快猫短视频

COMPANIES hoping to boost their profits by using 鈥渄ata mining鈥� to extract
invaluable market insights from their records are likely to be disappointed,
according to a new study of the technology in action.

In principle, data mining looks like an obvious way for any company to boost
its profits. Sales of products typically generate vast amounts of data, showing
what is selling best, where, and to whom. Correctly analysed, such data can hold
the key to targeting products more effectively, or breaking into new
markets.

But the drudgery of extracting the information from paper records has proved
a strong disincentive. Hence the attraction of filing all that information in an
electronic 鈥渄ata warehouse鈥�, and leaving a computer to make sense of it all
using tools such as neural networks鈥攑rograms that mimic the human brain鈥檚
ability to find relationships in masses of data (see 鈥淧anning for data gold鈥�,
快猫短视频, 25 May 1996, p 30).

Data mining has been a resounding success in academic research, allowing
astronomers to find distant galaxies in sky surveys filled with points of light,
for example. But a Europe-wide study of hundreds of companies using data
warehouses, from OTR, a London-based consultanting firm, says that they are
getting little from the technique.*

The companies had each spent up to 拢6 million setting up their data
warehouses. Yet almost three-quarters of the companies failed to make any money
from their investment. And of those that did, most found that the returns were
considerably less than they had expected.

Colin Jackson, the report鈥檚 author, blames managers who think that clever
computer techniques can make sense of anything thrown at them, whatever the
quality of the raw data. 鈥淭he idea that computers can act like some black box
oracle is a universal dream people keep coming back to, with no success at all,鈥�
he says.

The problem, says Jackson, is that the sales information collected by most
companies is a rag-bag that even the most sophisticated data-mining techniques
can make little sense of. 鈥淭he companies that have used data warehouses
successfully are those that have used them on very specific problems, such as
British Airways, which uses them for seat allocation,鈥� he says.