Coke's Voice-Powered Wireless Initiative

All manual-based processes are prone to human error. Even a conceivably simple task, such as picking a product off a warehouse shelf to match a customer order, can be fouled up by a human being. Just ask Coca-Cola Enterprises, Inc. (CCE). This bottler and distributor of Coke products experienced a marked decline in customer order accuracy as a result of human error in the warehouse picking process. CCE implemented a VoIP-based voice-activated picking solution to improve its order accuracy, streamline its picking processes, and increase employee productivity.

BRAND OVERLOAD HURTS ORDER ACCURACY
CCE maintains 64 bottling operations throughout North America, Belgium, France, Great Britain, Luxembourg, Monaco, and the Netherlands. These bottling facilities feed more than 430 distribution centers in these areas as well. Each distribution center contains a warehouse filled with hundreds of Coke products that must be handpicked and placed on pallets for delivery to fulfill the unique orders of CCE's customers (e.g. grocery stores, convenience stores, vending companies, etc.). Each CCE distribution center employs entry-level labor, known as pickers, to perform this process.

Traditionally, CCE's pickers would receive a printed customer order (known as a pick sheet) and would have to physically wander through the warehouse to locate the desired items (whether they be cases of Diet Coke, Sprite, Coca-Cola Classic, or other Coke products) and load them onto a pallet. This manual process was effective up until 2006 — the year that marked the beginning of Coke's brand explosion.

"For years, Coke only had about 150 distinct SKUs [stock keeping units]," says Mike Jacks, senior manager of logistics and transportation systems for CCE. "Picking from this limited number of brands was manageable for our employees. However, in 2006, trends in beverage drinking began to shift from your classic, sugar-based, carbonated beverages to flavored waters and teas, sports drinks, juices, and energy drinks. With all these new brands entering the mix, our warehouses are now pushing 400 SKUs."

This rapid increase in the number and variety of brands entering the warehouse was difficult for CCE's pickers to handle. For example, pickers saw the size of their average pick sheet grow from five or six distinct products to approximately 25. Plus, many of these products were new to the pickers — employees were unfamiliar with the packaging of the new products and weren't sure where new items were located in the warehouse. Combine this with the common distractions most pickers face (e.g. talking with coworkers), and it's easy to see how a picker could accidentally select 10 cases of lemon-flavored Dasani water, when a customer actually wanted 10 cases of raspberry-flavored Dasani water. Because of mistakes like this, CCE quickly saw its customer order accuracy rates plummet to 94%.
Now, a 94% accuracy rate might not sound all that bad on the surface. However, when you consider that Wal-Mart, CCE's largest customer, expects a minimum order accuracy rate of 99.8%, then the gravity of this accuracy shortcoming becomes apparent. This pain point becomes even more acute when you realize what CCE stood to lose by not meeting Wal-Mart's minimum order accuracy requirement.

"CCE has ASN [auto-ship notification] with Wal-Mart," says Jacks. "If we maintain a 99.8% order accuracy rate or higher as an ASN partner of Wal-Mart's, we receive preferential treatment when we make our deliveries. That means our trucks are allowed to dock ahead of other trucks, and our pallets aren't scrutinized by Wal-Mart dock workers. This can reduce the average time of delivery at each Wal-Mart by approximately 50% — from 50 minutes to 25 minutes. By not hitting 99.8% order accuracy, we stood to lose these benefits."

LEVERAGE EXISTING INVESTMENTS FOR NEW WIRELESS DEPLOYMENTS
To combat order accuracy problems as a result of picking errors, CCE executives decided to make implementing a voice-activated picking solution the company's primary technology initiative for 2007. A voice-activated picking solution provides a system where pickers receive customer orders over the phone, rather than a pick sheet. An automated teller then guides the picker through the warehouse to find the correct products, via a series of prompts and commands, and asks the picker to read the slot check-digit numbers of the product aloud to ensure they match the order. Voice-activated picking technologies have been used for years in the grocery and automobile industry; however, this type of solution was not common to the beverage industry.

CCE executives initially decided to pursue the project with a vendor that had a long history of providing traditional point solutions for voice-activated picking. These point solutions are traditionally carried over wired or dedicated wireless networks. However, CCE had already invested in a VoIP infrastructure for its call and sales centers in 2005. Jacks and other members of CCE's logistics team wanted to leverage this existing investment for the voice-activated picking solution. Conversations with traditional voice-activated picking solution vendors led CCE to believe that a VoIP-based solution was not possible. There were just too many challenges related to phone battery life and access point density. For example, CCE would need to provide its pickers with uninterrupted phone connectivity for 8 to 10 hours, and traditional voice-activated technology vendors didn't believe a VoIP phone existed that provided that kind of battery life. Furthermore, since pickers were on the move, the phone signal would have to jump from access point to access point without disconnecting. Plus, numerous pickers would often have to connect to the same access point simultaneously. It wasn't thought that a VoIP system could handle these requirements.

Just as CCE's logistics team was about to give up on its search for a VoIP-based voice-activated picking solution, Jacks spoke with Sudipta Bhattacharya, SAP's senior VP of supply chain at SAP's Logistics and Supply Chain Conference in early 2007. Bhattacharya suggested that CCE speak with Datria (see sidebar on this web page) because he knew the company specialized in voice-activated systems for mobile service dispatch and might be able to create a solution that worked with CCE's Cisco-based VoIP infrastructure.

SIDE-BY-SIDE PILOTS PROVE VoIP'S EFFECTIVENESS
Jacks met with Datria in May 2007, and in only six weeks Datria had a proof of concept system ready to deploy in CCE's pilot distribution center in Jacksonville, FL. The VoIP-based Datria proof of concept solution was tested alongside a proof of concept system developed by the point solution vendor CCE originally selected for the project. To accomplish this task, CCE simply had half of its Jacksonville pickers use one solution while the other half used the other. The company then compared the results of both proof of concept solutions.

Through the Datria proof of concept, CCE discovered that a wireless VoIP phone did exist that could provide 8 to 10 hours of continuous battery life — the Cisco 7921G Unified Wireless IP phone. Furthermore, with some tweaking, CCE eliminated VoIP-related access point issues by establishing a ratio that ensures the proper number of access points is used based on the number of pickers and the size of the warehouse.

Once these bugs were worked out, CCE realized that the results achieved by the VoIP-based voice-activated picking system and the point solution were almost identical. However, CCE didn't want to base its decision to pursue the VoIP solution on a single pilot. The company was worried that its success in one location might be a fluke. To ensure the VoIP-based picking solution actually worked, the company pitted the two proof of concept solutions against each other again in a subsequent pilot at the Fossil Creek, TX distribution center. The results of this pilot mirrored the first, and CCE decided to pursue only the Datria solution. This Datria solution was selected not only because it allowed CCE to leverage its existing VoIP infrastructure, but also because the hardware costs were lower.

"The traditional voice-activated picking solution we tested required pickers to use Windows CE handheld devices that cost approximately $1,500," says Jacks. "The VoIP-based voice-activated picking solution only requires a $400 cell phone for each picker. These devices are not only less expensive, but pickers are more familiar with cell phones than handheld computers, which helped to promote end user adoption."

The two pilot scenarios CCE implemented not only helped the company select the best solution, they also helped the company perfect its processes. For example, in both proof of concept solutions, CCE recognized that pickers often had to backtrack through the warehouse to locate the next item the voice-activated system prompted them to pick. This had a negative effect on employee productivity. As a result, CCE has standardized the pick path in all of its warehouses to coincide with the system. Now, pickers start at one location in the warehouse and are provided with a course through the warehouse that allows them to complete the order without backtracking, while minimizing overall travel distance.

VOICE-ACTIVATED PICKING COMBATS HUMAN ERROR
To date, CCE has deployed its VoIP-based voice-activated picking solution in 39 distribution centers with plans to roll out the solution to 61 more by the end of 2008. With the new solution, each picker logs on to a Cisco 7921G phone at the beginning of their shift and attaches it to a headset. The picker then says 'get work,' and the system randomly selects an order from the SAP system. A voice prompt then guides the picker through the warehouse to select the first item in the order. For example, the prompt might say, 'go to warehouse slot B 101 and pick 10 cases.' Once there, the system will prompt the picker to confirm the item and quantity picked are correct by reading the check digits aloud into the phone and saying 'check 241, picking 10 cases.' The voice recognition software ensures the numbers read match the order data contained in the SAP system. Once one item is confirmed, the system routes the picker to the next item, and so on, until the order is complete. If multiple products are stored within one bin, a picker can use the 'say material' prompt to hear the material description for the appropriate item to be picked, such as Coca-Cola Classic.

"Our voice-activated picking system has improved our order accuracy rates by nearly 5%," says Jacks. "We now maintain the 99.8% order accuracy rates mandated by Wal-Mart. One reason for this improvement is pickers are now in the 'picking zone.' Once that headset goes on and they are communicating with the system, the work doesn't stop. They know where they have to go, and they can't be distracted as easily by conversations with their friends. Furthermore, prior to voice-activated picking, we never had a confirmation system in place that verified that the items picked were correct. Our new solution ensures pickers confirm their selections while they're still in front of the pallet."

CCE has been so impressed with its voice-activated picking solution that it hopes to extend the solution to support other areas of the enterprise. For example, the company believes the solution could be valuable in addressing inventory control and fleet maintenance issues.
CCE's voice-activated picking solution is a prime example of how technology can be leveraged to help control manual processes and minimize the negative effects of human error. In CCE's case, human error was resulting in order inaccuracies that were damaging the company's reputation and standing with key clients. How much are human errors costing your business? Maybe it's time you took stock of your manual processes and found out.