Where do large data items start and what should be noted?

In this column, we'll explain how to choose your first http://www.aliyun.com/zixun/aggregation/13844.html "> Large data Project, staffing and planning." Our suggestions are derived from our many years of cooperation experience with a large number of customers from all walks of life. Instead of focusing on any specific technology, this series of articles examines organizational dynamics and lessons learned from the actual performance of these projects in the existing, often very busy IT infrastructures.

Each customer is unique, so use this series as a general guideline rather than a hard dogma. Please note that this series does not focus on general project management. Let's say you've been trained in project management, so we'll just look at the dynamics that can be applied to your large data projects.

Understand your main drivers

The first and most obvious question is "Why on earth do this"? You should provide a compelling use case that can be a competitive driver, a cost driver, or some other problem that has been identified, and the application of large data technologies is a key way to address these issues. Typical drivers include information types (for example, inefficient structured information sources) or information (IP log retention), but in some cases you need to determine exactly why this approach is used. One of the most important aspects to focus on is a significant ROI. That is, you should specify a value for the cost of a problem before you develop a solution plan. When calculating the cost of the problem, remember to include the cost of the work (including technical and human costs) to resolve the problem. You can then compare this to the "after" state to determine the overall value of the project. Now you must realize that the first phase of a project itself may not be able to achieve positive ROI. You must have the foresight to know the effectiveness of the project's entire use case. Perhaps, in adopting the second or third round of application technology, ROI can be positive.

One of our clients provides an example of such foresight, and the customer uses Infosphere Biginsights's natural language analysis feature to understand e-mail communications. Through biginsights analysis, this customer can identify customer satisfaction problems in advance, avoid these problems caused by customer dissatisfaction, resulting in the customer abandoned the enterprise.

At the outset of the project, our customers selected a set of data execution analyses that represented only one region of the country. After analyzing the data in the area, the result actually means that the ROI is positive. But we do not build models based on the positive ROI achieved in one region. The ROI plan for this project is based on running all e-mail messages across the entire system (rather than within a department) of the customer's coverage across the United States. By calibrating the ROI route and gaining recognition, you can ensure that everyone is focused on the practical goals.

Please note that we are not saying that simple experiments (first developing projects to understand new technologies) are not effective methods.

What we are referring to is that the experiment must be clearly differentiated from your first "business" project. Do not confuse the information learned in the experiment with the implementation of the project that will eventually be put into production.