If you think supply-chain management is a boring topic, listen to one of Hau Lee’s talks. Within minutes, the professor of engineering and management science at Silicon Valley’s Stanford University has the audience roaring with laughter. When it comes to anecdotes illustrating the pitfalls of today’s supply chains, he has the delivery of a stand-up comedian.
One story he tells is about Volvo. In the mid-1990s, the Swedish car manufacturer found itself with excessive stocks of green cars. To move them along, the sales and marketing departments began offering attractive special deals, so green cars started to sell. But nobody had told the manufacturing department about the promotions. It noted the increase in sales, read it as a sign that consumers had started to like green, and ramped up production.
Another classic in Mr Lee’s repertoire is the “bullwhip effect”, named after the way the amplitude of a whip increases down its length—just as variations in orders tend to get amplified along the supply chain. Why is it, for instance, that Procter & Gamble has to deal with widely fluctuating orders for its nappies, when babies’ consumption is generally quite steady? The reason is that each retailer bases his orders on his own, slightly exaggerated, forecast, thus increasingly distorting the information about real consumer demand. This is one of the most important causes of inefficiency in a supply chain.
Yet Mr Lee is not just some academic out to poke fun at clumsy companies and economic peculiarities. He and his colleagues at the Stanford Global Supply Chain Management Forum are in business to help firms run their supply chains more efficiently and effectively. This is becoming increasingly difficult. Customers are getting more and more demanding, looking for a customised solution delivered within days, and no longer willing to accept a commodity product in weeks. At the same time, most firms’ manufacturing processes are becoming increasingly dispersed and global.
Few of them have even begun implementing the technology necessary to reduce the “bullwhip effect”, such as software to speed up the information exchange with their partners and collaborate on planning. But Mr Lee and his colleagues are already concentrating on the ultimate prize of supply-chain integration: ways of constantly monitoring and improving the whole system by using all the available data.
The start-ups spawned by this kind of research are anything but typical dotcoms, says Bruce Richardson, an analyst with AMR Research, an IT consultancy. By supplying algorithms to balance all the variables of a supply chain, such as product prices, stock levels and customer demand, they are following in the footsteps of i2, the market leader in supply-chain management. Other models include firms that developed automated trading programs once real-time data became available on Wall Street.
Rapt, another start-up, is perhaps the best example to show that optimising a supply chain can be as challenging as managing a financial portfolio. The firm has 20 PhDs on its staff and boasts a long list of patented algorithms. It needed every one of them to build its hugely complex software. Based on, among other things, historical data and forecasts, it answers pressing questions for many manufacturers. What is the most appropriate stock level for each component of a product? And would it be better to buy a part on the spot market or get a supplier to guarantee a certain quantity?
Why Algorithms Are Good for You
For firms in volatile high-tech markets in particular, the answers to such questions are well worth paying for. For example, Sun Microsystems, a computer maker, used to get its demand forecasts wrong by 40-60%. The lead time for some components can be 16 weeks, but customers want delivery within four weeks. Product shortages are very costly, but unsold products lose value quickly. Using Rapt’s software for some of its products, Sun says, it saved $15m and increased revenue by $30m in the first quarter of last year.
SeeCommerce goes for a different market: monitoring an existing supply chain and finding weak links. By gleaning data from all parties involved and making them accessible online, the start-up’s software lets firms check continuously how well their supply chain is performing—a task that most manufacturing managers still carry out manually, based on printed reports. The program also allows managers to work together to solve, say, a chronic problem of getting replacement parts for cars into the hands of mechanics.
This was exactly DaimlerChrysler’s problem. Its warehouses could not fulfil all of the 220,000 orders that come in every day, which meant that customers had to wait to get their cars repaired—or go elsewhere. To improve the balance of supply and demand, SeeCommerce linked DaimlerChrysler’s Mopar parts division with many of its suppliers and logistics providers, which allowed Mopar’s purchasing and inventory managers to see what was wrong and speed up the flow of parts. In the year 2000, the car maker says, the system saved it $7.2m in reduced stock and $10m through improved order fulfilment.
Another firm, Vigilance, concentrates more on real-time alerts. The firm wants managers to be able to respond immediately if there is a rush order from an important customer, or if a critical machine fails, and to use the information to improve the business process. Over time, the firm hopes it can help its customers automate their reaction to many of these “events” so that humans have to get involved only in truly exceptional circumstances.
PowerMarket is also in the business of what insiders call “supply-chain event management”. But its strength is in helping electronics firms to respond—again something that most firms still do manually, even if they spend billions on components. If, for instance, a computer maker receives an unexpected order, PowerMarket’s “business process module” lets the managers in charge find the best source for the parts and the easiest way to reschedule production.
As yet, these firms have a handful of customers at best, but Mr Lee has already set the next goal: demand-based management. Traditionally, supply-chain management has assumed that demand is fixed. This is a foolish premise, because firms can influence demand with instruments such as pricing, promotions or lead times. What is needed, argues Mr Lee, are programs to steer demand that are linked to supply-chain systems—not only to avoid mishaps such as Volvo’s green cars, but also to maximise a company’s overall profit.
Stanford engineering professors make a habit of backing their hunches by founding start-ups, and Mr Lee is no exception. One of the companies he co-founded is Nonstop Solutions, whose goal is to speed up inventory flows. Another is DemandTec, which is working on optimising all the prices in a retail store, which can number in the thousands. The firm feeds its software with hundreds of variables—scanner data, pay rates, storage costs, seasonal ups and downs—to tell the manager of a grocery store exactly what he should charge for a bottle of lemonade in order to maximise profit or revenue.
The Best of Both Worlds
But products to optimise both supply and demand are only now being developed. Rapt recently introduced a sell-side version of its software. PowerMarket’s platform can be extended to include this aspect. Manugistics, a leader in traditional supply-chain management, in December 2000 acquired Talus Solutions, a software firm specialising in yield management (ie, charging more where the market will bear it and less if demand is low) for hotels and airlines. Demand for such products will surge, predicts Mr Lee: demand-based management will be “the next battleground for competitiveness”.
As a true supply-chain evangelist, he can already point to a number of success stories. Seven-Eleven Japan, a chain of convenience stores, instructs its cashiers to record the sex and estimated age of each customer so it can set out its shelves in the most convenient way, he says. That is why beer can now be found right next to ladies’ stockings: the data showed that those stockings are bought mostly by men on their way home from work. Similarly, Zara, a Spanish clothing giant, uses sales data to introduce new products all the time, about 12,000 each year. Its supply chain is so flexible that the lead time from designing a new piece of clothing to selling it in the shops is only two or three weeks.
Both examples, however, are special cases. Seven-Eleven Japan does much of its data mining manually. Zara runs its own manufacturing facilities, which gives it more control over its supply chain. All this goes to show that solving the increasingly complex supply-chain equation is not just a question of good technology, but also of the right corporate culture.