Internet industry Operations: What is data analysis?

Intermediary transaction http://www.aliyun.com/zixun/aggregation/6858.html ">seo diagnose Taobao guest cloud host technology Hall

Data analysis The vocabulary is fashionable, but as with all the concepts that have been scrambled over the years, it's hard to have a good answer when you calm down and ask a lot of people to explain what data analysis is all about.

The more common answer is that data analysis is data analysis. So how do you analyze and analyze what? Obviously the answer does not answer the actual question. However, it is this vague situation that hangs over the industry, especially in the field of data analysis in the Internet industry. It seems that the demand for data analysis is constantly improving, but what to analyze, what to analyze, how to use the results of analysis, do not think clearly, I am afraid even thought.

In my opinion, data analysis is not a job, especially to get data from the background, to do a graph work, but a product that can satisfy a certain kind of actual work need. For example, the data index system is used to guide the operation, so that the operators can evaluate the gain or insufficiency of their work based on index, and further understand the reason of the exponential increase or decrease through data drilling. Well, that's a bit of a mouthful. For example, a third party: TV is a user's leisure products, the data analysis as a TV bar.

In data analysis, the most important thing is to clear the purpose of data analysis, just like the television example above, to make it clear what TV is used for, don't be surprised, play games, watch movies, watch the game, watch soap operas, sing KTV and so on, the use of TV is different. Therefore, the purpose of data analysis determines the different ways and means, the starting point is always how to guide the work, whether it is the most basic understanding of the status quo and trends, or machine automatic learning algorithm improvement, always.

Speaking of which, what is the use of data analysis of this "product"? Too much, too much to let people too easily lost, the data will make people's ambition soar, see indicator A, will think of indicator B, understand these, and hope to drill, this full of, although also valuable, but input and output is very unreasonable. Therefore, the data analysis of this product, to the user should be "knowledge", without conversion into knowledge, all the data is worthless. I suddenly told you today Fahrenheit 105 degrees, do you feel valuable?

When it comes to knowledge, the best way to switch is 6 words: graphics, contrast, drill-through. A picture wins thousand words, the indicator grows or reduces, compares with oneself, compares with the control group. When these changes are found, the dimensions are viewed in different levels, and which levels cause these changes. In fact, very simple, simple than training what year, chain, mean, public number, variance, Gaussian distribution, ANOVA, nonparametric statistics, factor torque also with a rotation, Bayesian distribution and so on and so on, and so much simpler.

It must be goal oriented, not tool oriented. The latter is terrible, I once met a classmate, he is very happy told me he will use SAS, I asked why, the answer is can be programmed. I said well, thinking that there are meters, how complex the model, in fact, this complex model is to describe the calculation of statistics. Forgetting tools, forgetting models, guiding your work with goals, assuming that the knowledge that you want to transform is what users are given, what their business scenario is supposed to be, whether they need to look at data to evaluate performance, or need data to improve their work, and so on. Then tell him the knowledge, and it's over. If you really want to let me recommend what tool, I said Sql\python\r\spss\excel pick a line, if not, feel free to find me.

So to sum up: data analysis, is the data into the knowledge of the product. So there should be no data analyst role, but product manager.