User Operation | How to do user behavior path analysis?

User behavior path analysis is a type of data analysis method unique to the Internet industry. It mainly analyzes the flow patterns and characteristics of users in each module of the App or website based on the click behavior logs of each user in the App or website, and explores the user's access or click patterns, thereby achieving some specific business purposes, such as improving the reach of App core modules, extracting mainstream paths and characterizing browsing features for specific user groups, and optimizing and revising App product design.

This article will make some simple discussions on the user behavior path analysis methods, and will focus more on the introduction of some path analysis business scenarios and technical means, in order to serve as a starting point for discussion. Friends who are dedicated to Internet data analysis are welcome to give us comments and criticisms.

1. Path Analysis Business Scenario

An important ultimate goal of user behavior path analysis is to optimize and improve the conversion rate of key modules, so that users can easily reach the core modules according to the expected mainstream path of product design. Specifically, there are the following application scenarios in the analysis process:

1. Identification of typical user paths and analysis of user characteristics

User feature analysis often uses demographic data such as gender and region or operational data such as order price and number of orders. User access path data opens another door for us to understand user features. For example, for an application for making, uploading and sharing pictures, we can use the users' App usage data to divide them into creative users who like to make and upload pictures, interactive users who like to like and comment, lurking users who browse pictures silently, and consumer users who never upload pictures and only download them.

2. Optimization and improvement of product design

Path analysis is of great help in optimizing and improving product design. It can be used to monitor and optimize the conversion rate of each module in the expected user path, and can also discover some unpopular functional points. In a video creation and sharing App, users often perform a series of editing operations from the beginning of video shooting to the final release of the video. Through path analysis, we can clearly see which editing tools are familiar and favorite to users, and which operations are too lengthy and cumbersome. This can help us improve the editing operation module in a targeted manner and optimize the user experience. If the number of user creations during the path analysis process is closely related to the user's likes, comments, and sharing behaviors, you can consider enhancing the social nature of this app and increasing user stickiness and desire to create.

3. Monitoring of product operation process

The conversion rate of key product modules is itself a very important product operation indicator. Monitoring and verifying the corresponding operation activity results through path analysis can help relevant personnel understand the effectiveness of the operation activities.

2. Path Analysis Data Acquisition

The Internet industry has a unique advantage in acquiring data, and the data that path analysis relies on is mainly log data in the server. Every step a user takes while using the App can be recorded. At this time, what we need to focus on is an excellent deployment strategy, which should be closely related to the business we care about. Here I would like to recommend Zhuge io, a refined operational analysis tool based on user insights; by integrating Zhuge io's SDK into the App or website, you can obtain all user behavior data within the application. In fact, Zhuge.io believes that not all events have the same value in every App. Based on the need for in-depth analysis of core events, Zhuge.io recommends that you use a hierarchical custom event layout method. Each event consists of three levels: event, attribute (Key) and attribute value (Value). At the same time, Zhuge.io also provides data monitoring and deployment consulting services for developers, and can provide customers with personalized event deployment consulting and technical support based on rich industry experience.

3. The relationship between the funnel model and path analysis

The path analysis mentioned above is similar to the funnel model we are more familiar with. In a broad sense, the funnel model can be regarded as a special case of path analysis, which is a path analysis of specific modules and event nodes for a small number of people.

The funnel model usually describes the conversion rate of users at a series of key nodes in a website or app. These key nodes are often manually specified by us. For example, we can see the funnel conversion of the purchase behavior of a shopping app in Zhuge io. On this shopping app platform, buyers go through four key nodes from browsing to successful payment: product browsing, adding to shopping cart, checkout, and successful payment. From step 1 to step 4, fewer and fewer people go through these key nodes, and the conversion rate of the nodes presents a funnel-shaped situation. We can monitor and manage the conversion efficiency, operational effects and processes of each link, and conduct targeted in-depth analysis and improvement of links with low conversion rates. Other funnel model analysis scenarios can be used flexibly according to business needs. Zhuge io platform has a very powerful funnel analysis tool, which is a platform for you to give full play to your imagination of data. Welcome to refer to an analysis case based on the funnel model "New Funnel/Retention Method".

Path analysis is different from the funnel model. It usually tracks and records every behavioral path of each user, and analyzes and explores the user path behavior characteristics on this basis, involving the source and destination of each step, and the conversion rate of each step. It can be said that the funnel model is a pre-set, artificial, and proactive path of several key event nodes, while path analysis is an exploratory way to explore the overall behavior path, find out the mainstream path of users, and may even discover some interesting pattern paths that were not known in advance. From a technical point of view, the funnel model simply and intuitively calculates and displays the relevant conversion rates, while path analysis involves some more extensive aspects.

4. Common ideas and methods for path analysis

1. Simple traversal statistics and visual analysis exploration

By analyzing the user behavior path data obtained from the layout, we can count the event path click stream data of each user in the simplest and most direct way, and present it intuitively using data visualization methods. D3.js is one of the most popular data visualization libraries. We can use SunburstPartition in it to characterize the event path click status of user groups. Starting from the center of the circle and moving outward layer by layer, it represents the entire behavior statistics of users from the beginning of using the product to leaving; the sunburst event path diagram can quickly locate the user's mainstream usage path. By extracting path data between specific groups of people or specific modules and analyzing them using the sunburst event path diagram, deeper problems can be located. Flexible use of sunburst path statistics chart is a magic weapon for us in path analysis.

Zhuge.io can not only easily obtain the distribution data, but also provide customers with personalized sunburst event path map analysis, and can produce customized product analysis reports for customers' products.

2. Sequence path mining method based on association analysis

When talking about association rule analysis, we can’t avoid the classic case of “beer and diapers” in data mining. Regardless of whether the "beer and diapers" case is a "myth" made up by a Teradata manager, this case, to a certain extent, allows people to understand the process of shopping basket analysis (association analysis) and the business value behind it. All the goods purchased by each customer in the supermarket at one time are regarded as a shopping basket. The purchase behavior data stored in the database are analyzed using association rule algorithms, namely shopping basket analysis. It was found that 10% of customers bought diapers and beer at the same time, and among all customers who bought diapers, 70% of them also bought beer at the same time. So the supermarket decided to display beer and diapers together, which resulted in a significant increase in sales.

We might as well regard all the event points operated by each user every time they use the App as "a series of items" in a "shopping basket". Unlike the shopping basket mentioned above, all the event click behaviors here have a strict order of events. We can improve the Apriori or FP-Growth algorithm in the association rules so that it can mine frequent user behavior paths with a strict sequence, which is an important idea for user path analysis. We can carefully consider the product business logic reflected in the discovered rule sequence paths, and we can also compare and analyze the rule sequence paths between different user groups.

3. Social Network Analysis (or Link Analysis)

Early search engines were mainly based on retrieving the similarity between web page content and user queries or finding relevant web pages for users by looking up pages indexed in the search engine. With the explosive growth in the number of Internet web pages in the mid-to-late 1990s, early strategies were no longer effective and it was impossible to give reasonable ranked search results for a large number of similar web pages. Today's search engine giants such as Google and Baidu have adopted search engine algorithms based on link analysis as one of the solutions to this problem. Web pages are connected to each other through hyperlinks, just like social networks on Weibo are connected through following behaviors. There are well-known and influential big Vs in social networks, and there are also important or authoritative web pages on the Internet. Providing more authoritative web pages to the front of search engine results will make the search more effective.

We regard people in social networks as nodes, web pages on the Internet as nodes, and even each module event in our App products as nodes. Nodes are connected in their own ways to form a specific network graph. The analysis methods based on these network structures are collectively referred to as social network analysis.

There are some common analysis methods in social network analysis that can be applied to our path analysis, such as node centrality analysis, node influence modeling, community discovery, etc. Through centrality analysis, we can explore which module events are in a central position, or serve as a hub connecting two major types of module events, or become the final destination of most module events. Through community discovery, we can explore whether there are some "small circles" in this social network, that is, a small part of the behavioral path that users always like to operate, and this part of the path is relatively independent of most other modules.

Related reading:

1. User operation: new funnel model for conversion analysis!

2. User operation: How to use B-side operation thinking to increase user growth?

3. Product operation: How to use data analysis to drive product user growth?

4. APP user growth: One model solves 90% of growth problems!

5.How to increase users? Take Pinduoduo and Xiaohongshu as examples

6. Triggering user growth: Is user operation just about attracting new users?

7. User operation: What else can you do to attract new users without fission users?

8. User operation: how can financial products awaken dormant users?

Author: Zhuge io

Source: Zhuge io Data Coach