Shulou(Shulou.com)06/03 Report--

In a narrow sense, user behavior analysis is user behavior data analysis, but in a broad sense, this word includes user analysis, user behavior result analysis, user behavior analysis. The result of user behavior is different from the analysis of user behavior, one is the result, the other is the process. At present, the products of user behavior analysis in the domestic market are divided into user behavior analysis based on foreground data and user behavior analysis based on background data. User behavior analysis based on foreground technology focuses on user behavior analysis, while user behavior analysis based on background technology focuses on the result analysis of user behavior. These two types of products can be said to have a certain degree of one-sidedness, the completion of only part of the user behavior analysis. Based on this situation, let's talk about how to do a comprehensive user behavior analysis. This article will be mainly from the perspective of data sources and data collection methods.

A user behavior data source

Data is the premise of user behavior analysis, all analysis begins with data, and a skillful wife cannot make bricks without rice. The data sources of user behavior are foreground click data and background business data. Only using the business data in the background will lead to the lack of user behavior in the foreground, which is similar to the effect achieved by doing traditional statistics only through the database. Suppose there are two paths to place an order (denoted by the letter D), A-B-C-E-D and A-B-F-D. So only through the background data we know that a customer has placed an order and how much the order is, so if we want to know which path leads to make it easier for the customer to place an order?

Just using the foreground data, or the above example, although we can identify that the user placed an order from A-B-F-D, we can analyze how valuable the user is brought by this path, whether the user has returned after that, and how much he has spent on our products after that. It is impossible to rely on the foreground data for such a complex analysis.

If you want to analyze what behavior the user results rely on, so as to optimize the product and business to get better results in reverse, on the contrary, you want to measure the effect of the behavior pattern according to the user behavior. It is necessary to connect the front and background data for association analysis. For example, the behavior of high net worth customers with orders greater than 10,000 and their conversion rate must be measured by a combination of foreground data and background data.

How to associate the user data from the foreground and the background? This is a problem that needs to be solved.

Second, user behavior data collection technology

There are mainly two kinds of user behavior data collection techniques: buried point and non-buried point. Let's first popularize the two technologies, and explain how to choose them.

Burying point

The so-called burying point is to add N lines of code to obtain data on top of the original complex code logic for the purpose of data analysis. For example, if you want to get the number of clicks on an item, you have to collect the clicked product data in the click event and send out the data containing the product name and click event ({productname,clicktime}).

Advantages of burying points:

1) the biggest advantage of embedding point is that the data is generated by manual coding, which is more flexible and can better support some extended data.

2) because the buried point is preprocessed according to the logic of the buried point, the later analysis is friendly and the analysis effect is good.

Disadvantages of burying points:

1) the most important prerequisite for burying the site is that the goal must be very clear, that is, what kind of data needs to be collected must be determined in advance. Therefore, the most common problem of burying sites is leakage. Generally speaking, they must undergo careful checksum testing before release, because there is a problem with data collection once the version is released.

2) in the iterative process of the product, if the code ignores the change of the buried point logic during the iteration, which leads to the inaccurate logic of the subsequent analysis, and even leads to the product bug. Even more than for scenarios where the iteration of the product is relatively fast, the burying point is a × ×.

No burial point

Both buried point technology and non-buried point technology need to be changed on the original business code. No burying point is the automatic process of data collection through the characteristics of the programming language itself. For example, the foreground no buried point is actually to collect all the events that occur on the page by listening to JS events. The implementation of no burying point in the background is more complex, but it is very simple to say, in fact, it is to bypass the network data, and the data exchanged between the front and back ends will certainly pass through the network, so the network should contain the vast majority of business data.

Advantages of no burying point:

The main results are as follows: 1) compared with the benefits brought by the burying point mode, it happens to be the problem that the burying point is easy to produce. Because the full amount of data is collected, there is no need to pay attention to the buried point logic in the iterative process of the product, and there will be no phenomena such as missed burying or false burying.

2) No burying point method can greatly reduce the trial and error cost of operation and product because it collects all the data. The possibility of trial and error is high, and it can bring more enlightening information.

3) the last point, and the clearest point, is to reduce the communication costs caused by personnel mobility.

The defect of no burial point is also some doubt that there is no burial point:

1) it is suitable for most common scenarios, and a small number of scenarios that need to be buried cannot be covered.

2) collect all the data without burying point to increase the pressure on the data transmission and the server

According to the previous popular science about burial points and non-burial points, we all understand that both methods have their own advantages and disadvantages. There are also many articles on these two discussion points on Zhihu and other technology blogs, some of which are buried and some are criticizing non-burial points. With regard to technology, let's look at it rationally. the two are not life-and-death relations. what we have learned from our research is that there is no perfect solution to the problem without burying points. however, we are committed to solving the problem of burying points in a general way as much as possible, and minimize burying code. The less code is buried, the less likely it is to make mistakes. We choose to use the combination of foreground non-buried point and background non-buried point technology to obtain user data.

Third, how to use non-buried points to get through the front and background data.

The current analysis methods need to carry out a large number of buried points and associations in order to connect the foreground and background data. is it possible to automatically associate the foreground and background data through non-buried points?

There are no disadvantages at the front desk.

The front-end no-buried point is very appropriate for the relatively simple requirements of data analysis. if you only analyze the behaviors such as page jump, conversion rate, or control clicks, the front-end no-buried point is fully competent. However, if you want to do fine-grained analysis, it may be difficult to rely on the front end without burying points. The most critical reason is that no matter how advanced the foreground collection technology is, it is still difficult to collect some data, such as the amount of orders issued by customers. Many of these data are transmitted to the foreground in the form of html. In different technical implementations, the amount data displayed in the foreground may only be displayed data, and will not be transmitted to the background. Therefore, if you want to do some behavior according to the customer unit price, it is difficult to rely on the front-end non-buried point technology to achieve.

The introduction of no burying point in the background

In a typical foreground and background interactive program, the most common source of data is found to be wired data. Wired data should contain more than 90% fine-grained business data, but these data are stored in wired data in different formats. The difficulty is to design a data processing scheme that can generally use the data existing in wired data, which is the key to the implementation of non-embedding points in the background.

The foreground and background data can be connected automatically.

The connection point with no burying point in the foreground and background is cookie, which embeds special sessionid and uid in the user's request. Background data collection wired data, such as user information, through sessionid and uid association, you can know which user operates a certain order business and what kind of behavior it has on the page. In this way, we realize the automatic association of the front and back data.

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.