Shulou(Shulou.com)06/01 Report--

Many novices are not very clear about how big data analysts improve data sensitivity. In order to help you solve this problem, the following editor will explain it in detail. People with this need can come and learn. I hope you can get something.

Have you ever been particularly envious and afraid of such a person who can see at a glance the abnormal data in your PPT, can come up with a data to prove that there is something wrong with your little conclusion, and then quickly overturn the conclusion of your entire report with a data problem? the result is that you have done a whole month of analysis, completely destroyed, and start all over again.

Such a person is extremely logical and has the power of life and death in your report, and the most important thing is that he has a strong data sensitivity.

So, what is data sensitivity?

The so-called data sensitivity is actually establishing the connection between numbers and business in the brain, and excellent data sensitivity is to be able to see at a glance the problems of the data and the possible reasons behind them.

What do you mean to see at a glance?

If you are in the game industry, I will tell you that the secondary retention of this MMORPG is 20%. You can know what level my product is in the industry and what problems may exist in the early stage of the game.

If you are in the O2O industry, I will tell you that takeout orders have dropped by 10% compared to yesterday. You can quickly determine the impact of the problem and the possible reasons for the decline.

If you are in the e-commerce industry, I will tell you that the repurchase rate of my product is 40%. You can quickly tell what kind of product my product is and what level it is in the industry.

To this extent, it is called excellent data sensitivity. How to improve data sensitivity?

Secret book: familiar with business

The basis of data sensitivity training is that you must be very familiar with the business, and countless speculations and verifications are useful and valuable experience.

Next, I will give ways to improve the data sensitivity of analysts according to the three performances of high data sensitivity.

How ⒈ can quickly determine whether the data is high or low or wrong: memorizing the large numbers of key indicators, observing trends, and keeping an eye on outliers

This quick judgment is based on the usual memorization and use of business data. Please believe that everyone's memory is good or bad, but as long as you make efforts to memorize the key indicators of the business and understand their basic rules, after a period of accumulation, you will feel confident when you look at the data.

There are also skills for memorizing data. You don't need to remember them all. You just need to write down the big numbers of key indicators, ignore decimals, get into the habit of reading reports every morning, observe trends, keep an eye on abnormal numbers, and look at more cases of how others analyze the causes of anomalies. Slowly, after you have a basic concept of the data, understand the business logic behind the data, so that you can quickly find outliers when reading the report and track them in time.

But for newcomers who want to change careers, or new graduates, what if they don't have access to specific business at this time?

For newcomers or people facing a career change, both types of people lack a general understanding of the industry, and the first thing to do is to memorize the data and remember the industry average data and the definition of common indicators for the industry. the purpose of this is to have an overall understanding of the industry as a whole.

For example, in the game industry, you can query the game data released by YingYongBao, 360and hardcore, and have an overall arrangement and memory of the retention and payment data of various types of games on the market, such as Arppu, Arpu, next-day retention, three-day retention, seven-day retention, monthly retention, payment rate, and so on.

For example, in the e-commerce industry, the formula of flow * conversion rate * customer unit price * repurchase rate is the top priority and so on.

⒉ knows where all the indicators come from, knows their meaning and their relationship with each other, and then determines the cause of the abnormal data.

When you increase your sensitivity, think about three questions:

Where did the ❖ data come from?

Understand the business, analyze and trace the source, and also judge the reliability of the data source.

What are the dimensions of ❖ metrics?

Understand the evaluation criteria, different businesses have different key business indicators, use mind map to accumulate the index system of related business, summarize and ask why; the index system is often used to subdivide the data to find the cause, know the data composition can split the data more quickly and find the cause of the anomaly.

How does the ❖ data describe the business?

The application of indicators in the business, what is the normal level of business data, and how the data affected by holidays or event marketing should be compared more, combined with the year-on-year comparison to understand the significance of the data level, and so on.

⒊ gets the data and can quickly sort out the analysis framework and draw conclusions according to the analysis objectives.

If you are the head of operations of Baidu takeout, the number of mobile orders one day is 5% lower than that of the previous day, your boss asks you to give a reasonable explanation for this change, how do you respond?

Secondly, it is clear that the abnormal degree and impact of the change in the index, the order volume dropped by 5% compared with the previous day, whether it is so large that we must pay attention to it.

According to the data shown in the Baidu takeout round B financing plan in 2015, it has 30 million registered users, the number of daily orders exceeds 1.1 million, and the customer unit price can reach about 50 yuan. According to this data, assuming that the number of registered users reaches 60 million by the end of 2016, the number of daily order data exceeds 2 million, and the customer unit price remains basically unchanged, then a 5 per cent decline in order volume means a loss of 5 million revenue on the day. It is almost impossible for a natural single emergency to cause such a big loss, so it is enough to attract the attention of the team (here is just to take the payment data of the financing plan as an example, in fact, as the head of operations, these data are directly accessible internally).

After determining the need to attract attention, we need to seek the perspective of data analysis and investigate the causes of anomalies. We take the game industry as an example to analyze the possible reasons that affect abnormal changes in data. The mode of thinking here is actually the thinking mode of pyramid structure:

First consider the overall indicators, including the number of new users in a certain period of time, the overall payment rate, the total retention rate, user activity, the total conversion rate of each link, the utilization rate of search function, page turning rate, collapse rate and so on. Global metrics are used to analyze common causes that affect all users, and most of the problems are reflected in global metrics.

Then look at the sub-channel indicators: you can observe whether there are anomalies in different channels according to different user attributes (old and new users), user sources (download channels), user natural attributes (region, gender), network environment (network operators, network access methods) and other dimensions.

On the basis of the above two indicators, consider the user behavior data: focus on observing the behavior of users in different time periods and different demand types, so as to locate the data anomalies caused by the directional changes of a subdivided population.

Time factor: the impact of the external environment may also have an impact on product data, so it is important to observe month-on-month and year-on-year data.

Typical such as the "month-end effect", that is, a certain size of user groups due to the end of the month traffic depletion and reduce the Internet behavior, resulting in a decline in overall traffic. In addition, for a takeout product. Weather changes can also cause fluctuations in data, and orders are usually higher in cloudy and rainy days.

Similarly, "Monday effect" and "winter and summer vacation effect" are also common effects in the game industry. Game dau tends to go lower on Monday and higher in winter and summer vacation.

Other product line monitoring: changes in other product lines of Baidu Group may also be the reasons for the decline in order volume, for example, the 91 app market has changed the display location of App ads, or the algorithm adjustment of search engines has reduced the weight of common keywords used by netizens. (through the sub-channel data of download sources, you can obviously see which download sources have reduced data.)

Public opinion monitoring: including but not limited to collecting real-time public opinions on products by manual or machine means, from internal feedback channels such as customer service system to forums, post bars, Weibo, moments, etc. As a result, it is very likely to find special public opinions that lead to a sudden drop or surge in product data, such as what actions have been taken by competitors, major mistakes in the operation of the parent company, etc.

After locating the specific problems and causes, give the corresponding conclusions and solutions, such as repairing a certain bug, making the same strength discount counterattack against the competitor's marketing strategy, and so on.

Analyzing the cause of the problem is only the first step, and the most important thing is to propose a solution to the problem.

Is it helpful for you to read the above content? If you want to know more about the relevant knowledge or read more related articles, please follow the industry information channel, thank you for your support.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.