Jun 07, 2002June 7, 2002 – At a small stand on the crowded show floor at its Sapphire users conference in Orlando, five 200-ml bottles of Jack Daniels sat like beacons on a small shelf. Each bottle had a small RFID tag tapped to its bottom. Every time a passerby picked a bottle up, the movement was monitored by new software technology SAP is developing for Procter & Gamble and a retailer that didn't want to be named.
The demo was simple. Its implications are not. This condensed version of a prototype actually running in store today showed that it is possible to track and respond to changes in stock levels on store shelves. The system could dramatically reduce the number of times a retailer loses a sale because an item isn't on the shelf when the customer wanted to buy it.
In the demo, the consumption plan called for two of the five bottles to be "sold" on Wednesday and another on Thursday before the goods are replenished on Friday. The system was set up so that there has to be a safety level of one bottle at all times.
When one bottle is removed from the shelf on Monday, a reader built into the shelf reads the remaining tags and sends information to a server running something called Savant, software developed by the Auto-ID Center. Savant cleans the data coming from the shelves – eliminates duplicates for instance, if two readers pick up the same tag – and sends in on to Centauri.
Centauri is described by SAP as a distributed "geo-spacial database." Each store would have its own real-time database with inventory information. The smart inventory agent monitors the database and compares inventory levels against a plan. If inventory of a particular item gets too high or too low, it would alert staff.
In this case, the agent does nothing, because consumption is within the parameters of the plan. Even if three bottles are sold on Wednesday and Thursday, there will still be one item in stock.
But when another bottle is purchased on Tuesday, the agent sends an alert that the probability of an out of stock is high. That's because the system expects two bottles to be sold on Wednesday and another on Thursday. That's five bottles, meaning inventory would drop below the one-bottle safety level.
SAP's demo was highly simplified. Data from the shelf wasn't reconciled with point of sale data. Indeed, there was no point of sale. Just picking up an item was considered a "sale." And there was no back room, where inventory could be stored.
But the demo showed how the agent could monitor consumption and respond proactively to a potential out of stock. In the prototype SAP is running for P&G and the retailer, when shelf inventory dips below a predefined level, staff are alerted on handheld computers that stock needs to be brought out from the backroom.
When inventory at the back of the store drops below a certain level, because consumption exceeds the plan, the store's intelligent agent requests more product from the retailer's distribution center.
If there's no stock available, the agent can request product from the vendor's DC, and if there is none there, it can search regional DCs. The information can be fed into transportation applications so a shipment can be diverted to the store with the out of stock situation. This is what SAP means when it refers to "adaptive supply chain networks."
"The idea is to handle a local event at the local level," says Bob Betts, SAP's senior VP for global supply chain. "That's the cheapest way to do it because if you start having to change your manufacturing plan, that gets to be very expensive."
In a private demonstration for RFID Journal, Betts stressed that the technology is still a work in progress. SAP and its customers are learning how to take advantage of the real-time demand signals from shelves and designing systems that can respond appropriately.
But the system is far enough along to see how it can dramatically improve supply chain efficiency. By alerting the retailer and the manufacturer's advance planning and optimization software of a potential out of stock, the system can adjust to sudden changes in demand levels in real time – something that has never been possible before.
Say, for instance, demand surges and a shipment is needed from the vendor's distribution center, the DC will send the store information about how much product can be shipped, when and at what price. There might be several options. Maybe the vendor can supply five cases on Thursday for $25 each, or a whole pallet of Friday for $22 each. The store's intelligent agent can be optimized to choose the best option based on the plan.
When the new shipment arrives at the store, the tags on the goods are read as it passes through a barrier reader at the dock door. The arrival is automatically recorded, and the store's inventory is automatically updated.
"Today, deliveries are time-based," says Betts. "You ship product to the store on Monday and on Thursday. What we want to get to is delivery based on consumption in each store."
Can it work in practice? Can it scale? Not overnight, Betts concedes. The smart agent technology is complex because it uses software to take actions that have always been handled by people. SAP plans to take a step by step approach towards solving its customer's problems.
Betts also says that companies are going to have to get their own houses in order before they will be able to take advantage of this technology. "You have to have a high level of collaboration with your business partners," he says. "And you have to really be able to execute."
