Two decades ago, the idea of talking to a computer—and it talking back to you—was more science fiction than reality. But as years went by, computers started getting smaller, lighter, and more mobile. They started communicating with each other over wireless radio frequency (RF) networks. And then, the icing on the cake, they started understanding the spoken word. Today, it's safe to say that voice has certainly become a viable technology for the real-time supply chain.

In theory, the technology is fairly straight forward. Voice-directed uses a combination of voice and speech recognition software to convert spoken word into data—and vice versa—capturing and relaying information between a mobile worker and a warehouse management system (WMS), creating a real-time experience.

Through this integration of systems, pick-list reading is eliminated and pickers are directed where to pick and how much to pick via voice commands transmitted over an RF network into a wearable computer equipped with a headset. Pickers then relay their audio response back to the system via a microphone through the same headset.

The benefits of voice-directed systems have been widely documented. Vendors are reporting that clients are doubling their cases picked per hour, eliminating language barriers that may have hindered productivity, while the portability of voice makes it easy to install in a new or expanded facility.

According to users, its eyes-free, hands-free operation has ushered in a new level of safety. “When you're reading pick labels you're oblivious to your environment, and there could be a lift truck barreling down on top of you,” says Eddie McHenry, group general manager for APL Logistics. “It's kept our people focused on their task, listening to what the system is telling them to do,” he says.

Despite these well-publicized advantages, many questions still revolve around voice. Will it really improve productivity? How accurate is it? Is standardization set? How reliable is the hardware/software? How broad is the functionality? Is training difficult?

To help shippers get a better idea of whether voice is the right way to go, we've gathered several voice technology experts and users in an attempt to answer these questions and expand on other issues driving the development of voice technology in the warehouse today. There are voices in the warehouse, should you listen?

No matter who you speak to on the subject, productivity and accuracy are clearly marked as the main drivers of voice application implementations. (See “Four Seasons keeps it fresh” and “ODW goes faster.”)

What's driving the continued push for productivity? Larry Sweeney, VP of product management for voice technology provider Vocollect, firmly believes that “distribution operations are using voice to increase throughput without increasing the physical footprint of their DCs or building new facilities.” And it's true that faster, voice-directed picks can increase a DC's capacity using existing resources without buying new equipment or physical expansion.

However, the market is witnessing continuous issues with accuracy, an area that users say still needs to be improved. Voice-directed operations typically use a randomly generated two-digit number, also known as a check digit, to confirm that a picker is at the correct location. Unfortunately, the correct product might not be in that location.

“A real-time WMS will allow you to detect and trace this error to a specific worker for re-training and counseling to minimize recurrence and maintain confidence that the correct product will always be in the correct location,” says Fernando Gonzalez, senior industrial engineer for Gross & Associates. “But without scanning the actual product, there is no way to absolutely confirm that the correct item was picked.”

To improve accuracy, voice-pickers may take the additional step of saying the last four digits of the serial number to confirm picks. Some voice-pick systems come with wearable wireless scanning devices when scanning the item is absolutely necessary. The general rule: If it takes longer to speak it than scan it, then you should scan it.

Integrating voice software with a WMS used to be the critical path in most voice installations, but today experts and users agree that Service Oriented Architecture (SOA) has made life with voice much easier.

To understand SOA, consider the following analogy. Without SOA, a DVD may work in your living room DVD player, but not in the family room player. With SOA, that DVD will work in both. “What we've been doing in the past few years is providing customers with options: Whose units do they want to use? Which system do they want to interface with? How do they want pickers to pick? In order to do this successfully, we use SOA to leverage RF devices from the world's leading suppliers, then plug in our voice solution software into their chosen devices,” says Rich Hayden, director of business development for voice technology provider Voxware.

“Implementation period has been reduced from months to weeks,” he adds.

Standardization also results in flexibility. “With the market moving towards a hybrid or multi-purpose unit,” observes David Price, senior operations consultant for WMS provider Motek. “A worker may use the device for receiving in the morning, using the RF unit's keyboard and scanner, and then use it for voice-picking in the afternoon by attaching a headset.”

From an operations point of view, the latest Web-based software allows transactions to be displayed on a web browser or on any browser-equipped terminal device. Supervisors can then track product and pickers from virtually anywhere.

Most systems are speaker dependent, where a voice template of the worker must be recorded before going live. But speaker independent systems that do not require voice templates are also available. “Now, a virtual stranger can come off the street and start picking orders,” says Marceline Absil, VP of marketing and sales for systems integrator and voice-system provider Inther Integrated Systems. “But for now, the system is limited to a few languages.” Most systems now know to pick up key words and ignore casual conversation.

Most systems use a digitized voice to speak computer commands. “We've developed one using a natural voice, male or female, depending on your preference. With warehouse workers being mostly men, the sound of the female voice is usually preferred.” adds Absil.

Today, speaker-dependent voice systems are offered in over 20 languages. But many DC's limit their language options to one or two. “It can get tricky trying to determine how to divide a limited number of units with multiple languages, when the workforce is constantly changing based on days off, turnover, etc.,” adds Gonzalez.

Today, the equipment is better equipped to withstand the demands of warehousing. Weighing less than a pound, Inther and Vocollect's proprietary voice devices are lightweight and quite rugged.

A person moving back and forth from a freezer to a hot shipping dock used to cause condensation in the equipment, but manufacturers have designed a whole line of headsets that are water tight.

“You can drop it in your pool overnight, shake it off and still use it the next day,” says Sweeney. Headsets are even doing double-duty as earmuffs in a freezer environment

According to the providers, noise is no longer an issue. A good voice system can take a measure of the noise signature in the environment with the push of a button. Some headsets have dual microphones; one to record a picker's responses, another to pick up surrounding noise and filter it in real-time for optimum voice recognition.

Although the general consensus is that voice is most effective in a medium-to-high volume case picking operation, it can also be used for receiving, loading, inventory or any operation that requires the worker to interact in real-time with the WMS.

Pickers may be asked to confirm inventory at a pick position if it goes below a certain quantity, also known as cycle counting. A picker's response can trigger the WMS to download a new task to another voice-equipped worker to replenish the pick position and ensure that product will always be available for picking.

Voice can also be used for batch picking using check digits assigned to an order tote. If you say the wrong check digit the system lets you know that you are picking to the wrong tote.

“Put-to-store cross docking operations is another great application for voice,” adds Hayden. “Voice pickers are directed to sort inbound products to its appropriate outbound stores or destination.”

With each functionality, however, additional programming may be required and script dialogues will need to be expanded and verified.

Voice has certainly become more affordable in recent years. If you include the headset, RF device, software, implementation, and training, voice-picking will cost from $4,000 to $8,000 per picker.

Users typically report a full return on their investment in less than a year. Here's more good news: As hardware becomes more generic and software becomes more standardized, experts agree that prices will keep dropping.

In less than a week, many users report achieving or exceeding productivity targets. “The best part of voice is how quickly people can be trained,” says Gonzalez.

Training may consist of giving new users a two to three hour presentation which includes an overview of the system, a review of the typical dialogue and creation of the voice template which lasts anywhere from 20 to 45 minutes.More importantly, new diagnostic tools are invaluable in debugging the pick process. These “debug” tools list each voice transaction in detail so supervisors can track a worker's actions and determinewhy a worker may be slower or less accurate as his co-workers.

“I see RFID in Voice's future,” says Vocollect's Sweeney. “RFID is very good for capturing data, while voice is very good for directing work,” he adds. By the simple act of picking an item, the RFID reader registers that it has picked the correct lot.

Motek's Price agrees, “Voice and RFID equal speed and accuracy, but for now RFID components are too expensive for the warehouse.”

Voice is not an end-all solution. It is merely a tool that can efficiently capture and relay information between a worker and a warehouse management system. But if the business system you use to process voice can't meet your goals, then it's all just a bunch of noise.