Big data doesn't contradict statistics.

In general, "opposition" is bound to emerge for new IT keywords. Recently, "Big Data" has become an object of attack, such as the "Big Data Failure theory" and other arguments have increased markedly.

The industry has great expectations for big data, as evidenced by the large number of http://www.aliyun.com/zixun/aggregation/14294.html "> Big Data seminars and exhibitions." Over the years, in addition to cloud computing wave, the lack of hot topic of the IT industry, large data is the long-awaited large keyword, perhaps large data will become the revitalization of the industry's energy.

At the same time, the Japanese government has proposed a new IT strategy-"the development of administrative data to the private sector in order to continuously create new business." In other words, how to effectively use data to promote business success has become a national strategy.

Although the author is neither a strong pro faction nor the opposition, but through the past experience in the interview, the difficulty of processing data has a sober understanding. What's more, it's difficult to relate to big data.

Many people around the author of large data also have a variety of views, put forward various issues. Of course, these are for the IT industry's readers, it is natural things, the author said this may be a swim. But it is these well-known truths that are often important to be reckoned with. Therefore, the following author will again put forward a large data "trap" to explore how to avoid the use of large data failure.

Do you really need a lot of data

First, it has to be clear whether there is a real need for large amounts of data.

In one event, a statistical analyst said of the big data: "How statistical analysis is to understand the whole of business through a small amount of sampling." For example, television ratings surveys are a typical example of a survey that uses a very small number of samples to master the national ratings of Japan. A large amount of data is not required if the purpose is clear. ”

The comments come as a surprise to the writer, who is now the expert in statistical analysis, the "data Scientist". That is to say, as long as there is a certain amount of data, the number of unrelated data, the results of the analysis will not be very different. If so, there is doubt as to what the big data is.

Hearing the above point, people feel that the contradictions of large data should be more than the author alone. The idea was that through large data analysis, it was expected to discover new things that had not been recognized in the past, but sometimes the results were just facts that had been known. If an enterprise invests billions of yen for system development, it is hard to accept that it is only a conclusion that proves the experience of senior staff.

That is why it is necessary to reconsider the question of why large data is needed. For example, companies need to explicitly target large data in advance by combining large amounts of data outside the enterprise, such as companies with transactions, and social media, for what purpose.

Is there any problem with the "quality" of the data?

2nd, who will maintain a large number of data, that is, the quality of the data can be guaranteed.

I have heard such a thing. The general manager of an enterprise receives the advertising (PR) magazine of the IT vendor who deals with it every month, but the recipient's title is not "general manager", but the title of "managing Director" when he was the CIO of the company. Although the title is wrong, but still can receive, so did not care too much. But when the general manager of the IT supplier came to the company for a courtesy call, he raised the idea of wanting to change the title.

The new selling point for the IT vendor is big data, and the company's general manager said on the spot that it would be revised immediately. At first, it was a little bit of a snap to the IT vendors running the big data business, and it was going to be corrected. However, wait until the next one months he received the PR magazine, found that the recipient's title is still "managing director." The general manager, through two PR magazines, felt as if he saw the status quo of large data, so he was very disappointed to say: "In the final analysis it vendors do not maintain customer database."

Although the above example is customer data, not just customer data, when it comes to large data, it is necessary to deal with many kinds of external data of enterprise. However, whether the data is the latest data, and how accurate the data is, the "quality" of the data is important. It would be meaningless to analyze data of unknown origin. If customer data cannot be maintained at any time, it will not produce any value. Should not be the original thought is Baoshan Big data, into a pile of rubbish mountain.

Ignore the work drive of the workers on the spot

3rd, enterprises should not only strive to cultivate data scientists, but also need to enhance the ability of on-site staff analysis of data. If the employee who is directly in touch with the customer at the top of the shop becomes "good at numbers", they can often think about things and judge by the data, so the enterprise will be strong.

For example, a shop-head salesman in a supermarket has been inspired by a conversation with customers to boost sales by buying new products or changing the way they are displayed. For example, employees responsible for sales on express trains find it seems that "smoking-ready coffee is selling", and when he collates the sales of different trains, it turns out to be true. The decision to concentrate on coffee in the smoking compartment resulted in a significant increase in sales of coffee.

Of course, increased sales through the field may be smaller than the sales figures obtained with large data, and their analytical capabilities are far less than those of data scientists. But even so, if you extend this way to other sites, the accumulated numbers will be significant. At the same time, the most important thing is that this approach can enhance the work force of the on-site staff.

In fact, after a retail company has been unified by its sales analysis by the head office, the shop head employees lose their drive, and even appear to be retired employees. This means that only relying on superior orders, it will reduce the field of professional ethics. As a result, the company decided to give the on-site staff the function of free analysis and judgment, and the store head regained its vigor. While large data is very important, if you focus on some departments, you can cause a loss of working power at the site.

The above three points are actually not only important for large data, but are also applicable to the entire information system. Large data is the IT industry's long-awaited keyword, in order to make it grow, it needs down-to-earth efforts, and should not be tied to its gorgeous part. Because of this, the author thinks that the above three points need to be remembered again.