The key to successful stock management could lie with parallel supercomputers, which can analyse Terabytes of on-line data almost immediately – a workload that would choke even the largest mainframe – judging by the experiences of several US retailers. Parallel processors can be used to indicate with hitherto unprecedented precision when to cut prices, which […]
The key to successful stock management could lie with parallel supercomputers, which can analyse Terabytes of on-line data almost immediately – a workload that would choke even the largest mainframe – judging by the experiences of several US retailers. Parallel processors can be used to indicate with hitherto unprecedented precision when to cut prices, which goods to buy and where to stock them and have already boosted the festive profits of distributors such as Kmart Corp, Wal-Mart Stores Inc and Dayton Hudson Corp, according to the Wall Street Journal. The success of these super numbercrunchers lies in their design – a parallel array of hundreds of thousands of small processors – which is a much more effective means of capturing and analysing bulky loads of sales information than the two to four large processors typical of most mainframes. They enable users to monitor every sale in every store – what was sold, when it was bought, its characteristics and price and so build accurate and detailed sales profiles. Hallmark Cards Inc, for instance, which sells novelty Dr Spock Christmas decorations, thought it was running out of supplies and was at the point of withdrawing its US advertising campaign when its new computer came up with a more precise picture. It showed that existing stocks from stores with slow sales just needed to be redistributed to those catering for a more concentrated Trekkie clientele. The supercomputer used by Kmart Corp has also proven an invaluable source for managers needing fast and accurate information about how particular items are selling, and where, and indicating when changes in display and pricing are required. For example, it demonstrated that, during September, Buster, the talking hobbyhorse, was selling at double the rate hobbyhorses had sold last year. This trend was identified much earlier than would normally have been the case, enabling buyers to double their orders in time for Christmas.
It has also proven most effective at targeting price cuts, saving up to 20% on markdowns compared with the previous system. This is particularly important for high-profit seasonal merchandise that must be tightly managed. The computer indicated, for example, that panic price-cutting in one store was not necessary just before Valentine’s Day as stores actually sold 50% of their Valentine stock in the last two days of the season, with 30% on the actual day. However, if the computer saw that a given store had 40% of its inventory left on the last day, managers could be alerted to cut prices or carry out other promotions. The computer has also unearthed some interesting demographic patterns. It demonstrated that confetti was in far greater demand at Easter at stores in Hispanic areas, due to a traditional ethnic festival celebrated with handfuls of the stuff; and that Christmas shopping begins earlier and peaks sooner in affluent suburbs. It has uncovered the inexplicable such as the popularity of a brand of night light and a halogen car headlamp at Christmas – items that were not even on the basic seasonal list. But most curious of all – it found that if someone in a midwestern city buys disposable nappies at 5pm, their most common next choice will be a sixpack of beer. Crisps sales were seen to rise 17% when they were put in a kiosk near the nappy aisle!