The plight of today's supply chain manager could be fairly compared to that of Tantalus from Greek mythology. Trapped in the underworld and parked by a pool overhung with boughs laden with luscious fruit, Tantalus was doomed to spend eternity tortured by hunger and thirst in the midst of plenty. Each time he tried to drink, the pool drained away; each time he reached for a pomegranate or fig, the boughs receded. So it is for the average supply chain or distribution center manager yearning not for a sip of water or a pear, but for accurate data on the actual demand for the goods in his warehouse.
In theory, gathering demand data should be a matter of feeding all sales, tracking and inventory information gathered throughout the supply chain into a great ravening machine that links every party in the supply chain to every other party. But right now, there's a piece missing—the point of sale (POS) information gathered in retail outlets is almost never fed into that machine.
Why not? The main problem is that POS information is some of the least accurate you're likely to come across in the supply chain, answers Mark Johnson, vice president of marketing at G-Log, a vendor of supply chain management software in Shelton, Conn. "It's very nervous data, which is not very good for supply chain operations," he says. "The raw POS data still requires a lot of manual intervention before it can be digested into the supply chain."
Johnson gives this example: If a customer gets to the checkout counter and notices a defect in an item, often the clerk will swipe only the replacement item's bar code, although both items have come off the shelf. Multiply that over thousands of retail outlets over 90 days, and the result is a heavily distorted picture of stock on hand.
The other problem is that even perfect POS data will never be an absolute predictor of future demand. "Customers are notoriously fickle and their past demand patterns are less valuable in an era of rapid change in products, distribution and sales strategies," says John Fontanella, vice president of research at AMR Research in Boston, in a report titled The Demand Driven Supply Network: Striving for Supply Chain Transparency. For that reason, the data gathered as bar-coded items are swiped through the cashier's station will never be more than a part of the picture.
As good as it gets
Yet the fallibility of POS data hasn't discouraged Al Giunchi, director of distribution logistics at pet products manufacturer Hartz Mountain Group in Secaucus, N.J. For 10 years now, he's been extracting sales information from his company's main customer—Wal-Mart—and feeding it back into his own supply chain.
Every day, through the Internet, Hartz receives POS information on its products from thousands of Wal-Mart stores around the country and the 36 distribution centers that serve them. Through that mechanism,Hartz Mountain learns which products are selling, how much inventory is on hand in the individual stores, and what's available to top up the stock from nearby DCs. "We see inventory levels in stores and in the 36 DCs. We see what product needs replenishing and where that product is—on the East Coast or the West Coast.And it's in real time.You're looking at the product come off the shelf and out the store instead of out the DC," says Giunchi.
The Wal-Mart POS information isn't monitored directly by the logistics division. It's in the hands of a customer service team consisting of three people in Secaucus, and three in Bentonville, Ark., where Wal-Mart is headquartered. They, in turn, feed information about fluctuations in inventory levels and demand to the logistics group. When Giunchi wants to look at the data, he goes through a password-protected part of the World Wide Web (a step up from the early days when he used an EDI system).
Wal-Mart's sales forecasts tend to be almost uncannily accurate, says Giunchi. "Their computer system is second only to the government's. They know that a Wal-Mart in the Northeast is not going to need the same items as one in Arkansas. When 9/11 hit, they knew they'd sell more guns in Bentonville, Ark., than in Secaucus, N.J. Plus all of a sudden, there was a spike in gas can sales because people were hoarding gas.Wal-Mart knew all that. All that information started flowing through the system very quickly."
Responding to Wal-Mart's rapidly changing forecasts and constantly monitoring in-store inventory requires a lot of hard work, Giunchi says. "It takes a lot of maintenance, because there might be discontinued items or special promotions or delays for items coming in from, say, Asia or Brazil," he says. Returns, alone, occupy two members of the six-person team monitoring POS information. "The data does need scrutinizing and that's why you need six people looking at screens every day."
Aside from dirty data and shipment delays, the sheer size and nature of the consumer market means POS information is never going to allow anyone to stay exactly abreast of demand. "The problem is the vast number of variables in the system," says Giunchi. "You may think that a particular dog chew is going to knock people's socks off, but it doesn't. Or one product will unexpectedly take off and Wal-Mart will say 'I ordered 5,000 originally, but now I need 45,000 on the same day I wanted the 5,000'—and then the panic starts to set in. Or they order something in September and need it in time for Christmas," Giunchi continues. "So POS information helps maintain the flow to the stores of items that are already there. But it still doesn't help you if you're trying to push an item and you don't know if the customer is going to want it or not. There's no software in the world that's going to smooth that out."
All the same, Giunchi would welcome the opportunity to work with POS information from other major customers, instead of relying on vendor-managed inventory techniques, as Hartz Mountain does with Kmart,Walgreens and Winn-Dixie. Though popular, vendor-managed inventory programs, in which the products' supplier decides how much stock to put in the customer's distribution centers, don't get into the same detail as POS data. "VMI stops at the warehouse," says Giunchi.
Given the number of kinks that have yet to be worked out, it's no surprise that G-Log's Johnson says few companies are currently using POS data well. The ones that have mastered it include computer company Dell Inc. and Tesco, the British supermarket chain. Dell's selling structure, where customers order direct, typing their own information into a Web site, means its POS data are clean. Matters get a bit trickier when it comes to supermarket retail, where there are hundreds of thousands of SKUs to keep track of and more opportunities for mistakes. And it will be tougher yet to attain that level of sophistication in the retail sector.
Johnson says it's clear that feeding POS data into sales and manufacturing decisions works, because it's happening in industries like computer supply. But, in consumer retail, you're talking about adding an extra couple of zeros to the number of transactions, he says. "When you add dirty data, the complexity just takes off," Johnson says. "Transferring that into clean data and then translating it into orders that are digestible in the supply chain is a challenge, but it's not impossible. Absolutely not."
Despite the difficulties, there's still a lot to be said for feeding POS data into the system, Johnson adds. "The better the data you have, spanning the entire supply chain from factory to point of sale, the better you're able to reduce inventory and exposure to damage."
For Giunchi, the benefits of using POS information far outweigh the tribulations. "It gives us more intelligence. Whether we're able to perform with that intelligence is the key, and that's when we come into the real world," he says. "Planning and forecasting is so difficult. The weatherman doesn't get fired if he gets the weather wrong—it's Mother Nature's fault. But we don't have Mother Nature to blame in the world of business."