The challenge of Enterprise management: unstructured data analysis

Beijing August 23 News, according to foreign media reports, supermarkets inside how to design to maximize the increase in sales? On the face of it, the problem seems to be not related to data scientists. Consumer behavior is hard to quantify: unpredictable and seemingly unfounded. Why do some shoppers spend more time in certain channels than others? Why does every shopper have a different route to walk in the store? Why do some products sell well in the morning, but not at all in the afternoon?

The answer to these questions depends on unstructured data analysis, which cannot be uniform into a database or spreadsheet as a result of the unstructured data. Most business information is in the form of unstructured data. The data may come from emails, memos, videos, customer calls, tweets, Facebook messages and blogs, which are inherently hard to quantify. Because of the unstructured, these data are also more difficult to analyze, but in fact, partly because of the growing interest in large data, the volume of unstructured data has recently soared.

For businesses, dealing with this mess of new information can be challenging, but also a huge opportunity. Most senior executives (58%) rely on unstructured data analysis to make business decisions, according to a recent joint study by Capgemini Consulting and The Economist Information Department. The more data an enterprise analyzes, the more valuable insights can be gained. Technical vendors talk about the hidden value of unstructured data. They believe that understanding unstructured data is the latest frontier for data analysis.

Find a way to get information from all of these conversations, PowerPoint presentations and tweets, and businesses can get real value from big data. But many suppliers start at the wrong place. Business projects should not start with accessing data, and the first priority should be to identify the questions to be answered. Who is my best customer? Why did the product fail? These are business issues. But technology providers are looking for technical questions such as: How do we tap data? How can we find useful data in a haystack? The answer.

The understanding of unstructured data is only partially related to search. The real value comes from analyzing unstructured data with other structured information. How do you do that? Think about how a company can take advantage of the knowledge of all its employees. There are millions of different interactions in an organization that contain a lot of valuable information, but how do you focus on the most relevant information and turn it into real insights?

The first step is to find out what you want to solve, which may determine the efficiency of the sales process. How do we use less resources to do sales? As you analyze the PB-level unstructured data, from e-mail, staff blogs to voice calls, you first have to identify the important structural elements in order to identify specific interactions that can explain the problem. But to determine who and who to talk to, you need to start with an accurate tagging of "active", "conversation", and "passive talker."

Make sure the active speaker knows its user name, login name, and other specific employee ID. If this information comes from the same person in the real world, it is no use getting data from 5 different people. Second, you must be able to understand the topic you are analyzing. To successfully analyze the content must be structured. Third, who is your client? Do you use the correct entry to get all the information about the customer? You must ensure that you have access to all references for each department, abbreviation, and holding company. such as Royal Bank of Scotland and National Westminster Bank? British Sky Radio, Sky or News Corporation?

The complete structured framework is then combined with unstructured information collected from a large number of different conversations in your chosen group. This may be the case: the companies with the biggest sales are also the companies that the company's employees talk about the most, in which case internal communication is benign. But again, people talking about the most customers don't bring in enough revenue, and in that case you'll find the key to inefficiency in the sales process.

At present, many vendors treat unstructured data as a separate technical challenge. However, unstructured data can only be understood when relevant important structured information is identified. If a staff member is responsible for a specific topic 95% communication, it will undoubtedly make him or her the core information Center for this issue. But if this pattern does not identify the underlying structured information that proves that various types of login information and identity information belong to the same person, you may not be able to see this fact.

If data is equivalent to noise, it is impossible to make an accurate decision. So how do we use unstructured data to design the best layout for supermarkets? Retail companies, theme parks and even police departments use group analysis to predict the possible response of populations in some situations. In supermarkets, analyze the video to determine the route of the shopper's Mad store, record where they have stopped, where to put the goods in the basket, and how to respond when they encounter people or dead ends.

Then combine unstructured information with structured data-for example, where some goods are positioned on the shelves or at the checkout counter-to create a complete data-driven diagram of the buyer's behavior. Once this information is in place, it is possible to forecast future sales based on various situations. Do you have to change the goat cheese position and buy more wine? Would people buy more expensive beers if cheaper beers were moved to the back of the aisle? If there is congestion near the bakery, will the supermarket sell more bread?

Using this structured and unstructured data combination, you can find answers to these questions that are the basis for successful predictive analysis. But only by establishing a complete and usable data graph can we make accurate decision.

(Responsible editor: The good of the Legacy)