve come across unionized (high wage) food warehouses where the combination of voice and AS/RS had a payback of less than three years, while substantially lowering the costs-per-case picked. From my perspective, if you can get a payback of less than three years, the risk profile is acceptable.

But even in these warehouses, because they are so large, there is a lot of wasted labor that occurs when a selector travels with the pallet jack to and from the shipping docks. In existing food DCs, I believe “just right” will eventually also include the use of a new generation of intelligent (or robotic, if you prefer) automatic guided vehicles (AGVs). These AGVs will solve the empty travel issue and allow the selector to keep working. Seegrid and Kollmorgen offer these types of AGVs.

But what is “just right” today might be “too soft” tomorrow. Kiva Systems, for example, has developed robots that can retrieve a pallet and bring it to a worker who then builds mixed-SKU pallets at a station. The assembled pallets are then brought to the shipping dock by another robot.

However, I’m only half sold on this. These robots pick from pallets positioned on the floor that are not, and cannot be, stored vertically in racks. Because of this, the warehouse footprint required might be too large, unless it is built with mezzanine levels which include “bot lifts” (robot elevators). So, while the Kiva Robots have a good payback period (about two years) for small and medium-sized warehouses, I’m not convinced that it would be as good a solution when you add in bot lifts and mezzanine storage. But we should have some good proof points on this, either pro or con, within the next few years.

Bottom line: In warehousing, “just right” can have a very short half life.

Should Logistics Personnel Work in Retail Stores?

Inventory Management | By Steve Banker • 07/14/2009

I was briefed recently by Vocollect, a leading provider of voice recognition for warehouse picking. Vocollect is looking to expand from providing voice solutions for the warehouse to also providing solutions to support retail store operations. One of Vocollet’s clients, Cabela’s, a leading specialty retailer of hunting, fishing, camping and related outdoor merchandise, has logistics personnel working in its stores and they are responsible for managing inventory in the stockrooms and shelf replenishment. The Cabela’s implementation is also noteworthy because the company uses its warehouse management system (from Manhattan Associates) not just at its distribution centers, but also at its stores to manage stockroom inventory.

This strikes me as a very good idea. In college, I helped support myself by working in a grocery store where I was primarily responsible for the frozen food department. But when it got busy at the front of the store, I would often get pulled from my duties to help bag groceries. The store manager put more emphasis, perhaps rightly so, on keeping the check-out lines moving quickly than on keeping goods in stock on the shelves. This might make sense for a small-footprint store, but for larger stores, I think there is a good argument for having dedicated store inventory positions controlled by logistics.

Coincidentally, shortly after this visit from Vocollect, the Supply Chain Execution provider RedPrairie came in to brief me. RedPrairie is currently implementing a solution for a retailer that is doing something similar. This retailer has stores in busy metropolitan areas. These city stores don’t have sufficient storage, so there are warehouses in the city that serve as the stockrooms for the stores. These city warehouses are not distribution centers; each store has its own allocated inventory that is kept separate from the other inventory in the warehouse. The streets are too busy to make daytime deliveries, so they deliver the goods to the stores and stock the shelves at night. Similar to Cabela’s, the personnel running these retail storerooms/warehouses are managed by the logistics function.

Incidentally, a similar argument can be made for reverse logistics. The primary focus of most warehouse managers is to get that day’s orders out. Thus, if reverse logistics is co-located in the same warehouse, managers will likely pull people from the returns department to help with outbound execution if they fall behind schedule. Because of this, I believe reverse logistics operations should be housed separately, or if co-located, it should have its own dedicated manager and workers.

Finally, I think one advantage of having logistics in charge of store inventory fulfillment is that you could fairly implement a Key Performance Indicator for logistics linked to the percentage of time a store shelf is in stock. This is a key metric that drives store profitability. Currently, if problems exist with store-shelf replenishment and out-of-stocks, logistics blames store operations and store employees blame logistics.

A complete analysis of logistics trends, technologies, and services

Archives

Login