Big Data industry Investment: 1 dollars to earn 50 cents

September 24 News, according to foreign media reports, the survey results show that the recent market performance of large data was surprised. Partly because of the impact of media hype, many companies do not actually understand the value of large data industry on its investment. Then, as expected, a new analysis shows that companies receive much less in return for big data than they expect, or even less than their investment.

How much less? According to Wikibon's preliminary findings, the answer is surprisingly small.

Sure enough, Wikibon found that 46% of the big data investors struggled to sustain some of the success of their projects, while 2% of investors had to cancel their investments and end up in total failure.

But that does not mean that big data will inevitably end in failure.

One reason for the failure of Big data investments, Wikibon points out, is that "many companies invest in big data technologies like Hadoop without the precise and measurable business applications associated with the project." "They just hear the name" Big Data "and throw money at it, without thinking about what they really want to achieve. This is consistent with the results of Gartner's analysis.

For most businesses, large data equals Hadoop, and Hadoop is interpreted by IT staff as "an unmanned data dump." "Big data is often just a criteo of useless data," said Julien Simon Julien Simon, vice president of engineering at the company. ”

If you do not know the purpose of the data to store it, then this data will only increase noise and masking signals. The famous statistician Nette Slven (Nate Silver) assumes that:

If the number of messages increases to 2.5*1^18 bytes per day, the number of useful information is almost nil. Most of this data is noise, and the number of noise increases faster than the number of signals. There are too many hypotheses to be tested, the total data is increased too much, but the amount of the data is relatively stable.

In other words, adding more data is not a solution to the big data problem, but it is often the cause of the problem, which is why many companies receive very little in return for big data investments.

Wikibon's research shows that the best large data projects "are not being introduced into the IT world, but are used by business operations, which are often used in marketing or in small but strategic cases." "These companies will tap into big data experts and have a real expectation of what they can achieve with this technology."

These projects start out very small, but are then expanded on an initial basis.

In order to avoid spending unnecessarily heavily on large data projects, it is best to do so like the companies mentioned above. All the best large data science and technology resources are open, so before buying can be tested, and then according to their own needs to choose and develop the best technology.