You probably know there are eight models for data analysis, but do you know what they are specifically? How should we analyze and construct it? The author of this article analyzes and summarizes the eight major data analysis models, which can solve your doubts. I hope you can gain something after reading this article.

1. User Model

“It’s important to know not only what the user is thinking at the moment, but also what the user is thinking behind the scenes and what the user is going through.”

Traditional user model construction method:

• User model: It is established based on research results such as interviews and observations of users. It is rigorous and reliable but time-consuming.
• Temporary user model: It is built based on the understanding of users by industry experts or market research data. It is fast but easily biased. (In case of lack of time and resources)

In order to save time and reduce risks, product teams often push products to users as quickly as possible and conduct rapid trial and error. How do they construct a user model in this scenario?

1. Organize and collect any user-recognizable experience and data that has been obtained, map this information into user description information (attributes) or user behavior information, and store it to form a user profile;
2. Pay attention to fluctuations in your own data in real time and take timely actions;
3. Record user behavior data instead of simply labeling users;
4. 360° user profile covering the entire user life cycle.

Every step of user growth is recorded through behavior. Based on the different stages of the user's life cycle, targeted operational strategies such as attracting new users, conversion, and retention are adopted for new users, lost users, active users, and silent users.

2. Event Model

1. What is an event?

It refers to the user's behavior on the product. It is a professional description of the user's behavior. All program feedback obtained by the user on the product can be abstracted into events and collected by developers through tracking points.

In simple terms, it means: putting a piece of code into the corresponding page/button. When the user enters the page/clicks the button, the essence is to load the code behind it and load the event collection code at the same time, so that it will be recorded by the SDK.

2. Event Collection

Event: User behavior on the product

Attributes: Dimensions that describe an event

Value: The content of the attribute

Collection timing: user click, web page loading completion, server judgment return, etc. When designing the tracking point requirement document, the description of the collection timing is particularly important and is also the core of ensuring data accuracy.

For example: If there is no clear timing during the collection process, when the user clicks the registration button, the user may not actually register successfully because he entered incorrect registration information, but the data may still be recorded, which will not be accurate when counting successful registration events.

The correct description of the collection timing should be "the server returns the judgment of successful registration". (The Japanese official website collects the return activation success or failure page)

3. Analysis of the incident

Number of people: how many people triggered a certain event (behavior);

Number of times: how many times a certain event (behavior) is triggered;

Average number of times per person: average number of times a certain event (behavior) is triggered;

Active ratio: The ratio of the number of people who trigger an event in a time interval to the total number of active people in the current time interval.

4. Incident Management

When there are many events, they can be managed by category. At the same time, important user behaviors can be marked from a product business perspective so that common and important events can be found easily and quickly during analysis.

3. Funnel Model

The funnel model helps you analyze conversions and loss at each step in a multi-step process.

For example, the complete process for a user to download a product might include the following steps:

We can set the above process as a funnel to analyze the overall conversion situation, as well as the specific conversion rate and median conversion time of each step.

We need to monitor users who follow the process at each conversion level and look for optimization points at each level; for users who do not follow the process, we need to draw their conversion paths and find room to improve user experience and shorten the path.

Better use of the funnel model:

• Refine each link and display it down to the click? Between click and download? Between downloading and installing? Between installation and experience?
• Only with the awareness of embedding points and the overall concept can we effectively collect data, make decision-making basis for the optimization of the funnel in each link, and promote the optimization of each department.

4. Heatmap Analysis Model

1. What is a heatmap analysis model?

It reflects where users focus on the web page, especially for the official website homepage, where the information density is extremely high, and shows how users click and browse.

According to the calculation dimension, heat maps can be divided into click heat maps and browse heat maps:

1) Click on the heat map

It tracks the mouse clicks, counts the number of people and times, and distributes the heat based on percentages. Click heat maps are divided into two types: one is all mouse clicks, and the other is clicks on clickable elements on the page. The former can track clicks on all clickable and non-clickable locations on the page, while the latter only tracks clicks on clickable elements on the page.

2) Browse the heat map

Also known as an attention heat map, it records the percentage of time users spend on different pages or different locations on the same page, based on the length of time they stay.

2. New features in the heatmap analysis model

1) Analysis and comparison of specific groups of people

For example, for financial products, the focus of investing users and non-investing users is definitely different.

2) Focus on analysis

• Click-through rate = number of clicks / number of views of the current page
• Focus rate = number of clicks / total number of clicks on the current page

3. Application scenarios

• Landing page effect analysis
• Home page traffic tracking
• Key page experience measurement (product experience and download page)
• Custom retention analysis model

5. Custom Retention Analysis Model

1. Retention Definition and Formula

1) Definition: whether users who meet certain conditions have made return visits at a certain point in time

2) Formula: If the number of users who meet a certain condition is n, and the number of users who make return visits at a certain point in time is m, then the retention rate at that point in time is m/n

2. Three retention methods

1) N-day retention: that is, the number of days of retention, only the users who completed the return visit on the Nth day are counted

2) Unbounded retention (retention within N days): Retention will accumulate and calculate all users who have completed return visits within N days.

3) Bracket retention (customized observation period retention): N-day retention and Unbounded retention are both calculated based on independent days/weeks/months as observation units. However, sometimes we do not want to be restricted to this fixed time measurement and want to divide it into several observation periods:

• First observation period: next day
• Second observation period: Day 3-Day 7
• The third observation period: Day 8-Day 14
• Fourth observation period: 15th to 30th day

3. Customized retention

The above three retention methods are all limited in time, and the definition of retention is that the user opens the APP or enters the website.

Custom retention is based on the retention situation in the business scenario. For example, reading products will define users who have read at least one article as truly retained users, and e-commerce products will define users who have viewed product details at least once as effective retention.

1) Initial behavior

Initial visit and return visit are relative concepts.

2) Return visit behavior

It has an AND relationship with the setting of the initial behavior. The user's initial behavior can be understood as the previous behavior, and the return behavior can be understood as the next behavior.

The setting of initial behavior and return behavior is essentially to further screen the user base. In a growth sharing session of Didi, it was mentioned that "the daily retention of users who grabbed red envelopes and later took a taxi", that is, the initial behavior was grabbing the red envelope, and the return behavior was taking a taxi.

"3-day retention of users who grabbed red envelopes and then took a taxi" - that is, the initial behavior was grabbing the red envelope, and the follow-up behavior was taking a taxi. Look at the third-day retention of this group of people.

VI. Viscosity Analysis

1. Definition

Analysis of active users’ product usage habits, such as how many days a month they use the product, and how many users use the product for more than one day or more than seven days.

For example, when some products launch new features, users need to sign in to use them. This can be used to analyze user usage habits and evaluate the attractiveness and healthiness of the new features.

2. Function

Use retention analysis to understand how well products and features retain users, which features users like, and the differences in how different users use the same feature. This helps to scientifically evaluate products and develop retention strategies.

3. Examples

Stock APP, the number of times the function [Check Stock Market] is triggered by invested and non-invested users.

7. Full Behavior Path Analysis

Behavioral path analysis is divided into: funnel analysis and full behavioral path analysis.

1) Different from the funnel analysis model, the funnel analysis model analyzes the conversion of established behaviors. For example, for e-commerce products, it analyzes the conversion rate of each step from viewing product details to final payment.

2) Full behavior path analysis is to analyze the flow of users in each module of the APP or website, to explore the user's access pattern, and thus optimize the product or website.

It can usually be represented by a tree diagram, as shown in the figure below. For an online training website, users will mostly open the search for courses, so the search for courses needs to be optimized.

After searching for courses for the first time, the user did not find the courses he wanted and conducted a second search. Therefore, the keywords that users search frequently can be set as clickable elements, linking to related courses that users use frequently, guiding users to click to get the desired results.

8. User Segmentation Model

Clustering is the division and grouping of users with certain characteristics, while stratification is more of a management method for all users. In fact, we have been using the method of segmenting users, such as the RFM model we are familiar with:

1. The RFM model extracts three feature dimensions from the user's business data:

1) Recency

2) Consumption frequency

3) Monetary

Through these three dimensions, users can be effectively segmented into 8 groups with different user values ​​and response strategies, as shown in the following figure:

2. Four other dimensions for user segmentation

1) User attributes: objective attributes of users, labels that describe the real demographic attributes of users, such as age, gender, city, browser version, system version, operating version, channel source, etc.

2) Active time

3) Yes, No

4) Added on: When are the most new users?

Author: Li Qifang

Source: Data analysis is not a big deal