Application of the methodology of scientific research in the actual enterprise

Individual preliminary summarized the following five steps back will be updated according to the specific sound production practices. . .

1. Discover the real problem

"A half an acre of Fang Tang Kam open sky dark clouds hovering, and asked him what so clear, as there is water flowing." - Song Dynasty famous Confucian scholar, poet Zhu Xi's "concept of the book felt." Probably means: a large piece of half an acre pond calm like a mirror, the sky clouds glow on the water rippling, ask why the pond water so clear, not because there is a steady stream of water flows in. This poem shows that a poet's attitude toward research, whether it is done or doing scientific research technology, vitality is the most important.

The ever-changing Internet industry, technology updates fast iteration speed generations of practitioners have put forward a very high demand, once a slow, lazy, lazy, and just as likely to be eliminated as Ebb Tide industry. Therefore, to achieve life-long learning, it is not easy to set their own limits, constantly iterative progress, grow and update their knowledge, is particularly important.

Generally, there are two things absorb new paths: a direct path, an indirect path. The so-called direct approach is hands-on. Deng said the good old Grandpa: practice is the sole criterion for testing truth. I have to say: The practice can effectively obtain first-hand information, but it also has some limitations. Because the practice is based on personal experience, easy to fall into the trap microscopic empiricism. Sometimes, when you think they have happily surprising discovery or invention, the results of a survey only to find someone else has already found and fixed a few hundred years ago, you do not say embarrassing embarrassment. Human life is short, it should be efficient to do meaningful things. Do not like the general bull stormed rattling, and finally nothing. Therefore, we must learn to leveraging, like tai chi inside to force the power, learn to use the knack, sometimes done skillfully deflected the question, a surprise move, the indirect approach is particularly important. Indirect way is standing on the shoulders of giants speaking problems. Do literature research, literature review of existing knowledge in a particular field from the macro to control, form their own knowledge, and then select a key, the key is to find the problem. Then dig deep, always dug mine.

2. Gather information and diagnose problems

Gather information, this is not to say it, grasping a lot of Internet resources, Google, Baidu, books, papers, patents. The key is how to classify, keen to capture the details, to seize the key clue, then boils to solve the case.

Analyze problems, to talk about the methodology still play a few:

1. Seven Questions Analysis: Reference my blog post:  https://www.cnblogs.com/Kevin0626/p/12507383.html

2. Hawkeye analysis: image point understanding is like an eagle occupy the high ground overlooking a global, first, on a question of the macro-control, forming a large image in the mind inside. In addition, analysis on the micro level but also with. Like old Mao Yeye said: despise the enemy strategically, the enemy tactically. Both have a strategic plan, but also have the tactical thinking. That decision called strategic global strategy, tactics is the way to solve local problems. This will be done, not only to see the forest, we see trees, crude and refined.

3. hypothesis

In the business world, almost all decisions are made "what if" information in limited circumstances, that is, first to make a speculative question to answer.
Good assumption, can be infinitely close to the correct answer, even is the answer. Research is constantly rely on a "hypothesis - verification hypothesis" knowledge accumulated. In essence, such a hypothesis, test hypotheses practice, in fact, small run, fast iterative way of thinking. Development: Popular local optimum and global optimum understanding:  https://www.cnblogs.com/joeman/archive/2013/03/14/2958824.html most important, a good assumption was based on facts. For a business research, the support of a hypothetical fact, do not look at the attitude, look at behavior. I put effective behavioral divided into three categories: 1. quantifiable behavior. We often say, listen to his words and deeds. On behalf of aspirations attitude, behavior determine the outcome. This is the first class facts: You can measure the behavior. 2. Do not tamper behavior. Sometimes just look at the results of behavior is not enough, you have to distinguish these data are not a true reflection of the course of action. 3. secretive behavior. Secret, immersed in a scene where the user behavior information is difficult to obtain large data or by questionnaire. But this information even if the sample a little less, it is still of great reference value. Quantifiable behavior, can not be tampered behavior, secretive behavior. A good assumption, certainly not escape the fact that these three types of support. The more types of facts collected, the higher the quality of assumptions. In many cases, a few good assumption is not valid based on the fact, will be able to settle down, also may be more than a good assumption. It must be a dynamic process, along with the fact that you need to constantly enrich and adjust to iteration. One pair need to use a very useful model Considerations: Logical longitudinal and transverse logic. The purpose of longitudinal logical deduction is: look at this hypothesis the fact is not rich enough to make the assumption timely correction. The deduction method is to ask yourself a few times: "It's that right there is no place does not make sense??"











The purpose of the lateral logical deduction is: look at the same facts based on the assumption that there is no more, let assume a more complete, more problem-solving ideas. The deduction method is to ask several times: "Just that right there is no other possibility??."
For the life of the general problem, it did have a basis in fact, did it again vertical and horizontal logical deduction. Even if you basically come to a good assumption. But for the business problem is not enough. You will also need two guides to help you judge can not be used as guide decision-making that assumption.
First: Your assumption is not within the boundaries of the cost? Good assumption must be the best solution under conditions of limited resources. Customers ask, how to increase brand awareness? You say, five times more advertising. This assumption is very good, but it must have crossed the border costs. Do not assume that cost considerations, then the full facts, and then derive rigorous, are correct nonsense.
Second: Your assumption is not able to export decisions action? Customers say, our products are selling well, how is it? If you say, because the competition is too fierce. This is not exported decisions action. If you say that because larger competitors logo, packaging colors brighter and more attractive, and this time he did not know what to do.

4. Verify hypothesis

The learning process is a hypothesis - verification process. That is, people under the existing experience and knowledge, for knowledge to be absorbed in learning, there is a hypothesis (anticipation), then by thinking and verification, card true: prove true hypothesis (new knowledge is absorbed); or falsification: proof false assumptions (adjust existing knowledge and experience, so that the new information to be accommodated).
Expand, the popular understanding of machine learning algorithm:

https://www.cnblogs.com/xiaojingang/articles/4344425.html

Say to product development, product development, the biggest risk is that users do not need to do a thing. In particular, it puts a few months (or even years) of time behind closed doors, only to find no one to make use of a product can not ignore that tragic. All project founders cherished a large or small vision, but the vision is actually a very suspicious thing, before the success, you can not know what your vision is an executable and the great ideals widespread demand, or a foolish fantasy. Steve Jobs ahead of the 2003 iPod for Apple released a few months because of no confidence in the stock gave up a lot of options, and the board of directors to negotiate in exchange for the option to perform a smaller number of lower prices; now invincible Tencent year also almost sold for several million RMB. But they actually have an entrepreneurial who generally lack characteristics: timely reflection summary.
The actual process of hypothesis testing should follow the following scheme:
1. research needs, the implementation of the abstractions portrait users, user history, etc. FIG.
2. prototype, the prototype such app, service blueprint
3. Design MVP, i.e., the best possible product, allows users to use
4. verify prototype, prototype verification records can be optimized in
5. continue iterations, improve your MPV

5. solve the problem

In order to solve the problem is the internal mechanism is based on the movement of things, seek an effective solution. Solution is the most critical methodological problem. By analyzing the problem and would lay the foundation for method study the problem for us. Qualitative factors involved, we can use text qualitative analysis. Quantitative factors involved, we have established econometric model, quantitative analysis. Following is a brief introduction of several methodologies:

1. Feynman learning: reference my blog article:  https://www.cnblogs.com/Kevin0626/p/12495784.html 2. Pyramid Principle: Reference "classic efficiency of thought and expression mental model - Pyramid Principle" : https://baijiahao.baidu.com/s?id=1618559373273845188&wfr=spider&for=pc 3. brainstorming: reference MBA think tank Wikipedia - brainstorming entries: https://wiki.mbalib.com/wiki/% E5% A5% A5% E6% 96% AF% E6% 9C% AC% E6% 99% BA% E5% 8A% 9B% E6% BF% 80% E5% 8A% B1% E6% B3% 95



 

Guess you like

Origin www.cnblogs.com/Kevin0626/p/12521642.html