How to use data to guide business

About two months ago, Liu Yan, chief executive of Beijing-East, invited me to a trip to Beijing east to make an internal exchange with a roomful of people. Among them are the BI team, the marketing team, the PM and technical team, and the CTO Lee, who originally planned 1.5 hours of communication and finally 4 hours. Although it has been so long, I still want to share what can be made public, because these topics are representative.

Problem One: Data analysis please come to the math experts, but do not understand business, how to train them?

There are also many ph. D. In the Amazon, but if these doctors don't understand business, it's hard for them to use it. I was also not sensitive to business, this particular thanks to the work experience in the Dunhuang network, let me have the opportunity to train in Tsinghua University, and participate in the company's decision-making, 4 years down the business awareness of a lot. So, in my own experience, they have to be kept in touch with the company's actual business. For example, for the first one months, it would be better for them to answer the phone at the customer service department than to have them sit there.

Question two: Different departments have different BI team, marketing has one, the finance also has one, this kind of bi team scattered in different departments arrangement suitable unsuitable?

Although each department has a data team that helps each department understand its own situation, the company must also have a core, independent BI team.

There are two important reasons: First, the BI team has no interest in any other department, and the independent BI team is more conducive to impartial and independent analysis and research; second, the data is related to each other, and real data drivers need to connect the dots to the line or face. For example, why does the unit price of this month become smaller? It is possible that in addition to the high unit price of 3C products, the main push the price is lower than the daily necessities, but also may be the marketing department did a low price promotion ... This problem to look for bi, not just look at a department of data can analyze the answer, need to use linear data to see. Of course, this is only the most basic data analysis, if the rise to the use of data for the company to do strategic analysis, but also to the overall understanding of financial data, business data and user behavior data. Today, most electric companies lack a data architect, how much data they need, why they need the data, and there is no answer.

Moreover, it is particularly necessary to emphasize that a good bi team is good at asking a question: Is it possible ... BI team in the early stages of development, other departments to run what data to run what data, you go to help me see why today's buyers suddenly increase? But if the BI team stays at that level, it's just a machine running data, not a "staff" that drives the company's development. In the second phase, the BI team will think about it and ask if it's possible.

Question three: Why are data unreliable at times?

Many people only blame the data will deceive, very few people in the data analysis before, authentication to ask: Data from where? Is it accurate?

Now the biggest cause of inaccurate data is--not to remove interference and unreliable data, such as the biggest interference in behavioral data is the crawler, the second big interference is the staff click on their own, opponents click, and the 2nd is easy to ignore.

It's not surprising that you have to do cleanup before data, or use these unreliable data to determine how your site's product design works. That's why the BI team is using the technical team for reasons.

Question four: Must you use data to collect all customer information?

In fact, the user than we imagine willing to tell us more information, not necessarily everywhere to use data. One is to design a process to collect customer information, such as customers come in clearly can ask is a male or female, why to use behavioral data to see whether he is a male or female, the data can not play too badly. Second, the telephone direct communication, sometimes eight people divided into two groups of direct telephone to ask customers, and analysis of the results of the result is not much, sometimes the data should not be too stressed.

Question five: How to make an effective directory management?

I've spent 3 years researching directory classification management, but until now there are some ideas that still don't have a satisfying answer. I look forward to your comments.

Question six: From the customer's level, what is the difference between the traditional industry and the e-commerce industry?

The internet is wasteful, 100 people come in, only 2.5 to pay, which is a good site, how many people seriously think about improving conversion rate? And the shop downstairs to sell cigarettes, a person for three consecutive days to the fourth day did not come, the boss will be in mind.

Question seven: What impact does the logistics data have on the BoE trading data?

Logistics data does not belong to the front-end user behavior data and is not a back-end transaction data, temporarily not good conclusion.

Question eight: is the best buyer the most money?

No, measure customer value, in addition to from the purchase capacity of this dimension, but also should look at his social value in the network, such as some people although the total amount of purchase, but to more times than, he has many buyers in the network with countless contacts, can drive a lot of people come to buy things, So this customer is the core user of the platform.

SOURCE Address: http://blog.sina.com.cn/s ... 0100m4k6.html