How enterprise Big Data starts: from small data to big data

At present, there are many talk about large data at home and abroad, mostly talk about the operation level, or from the service side, the service side raised more. I want to communicate with you the problem is as a variety of enterprises, especially the client side of the enterprise, the big data and what their relationship, or as the enterprise side how to participate, this is the biggest problem facing the enterprise now.

The main answer to this question is that large data should start small. Because now many enterprises are facing the biggest problem is not how to use large data, but some of the internal small data integration problems, or small data are not good in the case of how to use large data. Large data should be gradually evolved from small data, is a normal ecology, not instantaneous change. The concept of large data is similar to the concept of media, which requires the enterprise to build itself, instead of relying on others from the beginning. Many companies talk about the media, like talking about other people's things. For example, when it came to the media, it was thought to be a platform offered by a third party, where everyone complained. Since the media is their own media, enterprises themselves to participate in. The same big data is not other people's large data, we assume that a third party provides a lot of data, there are a lot of information, CI, BI and many other modular things for us to use. If so, you have, competitors also have, you can get things, competitors can also get the case, it can not be called the core competitiveness. Large data as an enterprise to become its own competitiveness, enterprises must establish their own enterprise-level data.

To make a big data, you first need to know what your business is, or what the core of your industry is. We often find that many enterprises in the process of competition, ultimately not by the current competitors defeated, but many are not your competitors defeated. A very simple example, everyone thinks that Amazon is a dealer, but this is wrong, it is now the main revenue from the Cloud (cloud services). That is, companies need to find their core data (value), this is the most critical. Only on this basis, the establishment of their own large data is possible, to do some extension. Second, to find some internal peripheral data, to slowly grow it. A bit like a snowball, the first layer is the core, the second layer is peripheral related data. What's the third floor? is some structured data of external institutions. The fourth layer is socialized, as well as various now called unstructured data. These layers need to be found on a level-by-layer level, and to find something of value that is relevant to you. So that your big data can be built.

The first step is to find the core data. Core data now for many enterprises is actually CRM, its own user system, which is the most important. The second step, the peripheral data. For example, enterprises often online under the launch of some activities, in doing activities, the consumer information is simply provided in the form inside, or into the CRM system? The third step, regular channel data. For example, a company that sells fast-selling products can not get Wal-Mart data, Carrefour's data? Many foreign large data cases, said that consumers buy beer when they will buy a razor or something, or a mother and child product of the consumer she is buying this product today, indicating that she will inevitably buy another product. This has a preliminary excavation. These values how to come, this requires the enterprise to find the regular channel inside the data, with their own CRM, in order to do their next step to do marketing, promotion, product innovation and so on to establish a foundation. The fourth step, external social or unstructured data, is now called social media data. The main features of this information are unstructured and very large. What is the biggest value for an enterprise? Did you connect with your users when they spoke in social media? Here is a concept called DC (digital Connection). The so-called internet is actually a DC, but usually the kind of DC on the Internet is at the entertainment level. In the case of business, it is the enterprise must establish the DC relationship with consumers, its value can be played out. Otherwise, your data and a lot of CRM data are dead. Like the father of foreign CRM Paul Greenberg wrote four CRM related books, the first three are talking about databases, systems and so on. Fourth book, there is no more talk about those things, what to say? Talk about interactions, talk about DCs, and talk about how to build relationships with consumers.

With this database to do data mining, or in the process of building data, enterprises need to explore from what direction, nor aimless. The first thing to do is to follow your business, what is the problem with the business, or where is the main competitive point in the industry, which is critical. With this business relationship, then the assumption, that is, where the future competition point may be, big to the future of strategic competition, small to what. And then what to do next, these form a hypothesis, followed by some small sample test. Many companies look at large data is very scary, said I can not afford those big data, also can not afford so professional team, how to do? Do some small sample test yourself, even through the spreadsheet Excel can do data mining. It doesn't have to be that big, expensive data. Then the verification of large sample, the result can be applied to the reality.

One of the most important points in large data, especially in the internet age, is the failure warning. You find a rule that applies in reality, but you must set up some early warning indicators. That is, when the indicators reach the level of time, the previous discovery of the failure of the law, then you must find new, relevant, otherwise it will cause a waste. The author sees an article, which has an important conclusion. Everyone is saying that the value of large data is very useful, many companies say I accumulated how many TB, how many PB, but you based on the old data to come to a lot of conclusions is actually wasting your resources. You dig out a lot of data, many laws, if wrong, tomorrow press this to do, is wasted. Therefore, a failure warning is required. In such a process, ultimately you need to build an internal team, and their sensitivity to the data can be nurtured. It's worth it when you buy a big data service.

All of this work as a business needs to be done in-house, ultimately to bear fruit, there are some gains. Enterprise data start from small data.

Note 1: On the data mining process is directly up to large data, full data, or can be from small data, small samples, this aspect of the current debate. This paper is organized by Fudan University's "Big Data and Marketing Communication" Summit Forum, which only represents the author's point of view for reference.

(Responsible editor: The good of the Legacy)