11 errors that should be avoided during the implementation of the data warehouse, XIAOHUI@fudan.edu
1. Error project initiated:
On the implementation of the project manager on the data warehouse, there are two key characters to have an important impact on the implementation of the entire project. First, for more effective people who use corporate data and information technology, they can be the company's president, vice president of market, and Vice President of R & D. There should be a person called the Driver (Driver) between the project manager and the project sponsor, to control the progress of the project, and ensure that the project is moving in the right direction, this person should have the following three features: 1) People have been respectful in the implementation of other projects. 2) Hugging the technique. 3) Decisiveness, not old good people.
Table 1: The price that may pay when the data mining project lacks some key factors
The price of the problem is an IT supervisor rather than the business executive project outside IT looks like a technical test, not a commercial strategic investment sponsor budget. Due to insufficient funds, each encounter can lead to the entire project. Failure; will bring "troubleholders" and "budget dinosaurs". If there is no driving personnel, the data warehouse project manager directly launched a person who initiates people in the project soon, which will quickly and endless data definition, and there is no authority to coordinate all aspects. Driving personnel have not yet won the respect for various departments, because no one can guarantee the effectiveness of the data definition, people will not trust the data in the data warehouse, but no one will use it to protect the technology, not to doubt it project. It is seen as a technical test (or toy), and the business person is not used to use it to control the people's sorrow, and the rapid action project is slow. Management support softened, people's interest is reduced, people turn to find other solutions Method.
2. Develop an impossible goal:
Implementing data warehouses can be divided into two phases: First, to the company's high-rise lobby, convince them to implement your data warehouse plan. The second is to work hard to achieve the goal of the original promise after the budget. It is very likely that you have a big seaport in the original, such as "give me a million, you can make everyone you want very easily". " In fact, in many cases, data warehouses cannot provide all the data required by people, such as ancient historical details, may not be placed in a data warehouse.
3. ENGAGING IN POLITILY NAIVE BEHAVIOR:
Many data warehouse project managers will make typical mistakes in promoting data warehouses, "he can help manage people better formulate decisions", "When a self-esteem, the typical reflection is" " This guy believes that our decision is not good, and use his system to give us a patch, and then it is difficult to obtain these managers in the future implementation process.
4. Put all the data that can be found in the data warehouse:
Put many irrelevant data in the data warehouse, soon leading to huge databases, slow response, difficult to manage and maintain, so that the data warehouse is finally a Huawei, only in theory, and actually Difficult use.
5. It is considered that the design data warehouse is a design transaction database:
Since the purpose of the two is different, it should not be designed in the same database management software in designing the data warehouse, it is not designed according to the special needs of the data warehouse.
6. Select a technical rather than user-oriented person to do data warehouse project manager: Data warehouse is implemented in the process of dealing with people in various departments in the enterprise, and the data warehouse is also good for users to serve users, if it is only specially And the user's attitude is easy to cause the project to be terminated in the middle.
7. Only focus on traditional internal relational data, ignore external data, text, images, and even sound, video data:
A survey conducted by the White House in 50 major companies indicated that 95% of information used by senior management came from the company's external (news, friends, alliance), and the more high-level data on the company. . So good data warehouses should avoid the details of the high-level personnel directly to see a strip, but the summary data or a beautiful statement after statistical analysis, and combined with external data.
8. Delivery data with overlapping or conflict data:
The only fatal weakness of the data warehouse is that the data definition must be obtained, and it is easy to cause user confused and bored.
9. Believe in the commitment of soft-hardware suppliers to performance, capabilities and scalability:
The performance of each manufacturer's product commitment is usually obtained under the best conditions, or is completely exaggerated, and it is difficult to defeat in practical applications, so it is within the implementation process.
10. I believe that once the data warehouse is established and started, your task is completed:
Each user who brought benefits by using a data warehouse will continue to update the data in the data warehouse, which requires you to continue to adopt new technology and update equipment for faster response speeds. A typical error is to hand over the data warehouse to a project group, and it is considered that once the warehouse is established, the group can dissolve. After the data warehouse is completed, it needs to be carefully maintained for at least one year. The data warehouse should be regarded as a process rather than a goal.
11. Focus on dynamically generated queries, data mining and regular generation reports:
The typical process of using the data warehouse is 1) Put the data from all parts into the data warehouse, 2) Dynamic data queries, 3) Generate reports for commonly used query results. However, these reports are only truly utilized, and investment will truly benefit, while senior management people are very busy, they will soon send them tired of regularly send them in their hands, but will never go. A better way is to establish a hazard signal alarm system, define a variety of alert conditions in the data warehouse, and send information to a specific person only when these alert conditions are met.
